mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Hardware (https://www.mersenneforum.org/forumdisplay.php?f=9)
-   -   RDNA2 / Big Navi (https://www.mersenneforum.org/showthread.php?t=26134)

M344587487 2020-10-28 16:59

RDNA2 / Big Navi
 
AMD announcement: [URL]https://www.youtube.com/watch?v=oHpgu-cTjyM[/URL]


[url]https://www.youtube.com/watch?v=haAPtu06eYI[/url]



tl;dr:
[LIST][*]Infinity cache - "Graphics optimised, high density, high speed cache, based on the Zen L3 cache". graph shows 2.17x bandwidth of infinity cache + 256 bit GDDR6 compared to 384 bit GDDR6 on typical gaming workloads[*]Smart access memory - Zen3 processors can have full access to GPU memory? Showing single digit % performance increase in some games not specifically optimised for it, promises that workloads can specifically optimise for even better performance. Performance figures shown are with this feature enabled so this is one way they've fudged the numbers[/LIST] 6800
[LIST][*]60 CU's[*]1815 MHz game clock[*]2105 MHz boost clock[*]128MB infinity cache[*]16GB GDDR6, 256 bit?[*]250W TBP[*]November 18th[*]$579[/LIST] 6800XT
[LIST][*]72 CU's[*]2015 MHz game clock[*]2250 MHz boost clock[*]128MB infinity cache[*]16GB GDDR6 256 bit[*]300W TBP[*]November 18th[*]$649[*]Shown rivalling 3080 in games[/LIST] 6900XT
[LIST][*]80 CU's[*]2015 MHz game clock[*]2250 MHz boost clock[*]128MB infinity cache[*]16GB GDDR6 256 bit?[*]300W TBP[*]December 8th[*]$999[*]Shown rivalling 3090 in games[/LIST] At a glance it looks promising. Infinity cache has a chance of being insanely good for our purposes but who knows. The high price is also a good sign that they're confident that they're competitive, in gaming. Timely compute and Linux support, unlikely.

firejuggler 2020-10-28 18:05

The 6800 look like a fair competition to the RTX3070 price wise. Now we need more numbers... like FP 64 performance. For our "need".


edit: ok I found it :

[code]
FP64 (double) performance 1,010 GFLOPS (1:16)[/code]
Around 3 time the performance or a rtx 3070 (in FP64)
[url]https://www.techpowerup.com/gpu-specs/radeon-rx-6800.c3713[/url]

Mark Rose 2020-10-28 21:20

Still nowhere near the 3.5 TFL64OPS of the Radeon VII, sadly.

I do wonder how the Navi 10 mining GPU that showed up in recent kernel patches will do. It's Navi 10 though, so probably the same 1/16 FP64 performance.

mackerel 2020-10-28 22:22

The cache is the biggest "new" variable I guess. Be really interesting to see how that works out in non-gaming uses.

I think the 6800 will be positioned above 3070, if you compare their relative performance compared to the 3080 and 6800XT respectively.

moebius 2020-10-28 23:18

[QUOTE=Mark Rose;561376]Still nowhere near the 3.5 TFL64OPS of the Radeon VII, sadly.[/QUOTE]
Even a Vega 64 still has 0.792 Tflops at FP64. Maybe the fast HBM2 memory is missing.

M344587487 2020-10-29 00:37

[QUOTE=Mark Rose;561376]Still nowhere near the 3.5 TFL64OPS of the Radeon VII, sadly.[/QUOTE]Unfortunately that was never on the cards, Navi has always been about optimising for graphics performance over all else. Good DP is now the realm of CDNA2 and there's a 99% chance that CDNA2 is not the price-performance we're looking for.

[QUOTE=Mark Rose;561376]I do wonder how the Navi 10 mining GPU that showed up in recent kernel patches will do. It's Navi 10 though, so probably the same 1/16 FP64 performance.[/QUOTE]Wild guess, it's the exact same as an existing card just with all the extraneous features removed and maybe a bios tuned for efficiency. It could just be a ploy to redefine and push Navi 1 as a mining generation. Navi 2 seems to be of less interest to miners as Ethereum and Monero both require random access to a 2GB file so the infinity cache is scuppered, and the raw bandwidth of Navi 2 is not much higher than Navi 1. It's been a while but last I paid attention these two (and derivatives thereof being pedantic) accounted for the vast majority of GPU mining.

preda 2020-10-29 01:43

[QUOTE=M344587487;561339]Timely compute and Linux support, unlikely.[/QUOTE]

Yep, unfortunately: the hardware is done, time to start work on the software side..
(and by that I don't mean *our* software side, I mean AMD's software side)

axn 2020-10-29 02:05

[QUOTE=Mark Rose;561376]Still nowhere near the 3.5 TFL64OPS of the Radeon VII, sadly.[/QUOTE]

This might matter less than you think, to wit...

[QUOTE=mackerel;561382]The cache is the biggest "new" variable I guess. Be really interesting to see how that works out in non-gaming uses.[/QUOTE]

The 128MB cache is big enough to fit our current FFTs in their entirety. This could indeed be very interesting.

tServo 2020-10-29 12:43

[QUOTE=axn;561412]This might matter less than you think, to wit...



The 128MB cache is big enough to fit our current FFTs in their entirety. This could indeed be very interesting.[/QUOTE]

It probably depends on whether that cache is able to be manipulated and configured directly by the programmer or if management of it is left to some inflexible on chip algorithm.

Mark Rose 2020-11-09 15:47

[url=https://videocardz.com/newz/amd-radeon-rx-6800-allegedly-1-5x-faster-than-geforce-rtx-3090-in-cryptocurrency-mining]AMD Radeon RX 6800 1.5x faster than GeForce RTX 3090 in mining[/url]

If true, it's going to be fun finding one. I don't think it will have higher FP32 performance than the RTX 3xxx, but this bodes well for INT performance. Could be TF monster.

Mark Rose 2020-11-11 19:23

So the above rumour was a bust, but now there are some benchmarks: [url]https://videocardz.com/newz/amd-radeon-rx-6800-xt-and-rx-6800-geekbench-opencl-benchmarks-leaked[/url]

M344587487 2020-11-18 10:37

Release day is today, anyone brave enough to try and get one on day one? As much as I'd like to I just can't justify trying to get one on day 1, judging by how slow Navi 1 support on Linux progressed big navi could be a Linux paperweight for a while. Indications are that they'll sell out quickly and that demand will outstrip supply for months, if you're after one sooner rather than later then chancing today is your best bet.


edit: You may only have a few minutes to try and get an order in, if that.

M344587487 2020-11-18 13:37

Last warning for anyone trying to get one today. It goes live everywhere in 23 minutes, on the hour.

moebius 2020-11-18 13:47

[QUOTE=M344587487;563646]Last warning for anyone trying to get one today. It goes live everywhere in 23 minutes, on the hour.[/QUOTE]
Here is a quick list of German dealers where you can get the Big Navi probably at 3:00 p.m. Central European Time, at least for European users.
[url]https://www.pcgameshardware.de/Radeon-RX-6800-Grafikkarte-276952/News/AMD-Radeon-RX-6800-RX-6800-XT-kaufen-shops-1361986/[/url][URL="https://www.pcgameshardware.de/Radeon-RX-6800-Grafikkarte-276952/News/AMD-Radeon-RX-6800-RX-6800-XT-kaufen-shops-1361986/"]https://www.pcgameshardware.de/Radeon-RX-6800-Grafikkarte-276952/News/AMD-Radeon-RX-6800-RX-6800-XT-kaufen-shops-1361986/[/URL]

moebius 2020-11-18 14:10

Too funny, most of the retailer sites suddenly crashed on time for the appointment, or could not even be accessed.:smile::smile::smile::smile::smile:


I found an offer for AMD Radeon VII(used) at a dealer in Frankfurt, Germany for only 490 euros, 8 pieces available. All statements without guarantee
[URL="https://www.amazon.de/XFX-Radeon-16GB-Boost-1xHDMI/dp/B07NFGDZWQ/ref=sr_1_1?__mk_de_DE=%C3%85M%C3%85%C5%BD%C3%95%C3%91&dchild=1&keywords=radeon+vii&m=A1P0V3DD9VP5YQ&qid=1605709342&s=merchant-items&sr=1-1"]https://www.amazon.de/XFX-Radeon-16GB-Boost-1xHDMI/dp/B07NFGDZWQ/ref=sr_1_1?__mk_de_DE=%C3%85M%C3%85%C5%BD%C3%95%C3%91&dchild=1&keywords=radeon+vii&m=A1P0V3DD9VP5YQ&qid=1605709342&s=merchant-items&sr=1-1[/URL]

moebius 2020-11-18 15:08

[QUOTE=moebius;563652]I found an offer for AMD Radeon VII(used) at a dealer in Frankfurt, Germany for only 490 euros, 8 pieces available.[/QUOTE]
It seems like someone bought all 8, in any case they are no longer available. Damn fast the customers.

Xyzzy 2020-11-18 16:38

Since ROCm doesn't yet support the 5000 series GPUs, why would we throw money at a 6000 series GPU?

:confused:

M344587487 2020-11-18 17:04

[QUOTE=Xyzzy;563674]Since ROCm doesn't yet support the 5000 series GPUs, why would we throw money at a 6000 series GPU?

:confused:[/QUOTE]
My logic exactly, however initial outlook looks good for the 6000 series and the 5000 should be brought up to speed with it: [url]https://www.phoronix.com/vr.php?view=29708[/url]

firejuggler 2020-11-18 17:12

[QUOTE=Xyzzy;563674]Since ROCm doesn't yet support the 5000 series GPUs, why would we throw money at a 6000 series GPU?

:confused:[/QUOTE]
To make Nvidia release a Ti version wich will be better than the 6000 series

Mark Rose 2020-11-18 17:16

128 MB of L3 cache. I bet gpuOwl will benefit from that.

Mark Rose 2020-11-19 01:08

Looks like Radeon VII is still going to be the card to get. Also, the attachment feature of the forum doesn't like svg files, so here is a link:

[url]https://openbenchmarking.org/embed.php?i=2011176-FI-GPUCOMPUT49&sha=723bc7b2db04&p=2[/url]

From here: [url]https://www.phoronix.com/scan.php?page=article&item=amd-rx6800-opencl&num=1[/url]

Viliam Furik 2020-11-19 03:52

[QUOTE=Mark Rose;563719]Looks like Radeon VII is still going to be the card to get.[/QUOTE]

The data from the chart don't tell the whole truth when it comes to gpuOwl. Because of its dependence on memory bandwidth, Radeon VII doesn't reach anywhere near the potential 3,4 TFLOPS. In reality, even with the best cooling and memory and core clocks, it reaches at most 1,5 TFLOPS. My Radeon VII gives about 750 GFLOPS because it's heavily underclocked (1000 mV core voltage, -5% power limit, 1600 MHz core clock, 1000 MHz memory clock - MSI Afterburner settings).

If the gpuOwl could use the Infinity cache effectively, with its ~1,5 TB/s of memory bandwidth, even the RX 6800 could outperform the Radeon VII, but the RX 6900XT would most probably crush it. But that's only if the Infinity cache can be used.

Prime95 2020-11-19 04:03

Apologies for sidetracking the thread ---

[QUOTE=Viliam Furik;563724]My Radeon VII gives about 750 GFLOPS because it's heavily underclocked (1000 mV core voltage, -5% power limit, 1600 MHz core clock, 1000 MHz memory clock - MSI Afterburner settings).[/QUOTE]

Those settings look nothing like my optimized Radeon VII settings. All cards are set to ~1040mV (actual is ~800mV), 1373MHz, memory ~1170MHz. Your core clock is way too high compared to your memory clock for that to be an efficient setting.

Are you using Windows? Use Linux instead -- much better OpenCL support an easy 10+% performance gain.

moebius 2020-11-19 09:53

[QUOTE=Viliam Furik;563724]If the gpuOwl could use the Infinity cache effectively, with its ~1,5 TB/s of memory bandwidth, even the RX 6800 could outperform the Radeon VII, but the RX 6900XT would most probably crush it. But that's only if the Infinity cache can be used.[/QUOTE]Based on the data available to me, my estimate for the 6800 XT is 662 us/it at PRP for the exponent 77936867, which corresponds approximately to a Tesla P-100.
Real Benchmarks are of course highly desirable
[SIZE="1"][URL="https://mersenneforum.org/showpost.php?p=563183&postcount=2603"]https://mersenneforum.org/showpost.php?p=563183&postcount=2603[/URL][/SIZE]

Xyzzy 2020-11-19 18:48

[QUOTE=Mark Rose;563719]Also, the attachment feature of the forum doesn't like svg files, so here is a link:[/QUOTE]This should be fixed now if you want to test it.

The image engine that creates thumbnails says it can't process SVG files but they should be able to be attached.

:mike:

tServo 2020-11-19 20:00

[QUOTE=moebius;563735]Based on the data available to me, my estimate for the 6800 XT is 662 us/it at PRP for the exponent 77936867, which corresponds approximately to a Tesla P-100.
Real Benchmarks are of course highly desirable
[SIZE="1"][URL="https://mersenneforum.org/showpost.php?p=563183&postcount=2603"]https://mersenneforum.org/showpost.php?p=563183&postcount=2603[/URL][/SIZE][/QUOTE]

I applaud your enthusiasm but I don't think the RX 6000 cards are going to be as fast as you think, purely based on the difference in FP64 flops. That mystery cache would have to go a long way to make up for only 1:16 FP64 cores.

Unfortunately, the early reviews will only have gaming focus because, after all, these are cards for gamers and we won't know GIMPS performance until some lucky GIMPster is able to get one of these and posts the results. It seems the launch was pretty much like Nvidia's with everybody sold out in less than 60 seconds.

BTW, I hope you're right as Nvidia could use some compettition ( which means I'm completely dismissing Intel's imminent foray into this market. )

moebius 2020-11-19 20:25

[QUOTE=tServo;563783]I applaud your enthusiasm but I don't think the RX 6000 cards are going to be as fast as you think, purely based on the difference in FP64 flops. That mystery cache would have to go a long way to make up for only 1:16 FP 64 cores.[/QUOTE]
I have based my calculations on the 5700 XT which also has a 1:16 FP 64 cores performance with 0,609 TFLOPs which is amazing at gpuowl. If the NAVI technology doesn't get worse it would be quite possible!

tServo 2020-11-20 03:02

[QUOTE=moebius;563784]I have based my calculations on the 5700 XT which also has a 1:16 FP 64 cores performance with 0,609 TFLOPs which is amazing at gpuowl. If the NAVI technology doesn't get worse it would be quite possible![/QUOTE]

To what results are you referring?

moebius 2020-11-20 04:27

[QUOTE=tServo;563810]To what results are you referring?[/QUOTE]
[URL="https://openbenchmarking.org/embed.php?i=2011176-FI-GPUCOMPUT49&sha=723bc7b2db04&p=2"]https://openbenchmarking.org/embed.php?i=2011176-FI-GPUCOMPUT49&sha=723bc7b2db04&p=2[/URL] and
AMD Radeon RX 5700 XT[CODE]2020-10-29 16:19:40 GpuOwl VERSION v7.1-11-g97cfbd2
2020-10-29 16:19:40 Note: not found 'config.txt'
2020-10-29 16:19:40 config: -iters 200000 -prp 77936867
2020-10-29 16:19:40 device 0, unique id ''
2020-10-29 16:19:40 gfx1010-0 77936867 FFT: 4M 1K:8:256 (18.58 bpw)
2020-10-29 16:19:41 gfx1010-0 77936867 OpenCL args "-DEXP=77936867u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DAMDGPU=1 -DCARRY64=1 -DCARRYM64=1 -DMM_CHAIN=1u -DMM2_CHAIN=2u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0xa.c42d0d7cec038p-5 -DIWEIGHT_STEP_MINUS_1=-0x8.0e50c8817ddf8p-5 -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2020-10-29 16:19:44 gfx1010-0 77936867 OpenCL compilation in 3.53 s
2020-10-29 16:19:44 gfx1010-0 77936867 maxAlloc: 0.0 GB
2020-10-29 16:19:44 gfx1010-0 77936867 You should use -maxAlloc if your GPU has more than 4GB memory. See help '-h'
2020-10-29 16:19:44 gfx1010-0 77936867 P1(0) 0 bits
2020-10-29 16:19:44 gfx1010-0 77936867 PRP starting from beginning
2020-10-29 16:19:45 gfx1010-0 77936867 OK 0 on-load: blockSize 400, 0000000000000003
2020-10-29 16:19:45 gfx1010-0 77936867 validating proof residues for power 8
2020-10-29 16:19:45 gfx1010-0 77936867 Proof using power 8
2020-10-29 16:19:47 gfx1010-0 77936867 OK 800 0.00% 1579c241dc63eca6 1526 us/it + check 0.70s + save 0.10s; ETA 1d 09:02
2020-10-29 16:20:07 gfx1010-0 77936867 10000 0.01% fc4f135f7cf4ad29 2145 us/it
2020-10-29 16:20:22 gfx1010-0 77936867 20000 0.03% 3cd1bd9d5e09cbc5 1523 us/it
2020-10-29 16:20:37 gfx1010-0 77936867 30000 0.04% c4e0ff35e3290d98 1521 us/it
2020-10-29 16:20:52 gfx1010-0 77936867 40000 0.05% dffe1b1b0d748128 1520 us/it
2020-10-29 16:21:08 gfx1010-0 77936867 50000 0.06% 52e286945371ed29 1520 us/it
2020-10-29 16:21:23 gfx1010-0 77936867 60000 0.08% 0945da4dc08bdd95 1519 us/it
2020-10-29 16:21:38 gfx1010-0 77936867 70000 0.09% 7131fa4eb77f4bb2 1519 us/it
2020-10-29 16:21:53 gfx1010-0 77936867 80000 0.10% 8d76071d27ee4221 1521 us/it
2020-10-29 16:22:08 gfx1010-0 77936867 90000 0.12% 0bacff453b2f470e 1519 us/it
2020-10-29 16:22:24 gfx1010-0 77936867 100000 0.13% 6d7296b9e2830f50 1519 us/it
2020-10-29 16:22:39 gfx1010-0 77936867 110000 0.14% 8cbfd4435622bda7 1519 us/it
2020-10-29 16:22:54 gfx1010-0 77936867 120000 0.15% 79ae5dad855057ad 1520 us/it
2020-10-29 16:23:09 gfx1010-0 77936867 130000 0.17% 50c97bcbf876231f 1520 us/it
2020-10-29 16:23:24 gfx1010-0 77936867 140000 0.18% e1db15f897271496 1520 us/it
2020-10-29 16:23:40 gfx1010-0 77936867 150000 0.19% 127631386c6a9b17 1520 us/it
2020-10-29 16:23:55 gfx1010-0 77936867 160000 0.21% 25b7b6206fc6f085 1519 us/it
2020-10-29 16:24:10 gfx1010-0 77936867 170000 0.22% 416816b0d9f4bba8 1520 us/it
2020-10-29 16:24:25 gfx1010-0 77936867 180000 0.23% 6bee5d054f770861 1521 us/it
2020-10-29 16:24:40 gfx1010-0 77936867 190000 0.24% f37f068f014b18a0 1520 us/it
2020-10-29 16:24:56 gfx1010-0 77936867 Stopping, please wait..
2020-10-29 16:24:56 gfx1010-0 77936867 OK 200000 0.26% f0b04b45b0855bd2 1520 us/it + check 0.67s + save 0.09s; ETA 1d 08:49
2020-10-29 16:24:56 gfx1010-0 Exiting because "stop requested"
2020-10-29 16:24:56 gfx1010-0 Bye[/CODE]

M344587487 2020-11-20 14:42

As Linux support is reportedly looking good I'll try and get one but don't have any expectation of actually getting one. AIB models are released on the 25th which under normal circumstances you'd expect to cost some tens of dollars more than a standard card to pay for the customness, however AIB's could easily raise prices or more likely only release their top-end in numbers because they'll sell out of whatever they offer. I suggest that everyone who is even on the fence about getting one trying to get one on the 25th if you can, it increases the odds of getting benchmarks in hand and isn't much risk considering it can be returned or resold easily if it isn't a good fit.

[QUOTE=moebius;563735]Based on the data available to me, my estimate for the 6800 XT is 662 us/it at PRP for the exponent 77936867, which corresponds approximately to a Tesla P-100.
Real Benchmarks are of course highly desirable
[URL]https://mersenneforum.org/showpost.php?p=563183&postcount=2603[/URL][/QUOTE]Assuming perfect scaling with DP we get to 774 us/it, even that I think is pushing the boundaries of optimism as it assumes that Infinity Cache can perfectly make up the raw bandwidth deficit. Your estimate must be assuming that bandwidth not DP is the main limiter, and that IC is a miracle from the chip gods. I hope you're right but I'd rather be pessimistic and proven wrong than optimistic and proven wrong, YMMV.

moebius 2020-11-20 16:46

[QUOTE=M344587487;563836]Assuming perfect scaling with DP we get to 774 us/it, even that I think is pushing the boundaries of optimism as it assumes that Infinity Cache can perfectly make up the raw bandwidth deficit.[/QUOTE]In view of the fact that I was calculating with the (possible OC) values ​​at openbenchmarking.org, your gpuowl-value could of course also fit. The double precision values ​​there are slightly higher than the manufacturer's specifications.

tServo 2020-11-20 16:59

I finally found a reviewer that bothered to include compute stats in their results and the results are just so-so. The 6800xt performs even with the rtx 2080 ti and rtx 3070 but far behind the rtx 3080 and 3090.

[URL="https://www.legitreviews.com/amd-radeon-rx-6800-xt-and-radeon-rx-6800-review_223774/13"]https://www.legitreviews.com/amd-radeon-rx-6800-xt-and-radeon-rx-6800-review_223774/13[/URL]

moebius 2020-11-20 17:17

[QUOTE=tServo;563844]The 6800xt performs even with the rtx 2080 ti and rtx 3070 but far behind the rtx 3080 and 3090.[/QUOTE]
Do you work at Nvidia? This may even be true in certain computer games or maybe TF, but i'm more concerned with PRP performance and there are e.g. GA102 and Navi 10 XT graphics cards just on par.

xx005fs 2020-11-20 19:43

[QUOTE=tServo;563844]I finally found a reviewer that bothered to include compute stats in their results and the results are just so-so. The 6800xt performs even with the rtx 2080 ti and rtx 3070 but far behind the rtx 3080 and 3090.

[URL]https://www.legitreviews.com/amd-radeon-rx-6800-xt-and-radeon-rx-6800-review_223774/13[/URL][/QUOTE]


Sadly they used aggregated result, if only they listed the specific results for each scientific compute category (like GEMM, FFT, and N-Body) because it is well known that Ampere has good FP32 performance and have tensor core acceleration for GEMM. They also used FP32 in this case so it won't be representative of PRP.

tServo 2020-11-20 20:36

[QUOTE=moebius;563848]Do you work at Nvidia? This may even be true in certain computer games or maybe TF, but i'm more concerned with PRP performance and there are e.g. GA102 and Navi 10 XT graphics cards just on par.[/QUOTE]

No I do not nor have I ever work for Nvidia. I am retired. Nor am I an Nvidia fanboy because besides their cards, I have several Radeon Viis that run GPUOWL
My opinions were based on past performance of similar cards.

Of course "the proof is in the pudding" and we must wait for a GIMPster to enlighten us and anything before then is conjecture.
I really do hope the Big Navi cards come thru because Radeon VIIs are almost impossible to get, at least at a reasonable price, and it would be great to have another option.
I suspect that there will not be any card from Nvidia that will be good at running PRPs; even a high priced one like the Titan V.

moebius 2020-11-20 21:04

[QUOTE=tServo;563872] I am retired. Nor am I an Nvidia fanboy because besides their cards, I have several Radeon Viis that run GPUOWL [/Quote]
I'm also not a fan of any particular manufacturer. The scientific cards from Nvidia: A100, Tesla V100 and even the Titan V (at least with one instance running) are faster with gpuowl than the Radeon VII. Unfortunately, they cannot fully utilize their potential FP64 (DP) performance. No offense, I am creating the timing list for gpuOwl, that I have already posted in this thread, so that the gpuowl users have an approximate indication of which card is suitable for their budget.

Mark Rose 2020-11-20 21:23

I'm hoping AMD released a cut-down CDMA card as a Radeon VII replacement.

tServo 2020-11-21 00:51

[QUOTE=Mark Rose;563878]I'm hoping AMD released a cut-down CDMA card as a Radeon VII replacement.[/QUOTE]

That would be great, but wouldn't it directly compete with the Radeon Pro Vii ?

Mark Rose 2020-11-21 03:46

[QUOTE=tServo;563899]That would be great, but wouldn't it directly compete with the Radeon Pro Vii ?[/QUOTE]

The Radeon VII was discontinued in June 2019. They only made them for 5 months.

But surely AMD has seen their prices in the used market, so there's definitely demand for a replacement.

Xyzzy 2020-11-21 04:00

He is talking about the "pro" version.

[url]https://www.amd.com/en/products/professional-graphics/radeon-pro-vii[/url]

axn 2020-11-21 04:04

[QUOTE=Mark Rose;563905]The Radeon VII was discontinued in June 2019. They only made them for 5 months.

But surely AMD has seen their prices in the used market, so there's definitely demand for a replacement.[/QUOTE]

Radeon VII is not the same as Radeon Pro VII. The former was marketed as gaming card (LOL). The latter is workstation -- I don't think that one is meant to be a limited run.

CDNA, by definition, will not come to consumer space; at best, we can hope to see workstation cards (like replacement for R Pro VII)

Mark Rose 2020-11-21 05:55

I missed the Pro part :cmd:

moebius 2020-11-22 18:02

[QUOTE=tServo;563899]That would be great, but wouldn't it directly compete with the Radeon Pro Vii ?[/QUOTE] A question. Don't you think of proofs, are you using an older version of gpuowl? Saves a lot of computing time if we no longer have to do conventional double checks. I use gpuowl version v6.11-380-g79ea0cc on Ubuntu.

tServo 2020-11-23 01:50

[QUOTE=moebius;564029]A question. Don't you think of proofs, are you using an older version of gpuowl? Saves a lot of computing time if we no longer have to do conventional double checks. I use gpuowl version v6.11-380-g79ea0cc on Ubuntu.[/QUOTE]

I see that I was not precise in my terminology wrt 'compete'. I meant competition in the economic sense in that the Radeon Pro VII might be too close in performance to the card that Mark Rose proposed, the "cut-down CDMA card". This would confuse the marketplace if AMD had 2 cards that were quite close in specs.

I realize what proofs are for. What I'm unsure of is whether they are solving a problem that simply doesn't exist. With the advent of GEC, we may have a virtually error free primality test. I would be interested to know how many PRP tests that used GEC were found to be bad.
There may not have been enough double checking of these PRPs to know yet. Is PRP unfairly suffering from the same dark cloud of bad residues that hung over LL tests?
Or is the concern that some moron might just submit a bunch of fabricated completed PRP results with random residues?

PhilF 2020-11-23 02:03

[QUOTE=tServo;564068]With the advent of GEC, we may have a virtually error free primality test.

Or is the concern that some moron might just submit a bunch of fabricated completed PRP results with random residues?[/QUOTE]

Either. You said it yourself -- "virtually error free". But this project's goal is to be guaranteed error free. So GEC would have to get promoted to GGEC (Guaranteed Gerbicz Error Check) before the proof certifications could be dropped.

Prime95 2020-11-23 06:53

[QUOTE=tServo;564068]
Or is the concern that some moron might just submit a bunch of fabricated completed PRP results with random residues?[/QUOTE]

This is the much bigger concern that proofs solve.

M344587487 2020-11-25 11:35

PSA: AIB release day is today, for anyone wanting a a big navi card for a reasonable price in the next few months today is probably your best bet. If the initial release last week is anything to go by then every version will go live at 2PM UK time, 2 hours 25 minutes from this post on the hour. They'll go out of stock in minutes most likely. Good luck.

moebius 2020-11-25 12:22

[QUOTE=M344587487;564320]Good luck.[/QUOTE]"Verfügbarkeitsbot" for RX and RTX GPU'S [B]only for good old Germany[/B] (If it finds an offer in Shop/API, it will be posted in the thread) [URL="https://www.hardwareluxx.de/community/threads/rtx-und-rx-gpu-verf%C3%BCgbarkeitshinweise.1281755/"] https://www.hardwareluxx.de/community/threads/rtx-und-rx-gpu-verf%C3%BCgbarkeitshinweise.1281755/[/URL][COLOR="Red"] Go to last page[/COLOR]

moebius 2020-11-25 14:27

From Alternate.de a 6800 XT just appeared 5 minutes ago at the bot-list but not (anymore) or not yet available.

M344587487 2020-11-25 15:08

The UK never had any stock, the same appears to be true for most of Europe. The few listings that are up show some laughable pricing. If they can't stock anything today they're not going to be able to do timely restocks in the coming months, this is an actual paper launch. Probably best to ignore this gen until at least February.

moebius 2020-11-27 11:48

[QUOTE=M344587487;564340] If they can't stock anything today they're not going to be able to do timely restocks in the coming months, this is an actual paper launch. Probably best to ignore this gen until at least February.[/QUOTE]To see the bot thread, you now have to register in the hardwareluxx.de forum.There are still a few new offers for the RTX cards, but not a single one for the AMD RX 6800 (XT) .If someone buys a Nvidia RTX 3070 here, please don't forget to post a gpuowl benchmark with 77936867 in the thread below.

[URL="https://mersenneforum.org/showthread.php?p=564285#post564285"]https://mersenneforum.org/showthread.php?p=564285#post564285[/URL]

There is now a separate bot thread for AMD RX graphics cards, apparently there was one this morning for a short time.

[URL="https://www.hardwareluxx.de/community/threads/rx-gpu-verf%C3%BCgbarkeitshinweise.1282621/"]https://www.hardwareluxx.de/community/threads/rx-gpu-verf%C3%BCgbarkeitshinweise.1282621/[/URL]

moebius 2020-11-30 15:55

Today there were some Big Navi cards, among others in the AMD shop, you have to stay tuned to get one.

moebius 2020-12-10 19:03

[QUOTE=moebius;564839]Today there were some Big Navi cards, among others in the AMD shop, you have to stay tuned to get one.[/QUOTE]
This 2 RX grafic cards you can actually (today) buy (Delivery for private customers only to Germany). But it's not to cheap
[URL="https://www.mindfactory.de/Hardware/Grafikkarten+(VGA)/Radeon+RX+Serie/RX+6800.html"]https://www.mindfactory.de/Hardware/Grafikkarten+(VGA)/Radeon+RX+Serie/RX+6800.html
[/URL]

moebius 2020-12-13 19:37

[QUOTE=moebius;565887]This 2 RX grafic cards you can actually (today) buy (Delivery for private customers only to Germany).[/QUOTE] Now sold out again, they probably had around 200 AMD RX 6800 cards in stock!

moebius 2020-12-14 10:15

XFX Radeon RX 6900XT, 3 Piecec available, only Germany [URL="https://www.computeruniverse.net/de/xfx-radeon-rx-6900xt-gaming-16-gb-enthusiast-grafikkarte?utm_campaign=adgoal&awc=11355_1607940734_99fd72f6f68c94cd669b4ce7984f6ffe&utm_channel=affiliate&utm_source=aw&utm_medium=textlink&utm_content=text"]https://www.computeruniverse.net/de/xfx-radeon-rx-6900xt-gaming-16-gb-enthusiast-grafikkarte?utm_campaign=adgoal&awc=11355_1607940734_99fd72f6f68c94cd669b4ce7984f6ffe&utm_channel=affiliate&utm_source=aw&utm_medium=textlink&utm_content=text[/URL]

kruoli 2020-12-14 10:35

They are € 1.5K! They should be € 999.

moebius 2020-12-14 10:53

[QUOTE=kruoli;566143]They are € 1.5K! They should be € 999.[/QUOTE]
Unfortunately, you don't get any for less than 1150 Euro at the moment.Except maybe for this price with a lot of luck in the AMD shop. Apparently they also deliver abroad, for example France. I think tomorrow the 3 cards will be sold out anyway.

moebius 2020-12-14 11:56

6800 XT [url]https://www.computeruniverse.net/de/powercolor-radeon-rx-6800xt-red-devil-limited-edition-16-gb-oc-enthusiast-grafikkarte?utm_campaign=adgoal&awc=11355_1607946937_0080751b4ed460d4c4a5ac5269acfbfc&utm_channel=affiliate&utm_source=aw&utm_medium=textlink&utm_content=text[/url]

moebius 2020-12-14 15:37

Last info for today from me: 2 different Sapphire AMD RX 6800 [URL="https://www.alternate.de/Grafikkarten/Big-Navi"]https://www.alternate.de/Grafikkarten/Big-Navi[/URL] Europe-wide delivery

moebius 2020-12-16 06:52

1 Attachment(s)
Probable data for the AMD 6700 XT (Navi 22 XT)

M344587487 2020-12-16 09:12

The 5700XT that the 6700XT replaces has a 256 bit bus, implying that the 6700XT has infinity cache to make up for the bus deficit. Unlikely to be the full 128MB of Big Navi, I'm guessing 32-64MB. The question is if infinity cache is any good for us and if so what capacity?

moebius 2020-12-17 16:49

[QUOTE=kruoli;566143]They are € 1.5K! They should be € 999.[/QUOTE]
1.25 K but available today PowerColor Radeon RX 6900 XT [URL="https://www.mindfactory.de/product_info.php/16GB-PowerColor-Radeon-RX-6900-XT-DDR6-Drei-Luefter-256bit--bulk-_1388380.html"]https://www.mindfactory.de/product_info.php/16GB-PowerColor-Radeon-RX-6900-XT-DDR6-Drei-Luefter-256bit--bulk-_1388380.html[/URL]

moebius 2020-12-17 18:57

[QUOTE=moebius;566454]1.25 K but available today PowerColor Radeon RX 6900 XT [/QUOTE] Sold out now!

moebius 2020-12-18 13:19

3 Sapphire Radeon RX 6800 Pulse OC Gaming 16GB,
price depends on the country of delivery f.e. 849,- € Germany, 731,90 € Honduras [URL="https://www.computeruniverse.net/en/sapphire-radeon-rx-6800-pulse-oc-gaming-16gb?awc=11355_1608297496_3cf79256ea6bb30c7878f994a357b150&utm_campaign=adgoal&utm_channel=affiliate&utm_content=text&utm_medium=textlink&utm_source=aw"]https://www.computeruniverse.net/en/sapphire-radeon-rx-6800-pulse-oc-gaming-16gb?awc=11355_1608297496_3cf79256ea6bb30c7878f994a357b150&utm_campaign=adgoal&utm_channel=affiliate&utm_content=text&utm_medium=textlink&utm_source=aw[/URL]

2 Sapphire Radeon RX6800 Nitro+ OC Gaming 16 GB OC
[URL="https://www.computeruniverse.net/de/sapphire-radeon-rx6800-nitro-oc-gaming-160-gb-oc"]https://www.computeruniverse.net/de/sapphire-radeon-rx6800-nitro-oc-gaming-160-gb-oc[/URL]

1 GIGABYTE GeForce RTX3080 GAMING OC 10 GB OC
[URL="https://www.computeruniverse.net/de/gigabyte-geforce-rtx3080-gaming-oc-10-gb-oc-enthusiast-grafikkarte"]https://www.computeruniverse.net/de/gigabyte-geforce-rtx3080-gaming-oc-10-gb-oc-enthusiast-grafikkarte[/URL]

Delivery worldwide [URL="https://www.computeruniverse.net/getalldeliverycountrys"]https://www.computeruniverse.net/getalldeliverycountrys[/URL]

moebius 2020-12-18 14:29

3 Powercolor Radeon RX 6800XT Red Devil Limited Edition 16 GB OC again available
[URL="https://www.computeruniverse.net/en/powercolor-radeon-rx-6800xt-red-devil-limited-edition-16-gb-oc-enthusiast-grafikkarte?awc=11355_1607946937_0080751b4ed460d4c4a5ac5269acfbfc&utm_campaign=adgoal&utm_channel=affiliate&utm_content=text&utm_medium=textlink&utm_source=aw"]https://www.computeruniverse.net/en/powercolor-radeon-rx-6800xt-red-devil-limited-edition-16-gb-oc-enthusiast-grafikkarte?awc=11355_1607946937_0080751b4ed460d4c4a5ac5269acfbfc&utm_campaign=adgoal&utm_channel=affiliate&utm_content=text&utm_medium=textlink&utm_source=aw[/URL]

M344587487 2020-12-18 14:35

I don't have enough middle fingers to flip all of those prices off at once.

moebius 2020-12-18 14:41

[QUOTE=M344587487;566568]I don't have enough middle fingers to flip all of those prices off at once.[/QUOTE]Yes, you're right, but these are even better than the ones from the real rip-offs. I only post it because the availability at the dealers is currently still very poor. But you see it is still bought like stupid.

moebius 2020-12-18 16:23

[QUOTE=moebius;566567]1 GIGABYTE GeForce RTX3080 GAMING OC 10 GB OC
3 Powercolor Radeon RX 6800XT Red Devil Limited Edition 16 GB OC[/QUOTE] SOLD OUT!

PhilF 2020-12-18 17:21

[QUOTE=moebius;566567]3 Powercolor Radeon RX 6800XT Red Devil [B]Limited Edition[/B] 16 GB OC[/QUOTE]

In the context of GPUs, doesn't Limited Edition mean Inflated Price for No Good Reason?

moebius 2020-12-18 19:00

[QUOTE=PhilF;566576]In the context of GPUs, doesn't Limited Edition mean Inflated Price for No Good Reason?[/QUOTE]In this case, Limited edition could also mean that your old power supply is too limited for the new graphics card, which supposedly needs at least 850W. [URL="https://www.powercolor.com/product?id=1606105457#spe"]https://www.powercolor.com/product?id=1606105457#spe[/URL]

tServo 2020-12-19 13:38

[QUOTE=moebius;566586]In this case, Limited edition could also mean that your old power supply is too limited for the new graphics card, which supposedly needs at least 850W. [URL="https://www.powercolor.com/product?id=1606105457#spe"]https://www.powercolor.com/product?id=1606105457#spe[/URL][/QUOTE]

That's bizarre, for a 300W card a power supply of 850W seems too high.
It must be a typo.
Most recommendations are for 700W:

[URL="https://www.techpowerup.com/gpu-specs/radeon-rx-6800-xt.c3694"]https://www.techpowerup.com/gpu-specs/radeon-rx-6800-xt.c3694[/URL]

moebius 2020-12-20 02:18

Remaining merchant offers for the weekend for private customers

Computeruniverse.net
[URL="https://www.computeruniverse.net/de/sapphire-radeon-rx6800-nitro-oc-gaming-160-gb-oc"]1 Sapphire Radeon RX6800 Nitro+ OC Gaming 16 GB OC
[/URL][URL="https://www.computeruniverse.net/en/asus-tuf-geforce-rtx3080-gaming-oc-tuf-rtx3080-o10g-gaming-10-gb-oc-enthusiast-grafikkarte?awc=11355_1608430621_a044afe308119f507868867d7a0d0b8f&utm_campaign=adgoal&utm_channel=affiliate&utm_content=text&utm_medium=textlink&utm_source=aw"]1 ASUS TUF GeForce RTX3080 GAMING OC TUF-RTX3080-O10G-GAMING 10 GB OC[/URL]
Delivery: [URL="https://www.computeruniverse.net/en/getalldeliverycountrys"]Worldwide[/URL]

Cyberport.de
[URL="https://www.cyberport.de/gaming/gaming-komponenten/grafikkarten/sapphire-technologies/pdp/2e01-7jf/sapphire-amd-radeon-rx-6800-oc-nitroplus-gaming-grafikkarte-mit-16gb-gddr6.html"]Sapphire AMD Radeon RX 6800 OC Nitro+ Gaming 16 GB
[/URL] Delivery:[URL="https://www.cyberport.de/service/kunden-international/eu-laender-und-drittland.html"] Parts of the EU[/URL]

Alternate.de
[URL="https://www.alternate.de/XFX/Radeon-RX-6800-MERC-319-Black-Gaming-16GB-Grafikkarte/html/product/1698793?"]XFX Radeon RX 6800 MERC 319 Black Gaming 16 GB[/URL]
Delivery: only Germany

moebius 2020-12-20 02:29

Sold out!
ASUS TUF GeForce RTX3080 GAMING OC TUF-RTX3080-O10G-GAMING 10 GB OC

moebius 2020-12-20 03:57

acom-pc.de
[URL="https://www.acom-pc.de/Hardware-Software/Grafikkarten-Zubehoer/Grafikkarten-NVIDIA-PCI-Express/Manli-RTX-3080-Grafikkarten-GF-RTX-3080-10-GB-GDDR6X-PCIe-4.0-HDMI-3x-DisplayPort.html"]Manli RTX 3080 - 10 GB[/URL]
[URL="https://www.acom-pc.de/Hardware-Software/Grafikkarten-Zubehoer/Grafikkarten-NVIDIA-PCI-Express/Gigabyte-GeForce-RTX-3080-GAMING-OC-10G-Grafikkarten-GF-RTX-3080-10-GB-GDDR6X-PCIe-4.0-x16-2x-HDMI-3x-DisplayPort.html"]Gigabyte GeForce RTX 3080 GAMING OC 10G[/URL]
[URL="https://www.acom-pc.de/Hardware-Software/Grafikkarten-Zubehoer/Grafikkarten-NVIDIA-PCI-Express/ZOTAC-GAMING-GeForce-RTX-3080-AMP-Holo-Grafikkarten-GF-RTX-3080-10-GB-GDDR6X-PCIe-4.0-HDMI-3-x-DisplayPort.html"]ZOTAC GAMING GeForce RTX 3080 AMP Holo[/URL]
Delivery: only Germany

moebius 2020-12-20 12:56

Sold out!
ZOTAC GAMING GeForce RTX 3080 AMP Holo

moebius 2020-12-20 21:22

Sold out!
Sapphire Radeon RX6800 Nitro+ OC Gaming 16 GB OC

[I]Did anyone here got catched a exemplar of this new rare species?[/I]

moebius 2020-12-21 13:25

Sold Out
XFX Radeon RX 6800 MERC 319 Black Gaming 16 GB

[B]New Offers from today
[/B]
Cyberport.de
[URL="https://www.cyberport.de/gaming/gaming-komponenten/grafikkarten/sapphire-technologies/pdp/2e01-7jh/sapphire-amd-radeon-rx-6800-oc-pulse-gaming-grafikkarte-mit-16gb-gddr6.html?zanpid=11657_1608556736_ddd6f876ce2d016ea2e047c23cbc0a93&userid=101248&MC=101-101248&ext_publisher_id=101248&awc=11657_1608556736_ddd6f876ce2d016ea2e047c23cbc0a93&APID=101"]AMD Radeon RX 6800 OC Pulse Gaming 16GB[/URL]
Delivery: [URL="https://www.cyberport.de/service/kunden-international/eu-laender-und-drittland.html"]EU[/URL], Third country = shipping costs + 20 Euro

Alternate.de
[URL="https://www.alternate.de/GIGABYTE/Radeon-RX-6800-GAMING-OC-16GB-Grafikkarte/html/product/1697045?"]GIGABYTE Radeon RX 6800 GAMING OC 16GB[/URL]
[URL="https://www.alternate.de/ASUS/Radeon-RX-6800-ROG-STRIX-GAMING-OC-16GB-Grafikkarte/html/product/1695405?"]ASUS Radeon RX 6800 ROG STRIX GAMING OC 16GB
[/URL][URL="https://www.alternate.de/ZOTAC/GeForce-RTX-3080-AMP-Holo-Grafikkarte/html/product/1677985?"]ZOTAC GeForce RTX 3080 AMP Holo[/URL]
[URL="https://www.alternate.de/EVGA/GeForce-RTX-3080-FTW3-ULTRA-GAMING-Grafikkarte/html/product/1673524?"]EVGA GeForce RTX 3080 FTW3 ULTRA GAMING[/URL],
Delivery: only Germany

moebius 2020-12-21 13:53

Sold Out!
ZOTAC GeForce RTX 3080 AMP Holo
EVGA GeForce RTX 3080 FTW3 ULTRA GAMING,

New offer 3 minutes ago
Caseking.de
[URL="https://www.caseking.de/zotac-gaming-geforce-rtx-3080-amp-holo-10240-mb-gddr6x-gczt-166.html?__shop=2"]ZOTAC GAMING GeForce RTX 3080 AMP Holo,[/URL]
Delivery: [URL="https://www.caseking.de/ckShippingcosts?__shop=2"]Europe[/URL]

moebius 2020-12-21 14:56

Sold Out!
ASUS Radeon RX 6800 ROG STRIX GAMING OC 16GB

New Offer
computeruniverse.net
[URL="https://www.computeruniverse.net/en/evga-geforce-rtx3080-ultra-ftw3-10-gb-oc-enthusiast-grafikkarte?awc=11355_1608562413_064f9ead8bb30ff87acdfc778f90ca5a&utm_campaign=adgoal&utm_channel=affiliate&utm_content=text&utm_medium=textlink&utm_source=aw"]3 EVGA GeForce RTX3080 Ultra FTW3 10 GB OC[/URL]
Delivery: [URL="https://www.computeruniverse.net/getalldeliverycountrys"]Worldwide[/URL]

Cyberport.de
[URL="https://www.cyberport.de/gaming/gaming-komponenten/grafikkarten/zotac/pdp/2e13-1ha/zotac-gaming-geforce-rtx-3080-amp-holo-10gb-gddr6x-grafikkarte-3xdp-hdmi.html?zanpid=11657_1608563385_18b25ebfb9354f7726816dbb274933bc&userid=101248&MC=101-101248&ext_publisher_id=101248&awc=11657_1608563385_18b25ebfb9354f7726816dbb274933bc&APID=101"]ZOTAC GeForce RTX 3080 AM Holo[/URL]
Delivery: [URL="https://www.cyberport.de/service/kunden-international/eu-laender-und-drittland.html"]EU[/URL], Third country = shipping costs + 20 Euro

moebius 2020-12-21 15:44

Sold Out!
ZOTAC GAMING GeForce RTX 3080 AMP Holo
ZOTAC GeForce RTX 3080 AM Holo
3 EVGA GeForce RTX3080 Ultra FTW3 10 GB OC

moebius 2020-12-21 16:46

[B]New Offer[/B]

Alternate.de
[URL="https://www.alternate.de/GIGABYTE/Radeon-RX-6900-XT-16G-Grafikkarte/html/product/1702526?"]GIGABYTE Radeon RX 6900 XT 16G[/URL]
Delivery: only Germany

VBCurtis 2020-12-21 18:36

There is no evidence any forumites care about these offers and sold-outs. Unless someone corrects me to show interest, please ease up (or cease) with the daily shopping summary.

moebius 2020-12-21 19:05

[QUOTE=VBCurtis;566880]There is no evidence any forumites care about these offers and sold-outs. Unless someone corrects me to show interest, please ease up (or cease) with the daily shopping summary.[/QUOTE]
There are quite a few people here who read this thread regularly, and the offers that I post are not astronomical for dealer prices. If I could afford it I would probably buy a AMD 6800 XT. I know that people who tend to do TF with mfaktc would rather use an RTX 3080
[B]From now on I will not post any more offers[/B], these are only for information, and I think the people here were intelligent enough to assess the offers. The prices are currently too high for me, but they are acceptable for others. Best regards.

Xyzzy 2020-12-28 00:31

1 Attachment(s)
Here is a graph of various recent AMD cards.

The orange is theoretical FP64 performance in TFLOPS.
The yellow is the number of milliseconds per iteration for the M77936867 PRP benchmark.

You can see that the R7 is not like the others.

You may be able to extrapolate 6800/6800 XT/6900 XT performance using the slope from the Navi cards.

:mike:

moebius 2020-12-28 01:16

[QUOTE=Xyzzy;567499]You may be able to extrapolate 6800/6800 XT/6900 XT performance using the slope from the Navi cards.[/QUOTE]Aren't there too few points to really assess the function? :smile: So, you can only calculate a best-fit line with linear regression. I'm not a mathematician I just fought through analysis and linear algebra with Mathematica at university.

axn 2020-12-28 02:26

[QUOTE=Xyzzy;567499]You may be able to extrapolate 6800/6800 XT/6900 XT performance using the slope from the Navi cards.[/QUOTE]

The 128MB "Infinity Cache" is going to affect the constant of proportionality... a lot...

I am hoping that the 6800XT will become a viable replacement for R7 for PRP workloads. *fingers crossed*

Viliam Furik 2020-12-28 03:16

[QUOTE=axn;567509]The 128MB "Infinity Cache" is going to affect the constant of proportionality... a lot...

I am hoping that the 6800XT will become a viable replacement for R7 for PRP workloads. *fingers crossed*[/QUOTE]

Assuming the Infinity Cache allows the cards to unlock their full FP64 potential, the RX 6800 (~1 TFLOPS) should already be at least on par with Radeon VII (~1 TFLOPS, due to memory bottleneck).

axn 2020-12-28 04:45

[QUOTE=Viliam Furik;567510]Assuming the Infinity Cache allows the cards to unlock their full FP64 potential, the RX 6800 (~1 TFLOPS) should already be at least on par with Radeon VII (~1 TFLOPS, due to memory bottleneck).[/QUOTE]

That would be very nice indeed, but I'm keeping my expectations low. If we can hit a decent fraction of R7 thruput, that would be great -- say 70%/80%/90% respectively by 6800/6800XT/6900XT. Anything above that would be bonus.

M344587487 2020-12-28 11:49

Even 50% throughput from a 6800XT would be somewhat of a win, it'd still be the most viable consumer card out there in production and the IC is a big enough departure from the usual design that there is potential for further optimisation. There is an assumption that the 6700 and lower will also have IC but at lower capacities, another interesting wrinkle.



There is some potential with the RTX 3000's double-duty INT32/FP32 units too, but the optimisation effort may be punishing and they're still on a less efficient node so my expectations for that are low. IMO it's an optimisation effort that may not pay off now but could pay dividends in the future assuming nvidia stick to the design.

M344587487 2020-12-28 23:46

RX 6900XT
 
A kind soul with a 6900XT answered my request for some gpuowl benchmarks. Asrock Phantom Gaming 6900XT stock, Ubuntu 20.04 with kernel 5.4, ROCm 4.0. They've since wiped the install in an attempt to get ROCm's ML working (godspeed) so these are all the benchmarks we're going to get, but it's plenty to give us an idea of performance.


The 57M and 79M tests were done a few different ways (no ops, -maxAlloc 14000, -carry short, -carry short -maxAlloc 14000), they all had the same timings so I've only pasted the no ops results. maxAlloc didn't seem to stick or at least it didn't translate into doing P-1, no matter.

[code][57885161, ]
2020-12-28 17:09:56 GpuOwl VERSION v7.2-21-g28dbf88
2020-12-28 17:09:56 GpuOwl VERSION v7.2-21-g28dbf88
2020-12-28 17:09:56 Note: not found 'config.txt'
2020-12-28 17:09:56 config: -prp 57885161 -iters 200000
2020-12-28 17:09:56 device 0, unique id ''
2020-12-28 17:09:56 gfx1030-0 57885161 FFT: 3M 1K:6:256 (18.40 bpw)
2020-12-28 17:09:56 gfx1030-0 57885161 OpenCL args "-DEXP=57885161u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=6u -DAMDGPU=1 -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0.51445938099070077 -DIWEIGHT_STEP_MINUS_1=-0.33969836857173502 -DIWEIGHTS={0,-0.33969836857173502,-0.12800351106634347,-0.42421929575738759,-0.23962212328737542,-0.49792124750469385,-0.33695316124376262,-0.12437818131180543,-0.42182548460600072,-0.23646084847019155,-0.49583385258551427,-0.33419654070262389,-0.12073777923072029,-0.41942172117280535,-0.23328643063717852,-0.49373777931154078,} -cl-std=CL2.0 -cl-finite-math-only "
2020-12-28 17:09:58 gfx1030-0 57885161 OpenCL compilation in 2.35 s
2020-12-28 17:09:58 gfx1030-0 57885161 maxAlloc: 0.0 GB
2020-12-28 17:09:58 gfx1030-0 57885161 You should use -maxAlloc if your GPU has more than 4GB memory. See help '-h'
2020-12-28 17:09:58 gfx1030-0 57885161 P1(0) 0 bits
2020-12-28 17:09:58 gfx1030-0 57885161 PRP starting from beginning
2020-12-28 17:09:59 gfx1030-0 57885161 OK 0 on-load: blockSize 400, 0000000000000003
2020-12-28 17:09:59 gfx1030-0 57885161 validating proof residues for power 8
2020-12-28 17:09:59 gfx1030-0 57885161 Proof using power 8
2020-12-28 17:09:59 gfx1030-0 57885161 OK 800 0.00% 5727fe6a7225c273 459 us/it + check 0.26s + save 0.10s; ETA 07:22
2020-12-28 17:10:04 gfx1030-0 57885161 10000 0.02% 91565f36715e33e3 465 us/it
2020-12-28 17:10:08 gfx1030-0 57885161 20000 0.03% f2c610087d02c3ea 464 us/it
2020-12-28 17:10:13 gfx1030-0 57885161 30000 0.05% fe1565094c7f7b47 462 us/it
2020-12-28 17:10:17 gfx1030-0 57885161 40000 0.07% adb226c2322baa14 463 us/it
2020-12-28 17:10:22 gfx1030-0 57885161 50000 0.09% 96339cf030b79d74 464 us/it
2020-12-28 17:10:27 gfx1030-0 57885161 60000 0.10% 175901ec29adfa87 464 us/it
2020-12-28 17:10:31 gfx1030-0 57885161 70000 0.12% 7c2d3978b07c9f39 467 us/it
2020-12-28 17:10:36 gfx1030-0 57885161 80000 0.14% c2ee4a9ca385f917 464 us/it
2020-12-28 17:10:41 gfx1030-0 57885161 90000 0.16% a7a038f5438a2fa5 466 us/it
2020-12-28 17:10:45 gfx1030-0 57885161 100000 0.17% f1cbf8d474fd3237 467 us/it
2020-12-28 17:10:50 gfx1030-0 57885161 110000 0.19% 6d709cb8366f244d 464 us/it
2020-12-28 17:10:55 gfx1030-0 57885161 120000 0.21% 2172b8f3cc5b3272 465 us/it
2020-12-28 17:10:59 gfx1030-0 57885161 130000 0.22% 06fffcab14e3c81b 466 us/it
2020-12-28 17:11:04 gfx1030-0 57885161 140000 0.24% af31f96be3309024 466 us/it
2020-12-28 17:11:09 gfx1030-0 57885161 150000 0.26% f6ac00d9a2354121 465 us/it
2020-12-28 17:11:13 gfx1030-0 57885161 160000 0.28% fd84ac518a5eb59d 465 us/it
2020-12-28 17:11:18 gfx1030-0 57885161 170000 0.29% e91f9213bc5ea1a3 464 us/it
2020-12-28 17:11:23 gfx1030-0 57885161 180000 0.31% 63a2a2c5417898f9 464 us/it
2020-12-28 17:11:27 gfx1030-0 57885161 190000 0.33% 48ec91fc60cf2bde 466 us/it
2020-12-28 17:11:32 gfx1030-0 57885161 Stopping, please wait..
2020-12-28 17:11:32 gfx1030-0 57885161 OK 200000 0.35% de62d6db1ad5092d 466 us/it + check 0.27s + save 0.11s; ETA 07:28
2020-12-28 17:11:32 gfx1030-0 Exiting because "stop requested"
2020-12-28 17:11:32 gfx1030-0 Bye[/code][code][77936867, ]
2020-12-28 17:11:32 GpuOwl VERSION v7.2-21-g28dbf88
2020-12-28 17:11:32 GpuOwl VERSION v7.2-21-g28dbf88
2020-12-28 17:11:32 Note: not found 'config.txt'
2020-12-28 17:11:32 config: -prp 77936867 -iters 200000
2020-12-28 17:11:32 device 0, unique id ''
2020-12-28 17:11:32 gfx1030-0 77936867 FFT: 4M 1K:8:256 (18.58 bpw)
2020-12-28 17:11:32 gfx1030-0 77936867 OpenCL args "-DEXP=77936867u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DAMDGPU=1 -DMM_CHAIN=1u -DMM2_CHAIN=2u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0.33644726404543274 -DIWEIGHT_STEP_MINUS_1=-0.25174750481886216 -DIWEIGHTS={0,-0.25174750481886216,-0.44011820345520131,-0.16213409745771243,-0.37306474779553728,-0.061788266441989627,-0.29798072935699788,-0.47471232907613115,-0.21390437908665341,-0.41180199020062258,-0.11975874301407295,-0.3413572830988989,-0.014337887291734644,-0.26247586476052853,-0.44814572555075455,-0.17414732433395128,} -cl-std=CL2.0 -cl-finite-math-only "
2020-12-28 17:11:35 gfx1030-0 77936867 OpenCL compilation in 2.26 s
2020-12-28 17:11:35 gfx1030-0 77936867 maxAlloc: 0.0 GB
2020-12-28 17:11:35 gfx1030-0 77936867 You should use -maxAlloc if your GPU has more than 4GB memory. See help '-h'
2020-12-28 17:11:35 gfx1030-0 77936867 P1(0) 0 bits
2020-12-28 17:11:35 gfx1030-0 77936867 PRP starting from beginning
2020-12-28 17:11:35 gfx1030-0 77936867 OK 0 on-load: blockSize 400, 0000000000000003
2020-12-28 17:11:35 gfx1030-0 77936867 validating proof residues for power 8
2020-12-28 17:11:35 gfx1030-0 77936867 Proof using power 8
2020-12-28 17:11:36 gfx1030-0 77936867 OK 800 0.00% 1579c241dc63eca6 613 us/it + check 0.32s + save 0.13s; ETA 13:17
2020-12-28 17:11:42 gfx1030-0 77936867 10000 0.01% fc4f135f7cf4ad29 620 us/it
2020-12-28 17:11:48 gfx1030-0 77936867 20000 0.03% 3cd1bd9d5e09cbc5 618 us/it
2020-12-28 17:11:54 gfx1030-0 77936867 30000 0.04% c4e0ff35e3290d98 620 us/it
2020-12-28 17:12:01 gfx1030-0 77936867 40000 0.05% dffe1b1b0d748128 619 us/it
2020-12-28 17:12:07 gfx1030-0 77936867 50000 0.06% 52e286945371ed29 619 us/it
2020-12-28 17:12:13 gfx1030-0 77936867 60000 0.08% 0945da4dc08bdd95 620 us/it
2020-12-28 17:12:19 gfx1030-0 77936867 70000 0.09% 7131fa4eb77f4bb2 620 us/it
2020-12-28 17:12:25 gfx1030-0 77936867 80000 0.10% 8d76071d27ee4221 621 us/it
2020-12-28 17:12:31 gfx1030-0 77936867 90000 0.12% 0bacff453b2f470e 620 us/it
2020-12-28 17:12:38 gfx1030-0 77936867 100000 0.13% 6d7296b9e2830f50 622 us/it
2020-12-28 17:12:44 gfx1030-0 77936867 110000 0.14% 8cbfd4435622bda7 622 us/it
2020-12-28 17:12:50 gfx1030-0 77936867 120000 0.15% 79ae5dad855057ad 622 us/it
2020-12-28 17:12:56 gfx1030-0 77936867 130000 0.17% 50c97bcbf876231f 621 us/it
2020-12-28 17:13:03 gfx1030-0 77936867 140000 0.18% e1db15f897271496 622 us/it
2020-12-28 17:13:09 gfx1030-0 77936867 150000 0.19% 127631386c6a9b17 622 us/it
2020-12-28 17:13:15 gfx1030-0 77936867 160000 0.21% 25b7b6206fc6f085 623 us/it
2020-12-28 17:13:21 gfx1030-0 77936867 170000 0.22% 416816b0d9f4bba8 622 us/it
2020-12-28 17:13:27 gfx1030-0 77936867 180000 0.23% 6bee5d054f770861 623 us/it
2020-12-28 17:13:34 gfx1030-0 77936867 190000 0.24% f37f068f014b18a0 621 us/it
2020-12-28 17:13:40 gfx1030-0 77936867 Stopping, please wait..
2020-12-28 17:13:40 gfx1030-0 77936867 OK 200000 0.26% f0b04b45b0855bd2 620 us/it + check 0.34s + save 0.16s; ETA 13:24
2020-12-28 17:13:40 gfx1030-0 Exiting because "stop requested"
2020-12-28 17:13:40 gfx1030-0 Bye[/code][code]“[332220523, ]”
2020-12-28 16:57:10 GpuOwl VERSION v7.2-21-g28dbf88
2020-12-28 16:57:10 GpuOwl VERSION v7.2-21-g28dbf88
2020-12-28 16:57:10 Note: not found 'config.txt'
2020-12-28 16:57:10 config: -prp 332220523 -iters 50000
2020-12-28 16:57:10 device 0, unique id ''
2020-12-28 16:57:10 gfx1030-0 332220523 FFT: 18M 1K:9:1K (17.60 bpw)
2020-12-28 16:57:11 gfx1030-0 332220523 OpenCL args "-DEXP=332220523u -DWIDTH=1024u -DSMALL_HEIGHT=1024u -DMIDDLE=9u -DAMDGPU=1 -DWEIGHT_STEP_MINUS_1=0.31797529154814252 -DIWEIGHT_STEP_MINUS_1=-0.24126043453715812 -DIWEIGHTS={0,-0.24126043453715812,-0.42431427180125786,-0.12640892148665336,-0.337171884696568,-0.49708608381811953,-0.23683862754188784,-0.42095927188310595,-0.12131777912660047,-0.33330903355459196,-0.49415518582120899,-0.23239105099670423,-0.41758471958785059,-0.11619696644233313,-0.32942367036371434,-0.49120720704209719,} -cl-std=CL2.0 -cl-finite-math-only "
2020-12-28 16:57:14 gfx1030-0 332220523 OpenCL compilation in 2.89 s
2020-12-28 16:57:14 gfx1030-0 332220523 maxAlloc: 0.0 GB
2020-12-28 16:57:14 gfx1030-0 332220523 You should use -maxAlloc if your GPU has more than 4GB memory. See help '-h'
2020-12-28 16:57:14 gfx1030-0 332220523 P1(0) 0 bits
2020-12-28 16:57:14 gfx1030-0 332220523 PRP starting from beginning
2020-12-28 16:57:16 gfx1030-0 332220523 OK 0 on-load: blockSize 400, 0000000000000003
2020-12-28 16:57:16 gfx1030-0 332220523 validating proof residues for power 8
2020-12-28 16:57:16 gfx1030-0 332220523 Proof using power 8
2020-12-28 16:57:21 gfx1030-0 332220523 OK 800 0.00% b950798999630b08 3954 us/it + check 1.92s + save 0.49s; ETA 15d 04:53
2020-12-28 16:57:58 gfx1030-0 332220523 10000 0.00% 503cd91d7b8e30e5 3969 us/it
2020-12-28 16:58:38 gfx1030-0 332220523 20000 0.01% f2d3ffbb3586c527 3978 us/it
2020-12-28 16:59:18 gfx1030-0 332220523 30000 0.01% e7846100baf7ce53 3977 us/it
2020-12-28 16:59:57 gfx1030-0 332220523 40000 0.01% e305c82567149969 3969 us/it
2020-12-28 17:00:37 gfx1030-0 332220523 Stopping, please wait..
2020-12-28 17:00:39 gfx1030-0 332220523 OK 50000 0.02% 72885d5ee0a11128 3974 us/it + check 1.90s + save 0.50s; ETA 15d 06:39
2020-12-28 17:00:39 gfx1030-0 Exiting because "stop requested"
2020-12-28 17:00:39 gfx1030-0 Bye[/code]This was taken during a test, unless my eyes deceive me that looks like a nice stock voltage, relative to Vega at least:[code]Adapter: PCI adapter
vddgfx: 1.09 V
fan1: 1760 RPM (min = 0 RPM, max = 3300 RPM)
edge: +73.0°C (crit = +100.0°C, hyst = -273.1°C)
(emerg = +105.0°C)
junction: +90.0°C (crit = +110.0°C, hyst = -273.1°C)
(emerg = +115.0°C)
mem: +74.0°C (crit = +100.0°C, hyst = -273.1°C)
(emerg = +105.0°C)[/code]By my reckoning using Moebius's data, that's ~84% the throughput of a tuned Radeon VII using a stock 6900XT. Many factors that'll affect the actual ratio in performance and efficiency, but a good starting point IMO.

Xyzzy 2020-12-28 23:51

We wonder if the 6800 XT will scale down ~10% for the 72 versus 80 compute units. Everthing else on the cards is identical we think.

M344587487 2020-12-29 00:01

If we're bandwidth limited there's a chance it might scale down better than that, although binning will pull it back the other way and probably then some.

moebius 2020-12-29 00:30

[QUOTE=Xyzzy;567571]We wonder if the 6800 XT will scale down ~10% for the 72 versus 80 compute units. Everthing else on the cards is identical we think.[/QUOTE]Thank you, we have been waiting longingly for these values. And as I see my estimate is very close to reality (the only thing I can do pretty well)
[URL="https://mersenneforum.org/showpost.php?p=563735&postcount=24"]post #24[/URL]

axn 2020-12-29 03:40

[QUOTE=Xyzzy;567571]We wonder if the 6800 XT will scale down ~10% for the 72 versus 80 compute units. Everthing else on the cards is identical we think.[/QUOTE]

Hopefully better than that, since 6800XT should clock higher than 6900XT. They're both same TDP, so 6900XT can't clock as high.

axn 2020-12-29 03:46

[QUOTE=M344587487;567570]By my reckoning using Moebius's data, that's ~84% the throughput of a tuned Radeon VII using a stock 6900XT. Many factors that'll affect the actual ratio in performance and efficiency, but a good starting point IMO.[/QUOTE]
I am wondering how much more thruput can be obtained by:
1) Running two copies of gpuowl -- 128MB cache can hold two FFTs at current PRP leading edge easily.
2) Not using the GPU for display. Not sure if the benchmarks were done when the GPU was also driving display(s). That might affect the cache / performance.

M344587487 2020-12-29 17:01

[QUOTE=axn;567598]I am wondering how much more thruput can be obtained by:
1) Running two copies of gpuowl -- 128MB cache can hold two FFTs at current PRP leading edge easily.[/QUOTE]
I'm not so sure it'll help or what the memory utilisation is actually like, are you sure a wavefront test uses so little memory? (I vaguely recall hundreds of MB but don't know if all of it is accessed uniformly every iteration). Unless two workers can fully fit in cache there's probably cache contention bleeding performance like we see with CPUs. Still, worth a shot benchmarking everything every which way and letting the results speak for themselves.

When one of us gets a card they should run bandwidth tests across a spectrum of RAM utilisations to determine how strong the cache really is. Or test every FFT and estimate that way. You can bet that the "1.5TB/s effective memory bandwidth" is marketing speak for "the cache has a maximum throughput of 1.5TB/s", so that's probably the upper bound at best.

[QUOTE=axn;567598]2) Not using the GPU for display. Not sure if the benchmarks were done when the GPU was also driving display(s). That might affect the cache / performance.[/QUOTE]Normally negligible performance hit, if there is any it's within margin of error so undetectable on an R7 AFAIR. They did drive a display with the card during the tests mostly idling on desktop, if there is a performance penalty the numbers will be slightly better than we've got. I doubt cache has much to do with the framebuffer but I'd be happy to be proved wrong if it gets us free performance.

Viliam Furik 2020-12-29 17:12

[QUOTE=M344587487;567645]I'm not so sure it'll help or what the memory utilisation is actually like, are you sure a wavefront test uses so little memory? (I vaguely recall hundreds of MB but don't know if all of it is accessed uniformly every iteration).[/QUOTE]

Prime95 uses a little more than 8 bytes per FFT word, as I was told. So 6M (almost current wavefront FFT) times 8 B is about 48 MiB, which fits nicely 2 times in the 128MiB of cache if gpuOwl has the same cache requirements.

But I would like to know, whether gpuOwl has been told to use the cache, i.e. whether it can even use the L3 cache of these GPUs. I don't recall any confirmation from Preda. However, I expect he has implemented that.

Performance numbers indicate that it is indeed using the cache, but I want to be sure.


All times are UTC. The time now is 01:29.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.