LLR benchmark thread

Oddball · 2010-07-21, 20:14

I'm thinking of replacing my old Pentium 4 processor, so could you run some or all of these tests and tell me how long it takes to complete?

1.) k value between 100 and 200, n value between 500,000 and 501,000
2.) k value between 100 and 200, n value between 600,000 and 601,000
3.) k value between 300 and 2000, n value between 500,000 and 501,000
4.) k value between 300 and 2000, n value between 600,000 and 601,000
5.) k value above 1M, n value between 500,000 and 501,000

Any chip you have will do, but I'm especially interested in the AMD Phenom II X6 1055T and the Intel core i7 920. If you're not running your comps on stock settings, please say what your overclock is.

Thanks in advance.

VBCurtis · 2010-07-21, 20:40

You're asking the wrong question, but only barely. What you want to do is find what your current iteration time is for a range of FFT sizes. Asking someone to user any k from 300 to 2000 allows for more than one FFT size.

I suggest you post time per iteration (which allows you to find time per test by multiplying by the specific exponent) for FFTs at 48k, 56k, 64k, perhaps 80k. These should give you more useful data.

My machines are all overclocked, so I am not posting times (and I have only core2's and P4's anyway).
-Curtis

Oddball · 2010-07-21, 21:23

Quote:

Originally Posted by VBCurtis

You're asking the wrong question, but only barely. What you want to do is find what your current iteration time is for a range of FFT sizes. Asking someone to user any k from 300 to 2000 allows for more than one FFT size.

I suggest you post time per iteration (which allows you to find time per test by multiplying by the specific exponent) for FFTs at 48k, 56k, 64k, perhaps 80k. These should give you more useful data.

I see. Here are my times for a Pentium 4 at 3.2GHz:

FFT length of 40K: 0.98 ms/iteration
FFT length of 48K: 1.18 ms/iteration
FFT length of 56K: 1.43 ms/iteration
FFT length of 64K: 1.54 ms/iteration
FFT length of 80K: 2.08 ms/iteration

VBCurtis · 2010-07-22, 08:13

Core 2 quad at 3.20Ghz (stock is 2.66, DDR2 at 800):
80k 1.25ms
56k .88x ms
All four cores were running, two 80k and two 56k.

These scale nearly linearly, which is roughly to be expected from the code. Yours do also, save for 56k being a little high (possible background process, etc). This becomes less true for Mersenne-sized FFTs or really small FFTs, but for top5000 thru small megadigit searches, it's mostly linear. Say, 32k through 224k FFT sizes.

Thus, anyone who reports a chip and timing for a single FFT in your range tells you all you need to know. I'd personally like to see 80k, since that's the size both 6th and 7th drives are using right now. 9th drive is at 56k.
-Curtis

Kosmaj · 2010-07-22, 12:36

Just get the fastest Intel cpu based machine you can afford.
With as many cpu's and cores.

MooMoo2 · 2010-08-02, 00:11

Quote:

Originally Posted by Oddball

Any chip you have will do, but I'm especially interested in the AMD Phenom II X6 1055T...

Quote:

I see. Here are my times for a Pentium 4 at 3.2GHz:

FFT length of 40K: 0.98 ms/iteration

FFT length of 40K: 0.65 ms/iteration using one core of a 1055T.

2010-07-21, 20:14	#1
Oddball May 2010 499 Posts	LLR benchmark thread I'm thinking of replacing my old Pentium 4 processor, so could you run some or all of these tests and tell me how long it takes to complete? 1.) k value between 100 and 200, n value between 500,000 and 501,000 2.) k value between 100 and 200, n value between 600,000 and 601,000 3.) k value between 300 and 2000, n value between 500,000 and 501,000 4.) k value between 300 and 2000, n value between 600,000 and 601,000 5.) k value above 1M, n value between 500,000 and 501,000 Any chip you have will do, but I'm especially interested in the AMD Phenom II X6 1055T and the Intel core i7 920. If you're not running your comps on stock settings, please say what your overclock is. Thanks in advance.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Perpetual benchmark thread...	Xyzzy	Hardware	879	2022-11-23 16:36
Hardware Benchmark Jest Thread for 100M exponents	joblack	Hardware	285	2022-08-06 21:50
Sieve Benchmark Thread	Historian	Twin Prime Search	105	2013-02-05 01:35
sr5sieve Benchmark thread	axn	Sierpinski/Riesel Base 5	25	2010-05-28 23:57
Does anyone have i7 920? for Benchmark?	cipher	Twin Prime Search	2	2009-04-14 20:16

2010-07-21, 20:40	#2
VBCurtis "Curtis" Feb 2005 Riverside, CA 5637₁₀ Posts	You're asking the wrong question, but only barely. What you want to do is find what your current iteration time is for a range of FFT sizes. Asking someone to user any k from 300 to 2000 allows for more than one FFT size. I suggest you post time per iteration (which allows you to find time per test by multiplying by the specific exponent) for FFTs at 48k, 56k, 64k, perhaps 80k. These should give you more useful data. My machines are all overclocked, so I am not posting times (and I have only core2's and P4's anyway). -Curtis

2010-07-22, 08:13	#4
VBCurtis "Curtis" Feb 2005 Riverside, CA 13005₈ Posts	Core 2 quad at 3.20Ghz (stock is 2.66, DDR2 at 800): 80k 1.25ms 56k .88x ms All four cores were running, two 80k and two 56k. These scale nearly linearly, which is roughly to be expected from the code. Yours do also, save for 56k being a little high (possible background process, etc). This becomes less true for Mersenne-sized FFTs or really small FFTs, but for top5000 thru small megadigit searches, it's mostly linear. Say, 32k through 224k FFT sizes. Thus, anyone who reports a chip and timing for a single FFT in your range tells you all you need to know. I'd personally like to see 80k, since that's the size both 6th and 7th drives are using right now. 9th drive is at 56k. -Curtis

2010-07-22, 12:36	#5
Kosmaj Nov 2003 2×1,811 Posts	Just get the fastest Intel cpu based machine you can afford. With as many cpu's and cores.