Prime95 v30.3 - Page 45

Ensigm · 2020-11-29, 16:50

I wonder if anyone else is also interested in the P-1 benchmarks I posted here. I believe this could mean we can potentially extract more P-1 power on some machine configurations (by using different FFT implementations for Stage 1 and Stage 2).

BigDukeSix · 2020-12-21, 10:18

Is there a way to set the ETA to dd:hh:mm until it arrives less than one day?
And another one: is ist bad that I got told undefined architecture on my latest gen AMD Ryzen 5600?

Thanks in advance.

ZFR · 2020-12-21, 11:45

Quick question: can the constant writing of the 6GB proof file affect the lifespan of the ssd it is saved on? I have a second HDD connected, is it better to save the proof file on it instead?

retina · 2020-12-21, 11:55

Quote:

Originally Posted by ZFR

Quick question: can the constant writing of the 6GB proof file affect the lifespan of the ssd it is saved on?

Yes, all writes to any SSD reduce the lifespan.

The variable factors are how much lifespan you expect to get, how good the SSD controller is, and how much percentage each write reduces the life. Knowing nothing about your system or expectations, we can't answer those questions. Check the specs to see how it affects you.

ZFR · 2020-12-21, 13:15

OK, let me rephrase:

Roughly how many extra writes to disk are made (for the proof file only) on a full PRP test of an exponent in the 110,000,000 range. Proof power 8.

PhilF · 2020-12-21, 20:06

When I run GPUOwl, I copy the proof file over to my prime95 directory for easy upload.

Today, I had copied the file as usual, but stopped prime95 because I wanted to use the computer for other things. About an hour later prime95 uploaded the proof file, even though it was stopped.

This isn't a problem for me, but I wanted to make sure that this behavior was known and uploading while stopped is intentional.

Viliam Furik · 2020-12-21, 22:27

Quote:

Originally Posted by ZFR

OK, let me rephrase:

Roughly how many extra writes to disk are made (for the proof file only) on a full PRP test of an exponent in the 110,000,000 range. Proof power 8.

AFAIK, the residues used to produce the proof file are written once, into their folder, and then when the proof is generated, they are read, and the proof file is written once.

So about 6 GB of writes per test. SSDs have life expectancies in order of hundreds of TBW (Samsung uses ratio 1:600, that is drive capacity to the life capacity. For Pro versions they seem to use 1:1200, but that's on the 2,5" drives, M.2 seem to be all 1:600), which means you can safely write up to 25000 (realistically 20000-24000) proofs and residues to a "Samsung 970 EVO PLUS 250 GB", until it dies.

Unless you have a blazingly fast computer, that can do 20K+ tests within say 5 years, you don't have to worry about even scratching that amount. But if you really care about your drive, use HDD. They have a lot better capacity-to-price ratio, and the write size is limited by the drive's physical health. If you don't have an HDD, 1TB HDD comes at about $50.

ZFR · 2020-12-21, 23:25

Quote:

Originally Posted by Viliam Furik

So about 6 GB of writes per test.

Thanks. As you said, that's negligible.

I'm not really familiar with the maths behind the proof, so wasn't sure if it 6GB per test or per n iterations or every 30 mins, or....

Thanks for the reply.

ixfd64 · 2020-12-25, 23:59

FYI: George has released a version 30.4 for 64-bit Windows that fixes several ECM-related bugs: https://mersenneforum.org/showpost.p...&postcount=281

axn · 2020-12-26, 02:35

Quote:

Originally Posted by ixfd64

FYI: George has released a version 30.4 for 64-bit Windows that fixes several ECM-related bugs: https://mersenneforum.org/showpost.p...&postcount=281

Perhaps inappropriate for this thread, as the bugs were introduced and fixed in 30.4, and AFAICT, 30.4 is still "pre-release"

LaurV · 2021-01-04, 07:19

I have a small issue with the 32-bit version of the 30.3b6, is that still supported?

I have an old machine which still runs winxp32 (albeit the cpu is a 64bit i5-3470, but I can not change the OS) and till now it was doing ECM using an older version of P95. As I wanted to switch to PRPCF and PRPCFDC, I downloaded and installed the new one over the old one. Now, I am getting "access denied" every time I try to start it, and then it crashes (the program, not the computer). It was/is in own folder (not "program files"), it had/has all the rights, and it doesn't seem that it tries to access anything (files, libraries, etc, as well as I can monitor). It just crashing with the "access denied" error without opening any window.

Libraries issues? Any other known issue? If yes, is it fixed in 30.4?

2020-11-29, 16:50	#485
Ensigm Aug 2020 162₈ Posts	Benchmarking for P-1 I wonder if anyone else is also interested in the P-1 benchmarks I posted here. I believe this could mean we can potentially extract more P-1 power on some machine configurations (by using different FFT implementations for Stage 1 and Stage 2). Last fiddled with by Ensigm on 2020-11-29 at 16:52

2021-01-04, 07:19	#495
LaurV Romulan Interpreter Jun 2011 Thailand 9275₁₀ Posts	I have a small issue with the 32-bit version of the 30.3b6, is that still supported? I have an old machine which still runs winxp32 (albeit the cpu is a 64bit i5-3470, but I can not change the OS) and till now it was doing ECM using an older version of P95. As I wanted to switch to PRPCF and PRPCFDC, I downloaded and installed the new one over the old one. Now, I am getting "access denied" every time I try to start it, and then it crashes (the program, not the computer). It was/is in own folder (not "program files"), it had/has all the rights, and it doesn't seem that it tries to access anything (files, libraries, etc, as well as I can monitor). It just crashing with the "access denied" error without opening any window. Libraries issues? Any other known issue? If yes, is it fixed in 30.4? Last fiddled with by LaurV on 2021-01-04 at 07:25

2020-12-21, 10:18	#486
BigDukeSix Dec 2020 7₈ Posts	Is there a way to set the ETA to dd:hh:mm until it arrives less than one day? And another one: is ist bad that I got told undefined architecture on my latest gen AMD Ryzen 5600? Thanks in advance.

2020-12-21, 11:45	#487
ZFR Feb 2008 Bray, Ireland 2²×5×7 Posts	Quick question: can the constant writing of the 6GB proof file affect the lifespan of the ssd it is saved on? I have a second HDD connected, is it better to save the proof file on it instead?

2020-12-21, 13:15	#489
ZFR Feb 2008 Bray, Ireland 2²×5×7 Posts	OK, let me rephrase: Roughly how many extra writes to disk are made (for the proof file only) on a full PRP test of an exponent in the 110,000,000 range. Proof power 8.

2020-12-21, 20:06	#490
PhilF Feb 2005 Colorado 2×5×59 Posts	When I run GPUOwl, I copy the proof file over to my prime95 directory for easy upload. Today, I had copied the file as usual, but stopped prime95 because I wanted to use the computer for other things. About an hour later prime95 uploaded the proof file, even though it was stopped. This isn't a problem for me, but I wanted to make sure that this behavior was known and uploading while stopped is intentional.

2020-12-25, 23:59	#493
ixfd64 Bemusing Prompter "Danny" Dec 2002 California 5·11·43 Posts	FYI: George has released a version 30.4 for 64-bit Windows that fixes several ECM-related bugs: https://mersenneforum.org/showpost.p...&postcount=281