Consistent Error. Potential Bug? - Page 2

paulunderwood · 2021-06-24, 10:45

To reiterate, to solve your particular hardware error problem, try underclocking the RAM

Shat · 2021-06-24, 12:26

I don't quite understand, why would under clocking RAM solve the issue? I'm currently running default on 2133mhz 15-15-15-36 already, should I force it lower?

I bumped my vcore voltage by 0.0125v offset, that didn't help.

I now notice a consistent failure on 16K length now on Core 2, (if it passes the initial 480K, it'll fail here)

I intentionally did a 15-17K custom test. It fails after Test 3 16K on 1 of Core 2's worker too.

Shat · 2021-06-24, 12:43

I just underclocked my RAM to 1866mhz 13-13-13-31. Same issue.

I'm fairly confident now that it's constantly failing on FFT 16K now... Not necessary right after Test 3, but I'm seeing it right after Test 4 now.

Hmm... I don't know if this is a good way to test for my PC stability now.... I can't tell if this is a bug in the code or not.

drkirkby · 2021-06-24, 12:47

IMHO if you bought new hardware, and they are supposed to be compatible, then the seller should resolve the problem. I would keep copies of logs, screen shots or photos. You should not need to underclock RAM or other tricks to get the system working reliably.
If moving the RAM between sockets causes the suspect core to change, the problem is probably the RAM. But what I infer from what you wrote, this is relatively new and should work properly.

Shat · 2021-06-24, 13:00

Yea, but when I sent my CPU/Motherboard/RAM to retailer (Computer Alliance in QLD Australia), they supposedly tested and came back saying they can't reproduce the error....

I take it no one else has issue with 16K FFT failure? (Actually error might be related to a specific CPU too, the old Intel/DDR3 ran the old P95 fine...)

kruoli · 2021-06-24, 13:43

If they cannot reproduce it, you should ask exactly what they have tested and how they tested it.

Again, it is virtually certain that there is a problem with some part of the hardware. A bug would be reproducible on every attempt (if the parameters are identical) and the error would be identical, too. You e.g. stated a round off error of 2.352155227e+16, is this always identical?

Additionally, it is weird that only (or mostly) CPU core 2 is problematic. Since the code running is identical for each core, the other cores should occur this error roughly at the same rate if this was a bug.

Finally, you will have to write done exactly what you did, what Prime95 version you used and so on and send that to the retailer (after they told you what I asked for in the beginning of this post) together with the machine.

paulunderwood · 2021-06-24, 13:56

It would seem to me you have a duff CPU or maybe a duff motherboard.

chalsall · 2021-06-24, 15:14

Quote:

Originally Posted by paulunderwood

It would seem to me you have a duff CPU or maybe a duff motherboard.

More likely the former. But I concur with others here; it's bad hardware, not a bug in the software.

Please note that it has been documented that Intel, AMD, and Amazon (and ***many*** others) use GIMPS software for sanity checking. This is just yet another example of why George's code is so important in practical domains as well as rarified mathematics.

Uncwilly · 2021-06-24, 15:27

Quote:

Originally Posted by chalsall

Please note that it has been documented that Intel, AMD, and Amazon (and ***many*** others) use GIMPS software for sanity checking.

I found an problem with my Fujistu laptop right out of the package with Prime95. I then ran Memtest to confirm that there was a memory problem. I walked the machine back into the store and told them it did not pass the 2 tests. They gave me a replacement no questions asked. I took the replacement and ran Memtest for hours before allowing Bill Gates to boot. Then I let George hit the machine hard. It passed. If I deployed hardware professionally, I would burn in any machine for at least 12 hours on Memtest and 24 on Prime95 and any GPU with gpuOwl for 24.

drkirkby · 2021-06-24, 21:09

Quote:

Originally Posted by Shat

Yea, but when I sent my CPU/Motherboard/RAM to retailer (Computer Alliance in QLD Australia), they supposedly tested and came back saying they can't reproduce the error....

I would suggest it would be worth creating a video demonstrating the problem and upload the video to YouTube. Post a link here, asking people for comments. You may or may not feel it is worth re-making the video in view of comments you receive.

Once the video is clearly demonstrating the problem, provide the retailer a link to the video. I would have thought that a video would provide pretty convincing evidence to them, despite they were unable to reproduce the problem themselves.

Shat · 2021-06-25, 00:13

Thanks for all the feedback everyone!

I've started collecting a bunch of screenshot of failure on all the options that I've tried.

A video actually sounds like a good idea too. I'll do that!

2021-06-24, 10:45	#12
paulunderwood Sep 2002 Database er0rr 3,739 Posts	To reiterate, to solve your particular hardware error problem, try underclocking the RAM Last fiddled with by paulunderwood on 2021-06-24 at 10:47

2021-06-24, 13:00	#16
Shat Jun 2021 2²·3 Posts	Yea, but when I sent my CPU/Motherboard/RAM to retailer (Computer Alliance in QLD Australia), they supposedly tested and came back saying they can't reproduce the error.... I take it no one else has issue with 16K FFT failure? (Actually error might be related to a specific CPU too, the old Intel/DDR3 ran the old P95 fine...) Last fiddled with by Shat on 2021-06-24 at 13:01

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Potential Coppersmith attack on RSA	Alberico Lepore	Alberico Lepore	24	2019-09-13 07:30
Potential primality of F33, F34, and F35	siegert81	FermatSearch	37	2018-07-22 22:09
A potential cause of Windows low-memory messages	cheesehead	Software	14	2013-05-16 00:45
PrimeNet reports: consistent column widths please	James Heinrich	PrimeNet	11	2011-09-30 16:10
Low-Stress Job with High Potential? Mathematician	cheesehead	Lounge	20	2009-06-05 20:24

2021-06-24, 12:26	#13
Shat Jun 2021 2²×3 Posts	I don't quite understand, why would under clocking RAM solve the issue? I'm currently running default on 2133mhz 15-15-15-36 already, should I force it lower? I bumped my vcore voltage by 0.0125v offset, that didn't help. I now notice a consistent failure on 16K length now on Core 2, (if it passes the initial 480K, it'll fail here) I intentionally did a 15-17K custom test. It fails after Test 3 16K on 1 of Core 2's worker too.

2021-06-24, 12:43	#14
Shat Jun 2021 2²×3 Posts	I just underclocked my RAM to 1866mhz 13-13-13-31. Same issue. I'm fairly confident now that it's constantly failing on FFT 16K now... Not necessary right after Test 3, but I'm seeing it right after Test 4 now. Hmm... I don't know if this is a good way to test for my PC stability now.... I can't tell if this is a bug in the code or not.

2021-06-24, 12:47	#15
drkirkby "David Kirkby" Jan 2021 Althorne, Essex, UK 101101110₂ Posts	IMHO if you bought new hardware, and they are supposed to be compatible, then the seller should resolve the problem. I would keep copies of logs, screen shots or photos. You should not need to underclock RAM or other tricks to get the system working reliably. If moving the RAM between sockets causes the suspect core to change, the problem is probably the RAM. But what I infer from what you wrote, this is relatively new and should work properly.

2021-06-24, 13:43	#17
kruoli "Oliver" Sep 2017 Porta Westfalica, DE 7²·11 Posts	If they cannot reproduce it, you should ask exactly what they have tested and how they tested it. Again, it is virtually certain that there is a problem with some part of the hardware. A bug would be reproducible on every attempt (if the parameters are identical) and the error would be identical, too. You e.g. stated a round off error of 2.352155227e+16, is this always identical? Additionally, it is weird that only (or mostly) CPU core 2 is problematic. Since the code running is identical for each core, the other cores should occur this error roughly at the same rate if this was a bug. Finally, you will have to write done exactly what you did, what Prime95 version you used and so on and send that to the retailer (after they told you what I asked for in the beginning of this post) together with the machine.

2021-06-24, 13:56	#18
paulunderwood Sep 2002 Database er0rr 3,739 Posts	It would seem to me you have a duff CPU or maybe a duff motherboard.

2021-06-25, 00:13	#22
Shat Jun 2021 C₁₆ Posts	Thanks for all the feedback everyone! I've started collecting a bunch of screenshot of failure on all the options that I've tried. A video actually sounds like a good idea too. I'll do that!