Intel processors crashing Unreal engine games (and others)

[BlueWeasel]

Will there ever be an official tool that can test for the degraded processors?

Got a 14700K back in November and it seems to be fine. Have tried to undervolt it from day 1 so hopefully that has helped. But who knows about the future performance.

 

[AdamK47]

Will there ever be an official tool that can test for the degraded processors?

Got a 14700K back in November and it seems to be fine. Have tried to undervolt it from day 1 so hopefully that has helped. But who knows about the future performance.

What stress test software have you tried?

 

[DarthKyrie]

Will there ever be an official tool that can test for the degraded processors?

Got a 14700K back in November and it seems to be fine. Have tried to undervolt it from day 1 so hopefully that has helped. But who knows about the future performance.

If you are waiting on something official from Intel my advice would be not to hold your breath.

 

[BlueWeasel]

What stress test software have you tried?

The few that reportedly show the problem - Nvidia installer, Cinebench R23, etc. Is there some specific test that is more definitive?

I've had some weird crashes in games but they have been few and far between. Nothing of any consistency.

 

[Ranulf]

Might try the new SpaceMarine2 game. Reports are it pushes cpus up to 90C.

 

[lakedude]

Compile Black Myth Wukong?

 

[igor_kavinski]

Is there some specific test that is more definitive?

Write a script to do 10GB 7z compression and decompression concurrently, for a few hours. Easy to do.

Something like:

Compression script (create in notepad)
7z "compress" "source file" "destination file"
del "destination file"
7z "compress" "source file" "destination file"
del "destination file"
... and so on as many times as you want and rename this to cmptest.cmd

Decompression script
7z "decompress" "source file" "destination file"
del "destination file"
7z "decompress" "source file" "destination file"
del "destination file"
... and so on as many times as you want and rename this to dcmptest.cmd

Keep these in different folders.

Then run both cmptest.cmd and dcmptest.cmd and wait for them to finish. You can keep Task Manager core utilization graph and HWinfo open at the same time to see how the temps, voltages etc. are going during the whole process.

You will need to check 7z command line parameters to see what options to use. Recommend you try the most strenuous and slowest compression method. Probably LZMA2?

 

[GunsMadeAmericaFree]

If you are waiting on something official from Intel my advice would be not to hold your breath.

Maybe AMD will release one?

 

[AdamK47]

The few that reportedly show the problem - Nvidia installer, Cinebench R23, etc. Is there some specific test that is more definitive?

I've had some weird crashes in games but they have been few and far between. Nothing of any consistency.

I believe CoreCycler would be a good one to use for these CPUs. All core loads that peg the CPU hard at a constant voltage state don't seem to reveal the problem. Variable loads with variable voltages put sporadically across the cores would.

 

[igor_kavinski]

I believe CoreCycler would be a good one to use for these CPUs.

It won't try to set PBO on an Intel CPU, will it???

 

[DAPUNISHER]

Based on Wendell's results, if the 7z tests can make it through a few times, CPU should be good to go. This one is said part tongue and cheek part serious: Use it to run a Minecraft server.

 

[AdamK47]

Based on Wendell's results, if the 7z tests can make it through a few times, CPU should be good to go. This one is said part tongue and cheek part serious: Use it to run a Minecraft server.

I actually believe someone would be serious if they said a couple 7z tests should be sufficient.

These are the times we live in. People no longer stress test for hours with a varied array of tests. They want to have something quick and then forget about it.

 

[DrMrLordX]

I actually believe someone would be serious if they said a couple 7z tests should be sufficient. These are the times we live in. People no longer stress test for hours with a varied array of tests.

As has already been mentioned, some of these unstable Raptor Lakes can go for some time on constant workloads without error. Then you try to do a shader compilation and it just crashes.

 

[AdamK47]

As has already been mentioned, some of these unstable Raptor Lakes can go for some time on constant workloads without error. Then you try to do a shader compilation and it just crashes.

That's why you use a collection of stress tests over a long period of time.

It takes effort.

 

[igor_kavinski]

That's why you use a collection of stress tests over a long period of time.

It takes effort.

Shouldn't be required for stock settings. If users have to do this, Intel should send out the CPU for free and wait for the user to test and return if failed or give them the money if it passes. Honor code.

 

[BlueWeasel]

Thanks for the info.

Will definitely try a few runs with 7-Zip and CoreCycle. I know 7-Zip has a built-in benchmark tool but not sure if it pushes the CPU enough (compared to suggestion by @igor_kavinski) to be worthwhile.

Here's some other ones that can possibly indicate a CPU failure:

• Nvidia driver installer (CRC error)
• Cinebench R23 (single core benchmark sometimes fails)
• Chrome (installer fails)
• y-Cruncher

 

[igor_kavinski]

I know 7-Zip has a built-in benchmark tool but not sure if it pushes the CPU enough

It might. You can make it work harder by changing the options and it restarts with the heavier workload.

Not sure if it is checking data integrity during the benchmark, though.

 

[DAPUNISHER]

I actually believe someone would be serious if they said a couple 7z tests should be sufficient.

These are the times we live in. People no longer stress test for hours with a varied array of tests. They want to have something quick and then forget about it.

I do run 3DMark stress tests, but other than that I'm off and running. It's been enough to expose unstable negative CO offsets. Once it passes I have yet to experience any issues with games. I haven't ripped and transcoded a bluray in months.

 

[DAPUNISHER]

Someone please tell Intel it's a nothingburger - https://www.intel.com/content/www/us/en/support.html

I see the tech press is over a week late again with keeping up on happenings. I posted about owners having issues due to Intel having lack of stock for certain SKUs weeks ago. Some are reporting incorrectly (I think it was Paul's Hardware) that only 13th gen are a problem to get because they no longer produce them (wonder why?). I have seen a fair number of posts stating Intel would swap them for 14th but availability on those could takes weeks or more too. Seen them offer refunds, then reject them later. Or stop responding to the owner altogether. Big blue SNAFU.

 

[igor_kavinski]

12900KS is selling locally for $331. Never seen a KS sell this low, with two years warranty offered by the shop! 14700KF is only slightly more and also offers slightly more performance according to GB6. Was going for the 12900KS to mess with it and see if I could run my 7600 kit higher than the current 7000. But my request to convert it to four monthly installments got rejected so I rejected the notion too

 

[DAPUNISHER]

This proposal may sound radical, but hear me out. What if you saved the money for 4 months then bought it?

 

[DrMrLordX]

That's why you use a collection of stress tests over a long period of time.

It takes effort.

TBH I've never heard of anyone using 7zip or shader compilation for stability testing on an overclock, much less a stock CPU. Prime95, linpack, wprime, ycruncher, Cinebench, sure. Though from what I've heard some of these Raptor Lakes might also trip up on ycruncher.

 

[igor_kavinski]

This proposal may sound radical, but hear me out. What if you saved the money for 4 months then bought it?

Money isn't the issue. It's seeing that much get emptied from my account in one go!

Will try in October after the last installment of some RAM/SSD is paid. I think the stupid company offering the installments refuses to let people have multiple installments in parallel.

 

[igor_kavinski]

TBH I've never heard of anyone using 7zip or shader compilation for stability testing on an overclock

Because there never was a CPU so bad that it reliably failed in those two tests and on a grand scale too. There is no way that Intel didn't know what it was shipping. They knew they messed up and decided to tide it over through the customer support scumminess they are exhibiting right now to genuine owners of their defective products.

 

[lakedude]

There is no way that Intel didn't know what it was shipping.

Intel is a big company. It is likely some at Intel knew and were screaming bloody murder but they got labeled "Negative Nancies" and were ignored in favor of those with a more positive outlook.

This is one of the big problems with a democracy. Problems are overlooked and for some reason the overly optimistic people wearing rose colored glasses are better received than those who know better.

I once discovered a nationwide issue at my company that was causing real problems. It only got solved because I had the perseverance to endure 2 full days of people telling me it was a local issue (my fault) because no one else in the entire country was complaining. I found the problem in 5 minutes after our excellent operators noticed the irregularities but it took 2 days of frustration and dozens of phone calls to finally reach the engineer who made the coding mistake. Only thanks I got was to be the first to try the new beta software with the fix.