Intel processors crashing Unreal engine games (and others)

[DAPUNISHER]

@Det0x

Thanks for taking one for the team.

 

[igor_kavinski]

@Det0x

Thanks for taking one for the team.

Indeed. A worthy knight to have sat through the whole thing for us!

 

[H433x0n]

Since we are already speaking about Jay2Cents in this thread, he just released a new video.. But this time it actually had some substance

Performance with unlimited power which reviewers have been using in comparion-charts for the last two years: (all auto bios settings)
41273 points in Cinebench R23 @ 341w packet power

Performance with updated bios which followes the new "intel stock power limits"
37574 points in Cinebench R23 @ 253w packet power
View attachment 103836

Now we have to wait for the next bios update to check for how and if that one also affects the performance

With these Intel profiles the difference in silicon quality now manifests itself by producing different scores in benchmarks. Previously all of these chips scored relatively close, the difference was that the better bins would use less power for the same score. Now with a fixed power budget, there is a more noticeable gap in performance.

I have the same exact motherboard (Asus Z790-E Strix) and when running the same exact profile on the same 2402 BIOS I get a score of 38,800. I’ve got a significantly better bin than JayzTwoCents, his chip had a SP88 rating, while mine is SP104. So the difference between a top 10% and a bottom 10% bin is worth roughly 3.2% in nT performance.

 

[DAPUNISHER]

Could the big OEM be HP instead? After all Dell have been known as Intel's best friend which money could buy since the original Athlon days.

Anyway, that the comment about the large failure rate at a big SI at their intake QA: would love to know how they are testing.

No matter how good someone's intake QA is (and I appreciate they may have stepped it up just because of this) I cannot see it involving lengthy tests over days, so this System Integrator must have found some program or tool their intake QA uses.

Such a tool could really help end users.

About getting a full refund and considering a full platform swap: possibly going for a 12th gen replacement is the cheapest option for most customers. Normally resell value drops per "gen" (even counting rebrands), but in this case it might not.

I wanted to circle back to this.

Wendell stated that all the usual tools were running fine. Even y-cruncher was checking out. Turns out -

Raptor Lake hates this one weird trick! </clickbait>

All he had to do was take a 10GB file and compress and decompress it a few times and "sad trombone noises" It craps the bed revealing the degradation.

 

[igor_kavinski]

I have the same exact motherboard (Asus Z790-E Strix) and when running the same exact profile on the same 2402 BIOS I get a score of 38,800. I’ve got a significantly better bin than JayzTwoCents, his chip had a SP88 rating, while mine is SP104. So the difference between a top 10% and a bottom 10% bin is worth roughly 3.2% in nT performance.

It just KEEPS getting worse!

 

[igor_kavinski]

All he had to do was take a 10GB file and compress and decompress it a few times and "sad trombone noises" It craps the bed revealing the degradation.

I bet that the power usage and temps go through the roof for such a test and doing it a few times separates the men from the sad, sad underdeveloped creatures.

 

[DAPUNISHER]

It just KEEPS getting worse!

ROFLwaffles 🤣 It's laugh or cry I guess. I mean, how many silicon lotteries do you have to win just to play your games without contracting S.I.D.S (Sudden Intel Death Syndrome)?

 

[H433x0n]

It just KEEPS getting worse!

I would argue this is a better way of doing things than the alternative. I’d rather have a variance of 0-3% in benchmark scores at the same power consumption than similar benchmarks but with power draw varying wildly. This is what they should have had as the default from the beginning.

 

[DAPUNISHER]

Since I didn't put on this warpaint for nothing: We are bringing you live video of Raptor Lake (dumpster fire inside)

 

[igor_kavinski]

I’d rather have a variance of 0-3% in benchmark scores at the same power consumption than similar benchmarks but with power draw varying wildly. This is what they should have had as the default from the beginning.

Whatever

Maybe loss of performance is OK for you but I would rather let my 12700K die an honorable warrior's death, fighting till the last breath rather than have it gimped by some dumb microcode that is finally doing things sanely. Intel deserves the frickin' largest class action lawsuit in the history of Silicon Valley for failing to deliver advertised performance (the performance that was advertised in launch day reviews and with which Intel was TOTALLY fine coz it let them have a good showing in benchmarks).

And YES, before you decide to point it out to me, I know my 12700K probably isn't impacted but the purportedly irreversible microcode update may still have an adverse effect on its maximum achievable performance levels and so I will NOT be updating my mobo's BIOS anymore and I will torture my 12700K with Intel XTU and kill it before I settle for a gimped 14900K/KS. Heck, after I kill my 12700K, I will get a 14900KS and let it run as long as I can on the last known "performant" BIOS before I feel the need to update the BIOS to let the next victim of my cruel and unusual punishment live.

 

[H433x0n]

Whatever

Maybe loss of performance is OK for you but I would rather let my 12700K die an honorable warrior's death, fighting till the last breath rather than have it gimped by some dumb microcode that is finally doing things sanely. Intel deserves the frickin' largest class action lawsuit in the history of Silicon Valley for failing to deliver advertised performance (the performance that was advertised in launch day reviews and with which Intel was TOTALLY fine coz it let them have a good showing in benchmarks).

And YES, before you decide to point it out to me, I know my 12700K probably isn't impacted but the purportedly irreversible microcode update may still have an adverse effect on its maximum achievable performance levels and so I will NOT be updating my mobo's BIOS anymore and I will torture my 12700K with Intel XTU and kill it before I settle for a gimped 14900K/KS. Heck, after I kill my 12700K, I will get a 14900KS and let it run as long as I can on the last known "performant" BIOS before I feel the need to update the BIOS to let the next victim of my cruel and unusual punishment live.

I think you’re over reacting to a pretty benign statement but okay.

 

[coercitiv]

Maybe loss of performance is OK for you but I would rather let my 12700K die an honorable warrior's death

Start punching then:

Discussion - Variance in CPU quality and how it affects reviews and user experience

Der8auer tested multiple R5 7600 CPUs and found up to 50% difference in efficiency while gaming. From the start everyone should know this is not related to AMD, but rather a normal occurrence in all CPUs, and the variance is obviously bigger in budget SKUs. In the past, the same youtuber did a...

forums.anandtech.com

Spoiler

 

[IEC]

Buildzoid is probing the Intel 0x125 microcode update with an oscilloscope:

tl;dw
It looks like Intel did lower voltages for light loads by at least 0.05V (50mV). However, he thinks this is probably not going to be enough to prevent degradation (14900K stock VID goes up to 1.5V+) and will likely have an impact on frequency and performance. He is also not confident it will do anything beyond push degradation out to beyond the warranty period...

 

[igor_kavinski]

He is also not confident it will do anything beyond push degradation out to beyond the warranty period...

Crap. I guess Bartlett Lake it is, then.

 

[csbin]

LMAO

https://twitter.com/x/status/1816660855196647577

 

[DAPUNISHER]

He is also not confident it will do anything beyond push degradation out to beyond the warranty period...

That's the meat and potatoes. Can they pull off the shenanigans they always have? Or is the wheel of karma finally going to swing round and kick them in the ass? Stay tuned!

This scenario is why GN is already calling for extended warranties.

 

[Joe NYC]

That's the meat and potatoes. Can they pull off the shenanigans they always have? Or is the wheel of karma finally going to swing round and kick them in the ass? Stay tuned!

This scenario is why GN is already calling for extended warranties.

Yes, Intel definitely needs to extend warranties. It would be a signal that Intel is owning up to this issue, rather than dodging. So far, Intel has been dodging.

 

[Thunder 57]

What's the standard CPU warranty? Like three years? If they made it five that would send some confidence to buyers. Even five years is lacking as most people never need to RMA a CPU as they generally last longer than motherboards. It would send a message though.

 

[Ranulf]

What's the standard CPU warranty? Like three years? If they made it five that would send some confidence to buyers. Even five years is lacking as most people never need to RMA a CPU as they generally last longer than motherboards. It would send a message though.

I think its 3 years for a retail boxed cpu and 1year for an oem tray version. What are the warranties for OEM systems these days? 1 year or 3?

 

[gorobei]

Buildzoid is probing the Intel 0x125 microcode update with an oscilloscope:

tl;dw
It looks like Intel did lower voltages for light loads by at least 0.05V (50mV). However, he thinks this is probably not going to be enough to prevent degradation (14900K stock VID goes up to 1.5V+) and will likely have an impact on frequency and performance. He is also not confident it will do anything beyond push degradation out to beyond the warranty period...

caveat: BZ is running the test on an already degraded cpu and tweaking bios settings which has weird interactions with the microcode and gigabyte's bios not playing well together.

he also noted a significant amount of micro votage spiking/fluctuation possible caused by the e-cores and p-cores competing to see which one needs to be used in idle to everyday usage(post patch). the voltage to use the e-cores ends up knocking the p-core voltage up at low loads. he thinks it is a silly design.
under load the p cores are running outside where he is comfortable(1.4v) and the e-cores are pulling additional voltage increasing vdroop to where the p-cores dont get enough voltage.
he is worried that the aug patch is just intel trying to lower the voltage but not actually lowering the frequency enough to keep it stable.
slightly shorter for BZ standards summary at the 1h42m mark.

 

[WelshBloke]

Yes, Intel definitely needs to extend warranties. It would be a signal that Intel is owning up to this issue, rather than dodging. So far, Intel has been dodging.

If it's shown that the CPU was sold with a known fault they are going to have to replace them way beyond the warranty period in the EU!

 

[SteinFG]

If it's shown that the CPU was sold with a known fault they are going to have to replace them way beyond the warranty period in the EU!

They'll just say that the issue was in the microde, and it's fixed now.

 

[WelshBloke]

They'll just say that the issue was in the microde, and it's fixed now.

It's on them to show that it's actually fixed then and that it has the same capability as when it was sold.

 

[KompuKare]

If it's shown that the CPU was sold with a known fault they are going to have to replace them way beyond the warranty period in the EU!

It's on them to show that it's actually fixed then and that it has the same capability as when it was sold.

On the other hand European customers got nothing from Nvidia for Bumbgate while American customers got .. about $5. Nevermind then!

 

[DAPUNISHER]

If it's shown that the CPU was sold with a known fault they are going to have to replace them way beyond the warranty period in the EU!

How's the UK handle it?