Intel still doesn't know what is causing Core i9 desktop CPUs to crash

HardReset said:

Not Core 2 Duo but Pentium Pro. Some disagree but since they too agree that Nehalem is Pentium Pro based and Sandy Bridge is on high level almost identical to Nehalem, we can say Raptor Lake is indeed based on Pentium Pro. There just are not single totally new architecture between Pentium Pro and Raptor Lake.

Click to expand...

Yeah, it's a 15-year-old microarchitecture that Intel has been hacking on ever since. It's way past the point where Intel should have gone back to the drawing board to design a new microarchitecture. They should've gone back to the drawing board when all the big-name CPU vulnerabilities were being found.

If AMD can go back to the drawing board and design a whole new microarchitecture, so can Intel. This is a symptom of Intel being lazy!!!

Hell, at this point, Intel should be recalling every single last Core i9 ever sold and refunding buyers for selling what is a defective product.

Well, they got them out the door, I'll give them that much.

trparky said:

Hell, at this point, Intel should be recalling every single last Core i9 ever sold and refunding buyers for selling what is a defective product.

Click to expand...

Can I get to keep mine? It works like a charm

Strawman said:

Can I get to keep mine? It works like a charm

Click to expand...

But for how long?

trparky said:

But for how long?

Click to expand...

It's been almost 2 years now

I never liked the boosting feature on CPU's. I understand the pro's since every workload is not the same, but this is pushing silicon to it's limits and there's likely a chance by now there's a handful of Intel CPU's being degraded due to a misconfigured setting.

All because Intel wanted to be so bad number one in benchmarks, while eating excessive amounts of power compared to AMD's part.

Vanderlinde said:

I never liked the boosting feature on CPU's. I understand the pro's since every workload is not the same, but this is pushing silicon to it's limits and there's likely a chance by now there's a handful of Intel CPU's being degraded due to a misconfigured setting.

All because Intel wanted to be so bad number one in benchmarks, while eating excessive amounts of power compared to AMD's part.

Click to expand...

Just checking here but AMDs boost too, up to their thermal limit. Not had a problem with either my 3600 or the 5800X3D I replaced it with. Or the Nehalem i7 I had years before. This is a problem unique to these chips.

yukka said:

Just checking here but AMDs boost too, up to their thermal limit. Not had a problem with either my 3600 or the 5800X3D I replaced it with. Or the Nehalem i7 I had years before. This is a problem unique to these chips.

Click to expand...

My 5600x do not boost up to thermal limmit. Under 60C it runs at 4650Mhz and for each 1-2 degrees above 60C it drops 25-50Mhz and at 95 it reach thermal throttle. Only with boxed cooler it was reaching 95C.
With a $20 cooler no matter the load it do not go above 72C with a room temp of 21-24C. Gaming max 62C.
For a 65W CPU it's a lot compared to older parts same power, but die size is 1/3. IHS height and die placement not so well designed if you ask me.

Now back to subject, Intel knows exactly what it's going on and they are the ones to blame for not enforcing/anouncing board partners to keep with the CPU limits.
When test scores counted more than the rest...

My 13700k is not longer as hot as it was on release. I do not know where they fixed it or slowed it down, but it is much more like a proper higher tier cpu

HardReset said:

Not Core 2 Duo but Pentium Pro. Some disagree but since they too agree that Nehalem is Pentium Pro based and Sandy Bridge is on high level almost identical to Nehalem, we can say Raptor Lake is indeed based on Pentium Pro. There just are not single totally new architecture between Pentium Pro and Raptor Lake.

Click to expand...

I've still got some Nehalem CPUs in action to this day as well as Gulftown. i7 920/60/980X

loki1944 said:

I've still got some Nehalem CPUs in action to this day as well as Gulftown. i7 920/60/980X

Click to expand...

I miss my old Westmere i7 980x a 4.3 ghz. In 2010 it cost more than AMD'S gpus in 2024.
This is Intel's Bulldozer moment.

I guess the clock is simply too high.
when I did chip design, some digital circuit seems correct but the output is garbage or jittery.
the circuit couldnt switch 0-to-1 or 1-to-0 levels fast enough.
I had to put 1-clock delayer in some positions to ensure the circuit has correct output albeit having higher process latency.

trparky said:

It's way past the point where Intel should have gone back to the drawing board to design a new microarchitecture.

Click to expand...

They did; it was called Itanium.

They should've gone back to the drawing board when all the big-name CPU vulnerabilities were being found.

Click to expand...

Which I again note: Itanium remains immune to pretty much all these vulnerabilities due to not relying on x86 performance hacks.

gamerk2 said:

They did; it was called Itanium.

Which I again note: Itanium remains immune to pretty much all these vulnerabilities due to not relying on x86 performance hacks.

Click to expand...

Too bad that would've figuratively put decades of software in the trash.

However, that doesn't explain how AMD has been able to pull off a minor miracle whereas Intel seems to be a sinking ship.

Skye Jacobs said:

Why it matters: For months now, Raptor Lake gamers have been annoyed about a stability issue with the chips causing games to crash. For a moment, it seemed as if Intel was about to push a fix. Unfortunately, the company tossed water on those reports, saying there was more to the problem than was being portrayed.

Recently, users have struggled with stability issues related to Intel's Core i9-14900K & 13900K CPUs. Those with the processors have noticed increased game crashing, leading numerous developers to issue support notices. Intel said it would investigate.

Reports that Intel had found a fix originated with the German blog Igor's Lab. It claimed it obtained an internal document under NDA that said the instability underlying cause was "an incorrect value in microcode algorithm associated with the eTVB (enhanced Thermal Velocity Boost) feature."

For a brief moment, Raptor Lake users believed Intel had a fix in the queue. Unfortunately, Intel said reports identifying the fault were incorrect. It is still trying to determine the cause.

"Contrary to recent media reports, Intel has not confirmed root cause and is continuing, with its partners, to investigate user reports regarding instability issues on unlocked Intel Core 13th and 14th generation (K/KF/KS) desktop processors," it said in a statement. "The microcode patch referenced in press reports fixes an eTVB bug discovered by Intel while investigating the instability reports. While this issue is potentially contributing to instability, it is not the root cause."

Igor's Lab has since added an addendum to its report.

Intel developed eTVB technology specifically for Intel's Raptor Lake S processors to give them a burst of performance in games and applications by overclocking the processor's cores above the maximum turbo frequency under certain conditions. Igor originally reported that Intel detected a displacement in the minimum operating voltage on the Core i9 parts due to exposure to elevated core voltages and that a firmware update was forthcoming, which was only partially correct.

In the meantime, gamers have developed workarounds, with some advice from developers. The crashes vary depending on the title, with some producing "out of memory" errors, others exiting to the desktop, and some locking up the computer entirely. Additionally, some of the chips work fine for a week to about a month but eventually start having stability issues. Unreal Engine seems to be a common denominator among the games.

Gamemakers have different advice on how to avoid the issue, but all are related to clocking. Epic Games issued a support notice to encourage users to adjust BIOS settings. Vermintide 2 and Warhammer 40,000: Darktide developer Fatshark advised using Intel Extreme Tuning Utility (XTU) to underclock the Performance Core speed from x55 to x53. Gearbox said to remove any overclocks in BIOS or use XTU, though changes made in the BIOS persist after a reset, unlike those made with Intel's software. Outpost: Infinity devs suggested downclocking from 5.5GHz to around 5GHz to avoid crashes.

Permalink to story:

Intel still doesn't know what is causing Core i9 desktop CPUs to crash

Click to expand...

Quite a while ago I replaced my 12600k with a shiny new i9 13900k, it has been cooled by an EK waterblock, the CPU is not overclocked, it resides in an EVGA z690 Classified motherboard. No crashes, have played many hours of gaming, Cyberpunk 2077, The Witcher3, many others, heck, I even gave that boring Starfield a try, no crashes. Perhaps it is the OS some use? I use Linux, not windblows

trparky said:

Too bad that would've figuratively put decades of software in the trash.

Click to expand...

It had an x86 emulation mode. At a 20% hit sure, but as Itanium sped up and x86 died off, no one would have cared.

x86-64 was a mistake in that it kept x86 going to the point where its basically impossible to replace now.

trparky said:

Hell, at this point, Intel should be recalling every single last Core i9 ever sold and refunding buyers for selling what is a defective product.

Click to expand...

Nah. First i9s were 9000-series.

yukka said:

Just checking here but AMDs boost too, up to their thermal limit. Not had a problem with either my 3600 or the 5800X3D I replaced it with. Or the Nehalem i7 I had years before. This is a problem unique to these chips.

Click to expand...

That's not how AMD's boost works. It has several limits (temp, power, current, frequency) and if it hits *any* of those it won't go higher. You can of course adjust them. Thing is, if you have enough cooling you'll never hit the temp limit because power limits will apply first.

godrilla said:

I miss my old Westmere i7 980x a 4.3 ghz. In 2010 it cost more than AMD'S gpus in 2024.
This is Intel's Bulldozer moment.

Click to expand...

Nah, Intel's Bulldozer moment is gen 13 and 14 i9s. Except no Bulldozer was even rated for what i9s draw.

gamerk2 said:

x86-64 was a mistake in that it kept x86 going to the point where its basically impossible to replace now.

Click to expand...

Why we would want to get rid of x86-64? FPU side already has basically no x86 (x87) stuff because x86-64 compilers only create SSE2 or higher FPU code (x87 will work but compilers never created such code). Only major problem is "only" 16 GP and FPU registers but it can be solved with new instruction set.

Only integer side still uses x86-based instructions but again, it too can be solved with new instruction set that will some time replace x86-64.

yesnt said:

That's not how AMD's boost works. It has several limits (temp, power, current, frequency) and if it hits *any* of those it won't go higher. You can of course adjust them. Thing is, if you have enough cooling you'll never hit the temp limit because power limits will apply first.

Click to expand...

Yeah I was just making the point that the boosting doesn’t usually impact reliability. These ones are “special”