intel 13th and 14th gen xx900 and xx700 may have defective cores causing crashes in gaming

Randall Stephens said:
Is the board forcing the voltage or is the chip asking for more than it can take? I’m genuinely curious.
Click to expand...
The chip will happily suck up everything you throw at it. Intel builds the “oh shit” protections on the CPU but just like everyone else the BIOS controls the power flow.

Processors are like goldfish, they will eat everything you throw at them till they pop.
 
If someone here had this Intel problem, can he try with year earlier Nvidia driver?
 
Ranulfo said:
FYI, H forum thread on this in the cpu section with more info:

https://hardforum.com/threads/13900...-shader-compilation-in-two-ue5-games.2031265/
Click to expand...

Apparently this is a motherboard issue. It seems Intel never tests its processor with this setting but this is the first time this is causing issue

vick1000 said:
How many of you are forcing Intel's power limits in your BIOS? I find the 253w default limit too much. My B760 seems to handle it on the 13600K, and the Deep Cool AK620 keeps things between 85-90c, but don't like that range. So I enforce it to keep temps more reasonable in the 70s at 1.235 vcore (no undervolt on this mobo). I know Intel wants you to push the thermal limits to get more performance, but at such diminishing returns, I dislike the notion of that much wasted power.
Click to expand...
 
AMD users went through a similar situation when Zen 4 X3D processors came out. In that case, though, the issue was addressed relatively quickly compared to Intel.
 
d3athf1sh said:
So, this all started with a guy in the steam forums that claims he has tracked down the cause of "Out of video memory trying to allocate a rendering resource" errors and crash to desktop while gaming to being defective cores that actually seem to have been fine out of the box but then seem be degrading over time till they start throwing whea errors and to quote him " Inside those entries, look for 'Translation Look Aside Buffer' and 'Internal Parity Error'. There will be lots. That's a bad CPU."

he says he's been through 3 or 4 processors of both 13900k and 14900k varieties and it's always the same, fine for ~3-4 months then they all end up the same way, throwing errors and crashing. besides a lot of people on different forums just chalking it up to bad drivers or gpu. one of the main solutions that's been a temporary fix has been to undervolt and underclock the cpu. i know this sounds fishy at first but i'm trying to summarize a very long post and probably not doing a good job. but Tom's Hardware has now picked up the story and has reached out to intel for a response.

if this interests you i will post a link to his original post which is long and detailed and itself has a lot of links but also a link to the Tom's article as we await for a response from intel

https://steamcommunity.com/app/315210/discussions/0/4204741842669464852/?ctp=2

https://www.tomshardware.com/pc-com...blame-other-high-end-intel-cpus-also-affected
Click to expand...

View: https://www.youtube.com/watch?v=8yatSqh5hRA
 
Bro Buildzoid is constantly running unrealistic loads like linepack and OCCT with RAM timings that he creates himself lol. He's so far down the rabbit hole I'd question whether or not any realistic loads for creators or editors would even be compromised. Haven't seen any complaints on this forum for that matter. He says "high loads" lol this is coming from a guy who runs stress tests all day all night like his life depends on it pretending to play it cool but really the guy is in the deep end 🤯
 
xDiVolatilX said:
Bro Buildzoid is constantly running unrealistic loads like linepack and OCCT with RAM timings that he creates himself lol. He's so far down the rabbit hole I'd question whether or not any realistic loads for creators or editors would even be compromised. Haven't seen any complaints on this forum for that matter. He says "high loads" lol this is coming from a guy who runs stress tests all day all night like his life depends on it pretending to play it cool but really the guy is in the deep end 🤯
Click to expand...
is it really that unrealistic though? be honest
 
xDiVolatilX said:
Bro Buildzoid is constantly running unrealistic loads like linepack and OCCT with RAM timings that he creates himself lol. He's so far down the rabbit hole I'd question whether or not any realistic loads for creators or editors would even be compromised. Haven't seen any complaints on this forum for that matter. He says "high loads" lol this is coming from a guy who runs stress tests all day all night like his life depends on it pretending to play it cool but really the guy is in the deep end 🤯
Click to expand...

Yet what really led to this problem being discovered or discussed was constant crashing while playing games and pre-caching shaders for games. Not stress tests like cinebench etc..
 
Ranulfo said:
Yet what really led to this problem being discovered or discussed was constant crashing while playing games and pre-caching shaders for games. Not stress tests like cinebench etc..
Click to expand...

Because of the motherboard settings. It's an important scenario that specifies why it happens because of the motherboards entirely, not the CPUs. Builzoids issues are created by himself, that's different.
 
Good customer service? Accepting returns also allows them to analyze the damaged chips.

But yeah, they should claw back those losses from the guilty Mobo vendors.

4096A? Seriously?
 
d3athf1sh said:
if there's nothing wrong with the chips, why is intel accepting returns?
Click to expand...
Damage control. Even if the issue isn't Intel's fault Intel will be blamed for it by at least some people which makes for really bad press and damage to the company and brand which goes far beyond the cost of RMAs.
 
GoodBoy said:
Good customer service? Accepting returns also allows them to analyze the damaged chips.

But yeah, they should claw back those losses from the guilty Mobo vendors.

4096A? Seriously?
Click to expand...
4096A at, what, 1.4V? Sounds like a fun time.
 
Following the implementation of the Intel baseline profile, a performance drop of up to nearly 30% can be observed in power-intensive multi-threaded workloads. Cinebench R23 is a benchmark where the multi-core score drops from 40,021 to 28,811 points. Lightly threaded workloads are largely unaffected as is the single-core score.

Gaming workloads see a modest decline in performance, mostly in the 5-10% range. This is because most games don’t push the power consumption past the official Intel spec. Consequently, even at the official Intel power specifications, the 13th and 14th Gen CPUs can boost to their marketed boost clocks (or very close to them).

https://www.hardwaretimes.com/13th-...ec-than-board-partner-optimized-power-limits/
 
Intel should give us a 'baseline" set of benchmarks using the popular consumer tools available and make a concerted effort to normalize all other variables. They should be completely transparent in all material and settings used. So, we can see what they expect their CPU's to do at their stack settings. Do that with multiple MB's and they can show us all how their chip at their stock settings is the "fastest" whatever they always claim. Publish the data and let others replicate their work. I seems to me we used to get reviews like that from third parties years ago.

But, for now, they "fixed" it, while they still don't know "why" the issue occurs, because "motherboards".

1714246672032.png

They fought overclocking years ago, then relented and gave us OC marketed chips. Now, they rely on those chips pushed as hard as possible to maintain their marketing claims of dominance, usually with single core performance now. They let third party reviewers handle the marketing part for them. They really do need to tread carefully here.

1714247093853.png


Cool, I'll wait.
 
So in the past Intel claimed that having unrestricted power limits was "in spec" as long as you didn't change the CPU multiplier but now that consumers are reporting crashes they're back pedaling and saying that the problem is with motherboard partners that are not following the Intel power specs for 13th and 14th gen processors... You can't have your cake and eat it too Intel!
 
Marees said:
Have we got a clear recommendation from Intel on the power limits to be implemented as default by the motherboard manufacturers ?
Click to expand...
Not really. Check out the video that CAD4466HK posted. Intel has "recommended" values that are posted on their site:
https://edc.intel.com/content/www/u...cessor-line-thermal-and-power-specifications/

It seems that motherboard partners like Asus and Gigabyte and implementing different power limits under the "Intel baseline Profile" or "Intel Stock" power limits.
 
These boards should ship with the base spec as the default BIOS. If they want to offer tweaks, have it be labeled like an "extreme" option or something when you go to load defaults.
 
Last edited:
Intel has no one to lame but themselves. They have been saying publicly what the board manufacturers are doing is in spec for their cpu's. They've been saying it for years. Intel allowed them to push the chips with MB designed "normal" or "default" settings as they both benefit from the bigger bar better charts from influencers and trusted builders alike. That's why we are looking at them and not faulting the MB's.

This situation brought out the desire consumers have for MB makers to include a bone stock Intel setting, label it clearly and make it the bios default behavior. They are responding and several now have updated bios settings. So, this is happening and that's a good thing. But, that won't solve all Intel's problems.

Soon, we will see more fallout for Intel as their high end cpu's drop in performance charts across the spectrum of reviewers as they retest with updated bios's on "stock" settings and the "Intel stock" i9 acts like a juiced i7, or worse, an i5 or i3, in some workloads. According to HUB's charts, Single core performance in the C24 is not impacted much, but, the Multi core is. Gaming will follow depending on how the games you play use the cpu and what the user is doing while gaming. And we still don't know for sure if the default intel stock low power setting will keep the chips from tanking. There is possible an unresolved root cause mitigated for many, but not all, by lowering the juice.

This is not new. Many of us purposefully bought chips in the past that were great for overclocking but were a few tiers down from the highest end. Intel made that largely moot when they created more market segmentation with core counts/hyper threading and locking some chips into power/performance envelopes and letting other higher end chips go unrestricted and spike the price.

I wish HUB would have compared the intel stock settings to the MB "preferred" settings on each MB, versus comparing them to each other. Maybe a follow up video is coming.

They should've called 14th gen Icarus Lake. Their wax is melting.
 

Intel Statement on Stability Issues: "Motherboard Makers to Blame"​

by AleksandarK Today, 01:00 Discuss (43 Comments)
A couple of weeks ago, we reported on NVIDIA directing users of Intel's 13th Generation Raptor Lake and 14th Generation Raptor Lake Refresh CPUs to consult Intel for any issues with system stability. Motherboard makers, by default, often run the CPU outside of Intel's recommended specifications, overvolting the CPU through modifying voltage curves, automatic overclocks, and removing power limits.
 
Marees said:
Following the implementation of the Intel baseline profile, a performance drop of up to nearly 30% can be observed in power-intensive multi-threaded workloads. Cinebench R23 is a benchmark where the multi-core score drops from 40,021 to 28,811 points. Lightly threaded workloads are largely unaffected as is the single-core score.

Gaming workloads see a modest decline in performance, mostly in the 5-10% range. This is because most games don’t push the power consumption past the official Intel spec. Consequently, even at the official Intel power specifications, the 13th and 14th Gen CPUs can boost to their marketed boost clocks (or very close to them).

https://www.hardwaretimes.com/13th-...ec-than-board-partner-optimized-power-limits/
Click to expand...
Apparently using Intel's recommended setting of 253W keeps Cinebench R23 multicore scores around 40k.

1714401931629.png
 
You must log in or register to reply here.
Back
Top