NVIDIA Points Intel Raptor Lake CPU Users to Get Help from Intel Amid System Instability Issues

[efikkan]

BTW my friend who is a distributor for Intel has been saying he had to take in lots of RMA for Intel 14th gen, so the article is actually very legit

What we need is information like this, but with actual data.
There are always some failures with mass-produced electronics, but real statistics tells us whether there is a widespread defect/design flaw, or if this is just noise from a few people shouting very loudly.
Most of us probably remember the RTX 2080 Ti debacle, which eventually turned out to be tied to EVGA's designs, and excluding these left them with completely normal failure rates. But people still remember it as a Nvidia issue, when in this specific case it was an AiB vendor issue.

I would very much like to see statistics (even if it's just relative numbers) of how many CPUs are RMAed, and of these, how many are have been running completely stock (incl. memory) and are still confirmed to be defective. If the resulting figure is anywhere close to 1% (with a good sample size), then there is certainly a hardware issue. Also, if proven true, it will be interesting whether the resulting bad chips are tied to the same batches, or if it's evenly spread across everything (a widespread quality issue or design flaw).

I wonder if this is the cause of my issues lately (13900KS / 3080Ti). I did a BIOS update a couple months ago, and haven't been able to run any games since. Some crash right when 3D is activated, others completely at random, but never more than 5 minutes from 3D load start.

If upgrading or downgrading your BIOS doesn't resolve the issue, and you run everything within Intel's power limits and your memory at stock JEDEC speeds, and the problem is still as reproducible as you portrait it, then I would ask if either you (or someone in the same situation) would be willing to try installing a Linux distro like Ubuntu on a separate drive, just for testing purposes. Because if you do, then reproducing the same error on a completely different software stack will eliminate the software and lead you to the conclusion you have defective hardware (in which case you should RMA it). But this is only if you feel comfortable with this.

Spoiler

Installing a Linux distro like Ubuntu will take about ~20-30 min. All base drivers are there by default, you only have to install the Nvidia's drivers, for those who have Nvidia GPUs (you find this under Software & Updates -> Additional drivers).

Make absolutely sure you don't overwrite your current system or files. Disconnecting the drive is an option. Install the test system on a separate drive (even if it's an old HDD, as this is just a test). Also, choose to install the boot loader into the separate drive, so you can only boot from it through the BIOS boot menu.

Add AssRack(they burn non 3d cpu's) and MSI(they blame amd for any bios issues) to the list too.

I think there is an overall lack of quality in motherboards. Even expensive "gaming" motherboards from Asus aren't always as stable or compatible as we would expect, so I stay away from those. But I'm not a fan of the dynamic boosting from both CPU vendors either, as I fear they are pushing the limits too far.

 

[MaMoo]

What we need is information like this, but with actual data.
There are always some failures with mass-produced electronics, but real statistics tells us whether there is a widespread defect/design flaw, or if this is just noise from a few people shouting very loudly.
Most of us probably remember the RTX 2080 Ti debacle, which eventually turned out to be tied to EVGA's designs, and excluding these left them with completely normal failure rates. But people still remember it as a Nvidia issue, when in this specific case it was an AiB vendor issue.

I would very much like to see statistics (even if it's just relative numbers) of how many CPUs are RMAed, and of these, how many are have been running completely stock (incl. memory) and are still confirmed to be defective. If the resulting figure is anywhere close to 1% (with a good sample size), then there is certainly a hardware issue. Also, if proven true, it will be interesting whether the resulting bad chips are tied to the same batches, or if it's evenly spread across everything (a widespread quality issue or design flaw).

I completely agree. This is a statistical problem.

I get the impression that this issue is related to the amount of operating margin that the 13th and 14th gen chips have been engineered with, relative to the 12th gen. They all run on the same motherboards to date and I have not heard anything like this for the 12th gen. So whatever motherboard "boost" or "OC" that applies to all 12th to 14th gen chips eats away the safety margin in the 13th and 14th, but not the 12th, because it has a bigger physical safety margin.

 

[mechtech]

Neither of my i9's have these issues.

My 14900K is locked to intel limits though, Asus default settings were insane.
My 13900HX of course is low power being in a laptop.

Both are rock solid stable.

What were Asus default settings (stock or OC to the max?)

 

[phanbuey]

What were Asus default settings (stock or OC to the max?)

Asus aggressively overvolts and puts 4096W on turbo runtime. That's not unique to Intel though, they do that to AMD as well. Most of the stability and heat issues of the 13th and 14th gen were due to default motherboard settings.

Once you dial in the settings the chips run great.

 

[Darmok N Jalad]

I'm surprised we don't see a lot more of this with the demands we're placing on PSUs and power delivery. There are a lot of watts getting chewed up, and it's not just the watts, but the spikes. If both the CPU and GPU spike at the same moment, that's a big ask for the electrical system, even when it's engineered well. Naturally, dialing back those demands helps, and it doesn't even cost you that much performance. It certainly seems like factory recommended settings would be the ideal default, but then you won't see those shiny benchmark scores that reviewers are getting, and then queue the outrage.

 

[nguyen]

Asus aggressively overvolts and puts 4096W on turbo runtime. That's not unique to Intel though, they do that to AMD as well. Most of the stability and heat issues of the 13th and 14th gen were due to default motherboard settings.

Once you dial in the settings the chips run great.

I have used 4790K on MSI board, 8700K on Gigabyte and now 13700K on MSI board, the default settings are always no power limit.

Funny thing is Asus used to be the only brand that enforce Intel limits, until recently

If Intel was not aware that pretty much all motherboard makers remove power limits on their Z board, that's just sheer incompetence

 

[ThrashZone]

That’s actually part of improper behavior we are talking about. MCE definitely should not be enabled automatically with XMP.

Hi,
Yeah well last I saw a popup after activating xmp profile pops up and asks for permission to do so
If you or anyone else doesn't want the additional feature enabled I'd say click no hehe

Otherwise disable it or stick with optimized defaults F5 I believe where none of this is enabled and this is called Stock configuration lol

Asus aggressively overvolts and puts 4096W on turbo runtime. That's not unique to Intel though, they do that to AMD as well. Most of the stability and heat issues of the 13th and 14th gen were due to default motherboard settings.

Once you dial in the settings the chips run great.

Yeah that is a very brief window where that may happen
You can't even make it all the way through R20 without dropping off that.

 

[Gatt]

I have used 4790K on MSI board, 8700K on Gigabyte and now 13700K on MSI board, the default settings are always no power limit.

Funny thing is Asus used to be the only brand that enforce Intel limits, until recently

If Intel was not aware that pretty much all motherboard makers remove power limits on their Z board, that's just sheer incompetence

ASRock is doing that since almost a year and a half. I had to post the thing on their forum and one tech showed up and said: really?

A couple of Bios' later they allowed the choice between no-limits and Intel settings. But the default mode (i.e. after every bios upgrade) was (and I guess still is) no-limits.

 

[Psychoholic]

Asus aggressively overvolts and puts 4096W on turbo runtime. That's not unique to Intel though, they do that to AMD as well. Most of the stability and heat issues of the 13th and 14th gen were due to default motherboard settings.

Once you dial in the settings the chips run great.

Exactly this... 4096W and i THINK the timer is unlimited too although i'd have to double check that.
Also Asus default Current settings (iccMax) were 511Amps; I think intel is either 320a or 400a

 

[ThrashZone]

Exactly this... 4096W and i THINK the timer is unlimited too although i'd have to double check that.
Also Asus default Current settings were 511Amps; I think intel is either 320a or 400a

Hi,
No it's not unlimited
Timer limit is 448 seconds

 

[mechtech]

Asus aggressively overvolts and puts 4096W on turbo runtime. That's not unique to Intel though, they do that to AMD as well. Most of the stability and heat issues of the 13th and 14th gen were due to default motherboard settings.

Once you dial in the settings the chips run great.

So is it one setting to change or do you have to run the gauntlet and change a dozen settings in bios?

 

[Upgrayedd]

So is it one setting to change or do you have to run the gauntlet and change a dozen settings in bios?

Intel Processor Instability Causing Oodle Decompression Failures

RAD Game Tools' web page. RAD makes Bink Video, the Telemetry Performance Visualization System, and Oodle Data Compression - all popular video game middleware.

www.radgametools.com

There's an update towards the middle of the page about what to change.

Also just an FYI for Intel users. I recently built a 14700K and Asus mobo system and any instability I experienced was coming from a Thermal Grizzly contact frame. I put the stock ILM back on and it was solid after that. I had Intel's limits in place the entire time, not the mobo auto setting.

 

[matar]

No issues for me Nvidia as i am on still on comet lake i9-10900KF and i am very happy with it @5.2Ghz

 

[phanbuey]

So is it one setting to change or do you have to run the gauntlet and change a dozen settings in bios?

There are 2 settings "enhanced turbo" and equivalents (each mobo names it someting different) which runs all cores at the 2 core turbo, and the 4096W setting - it directly affects prime95 stability at lower voltage -- ie. capped @250W you will be stable at lower volts while you will crash at 4096W with even higher voltage -- add in the "enhanced turbo" and I can completely see how people are erroring out of games.

 

[Dr. Dro]

Exactly this... 4096W and i THINK the timer is unlimited too although i'd have to double check that.
Also Asus default Current settings (iccMax) were 511Amps; I think intel is either 320a or 400a

For my i9-13900KS MSI offers three modes on my Z690 Ace:

1. Intel stock, listed as "boxed cooler", at 253W/320A
2. Tower air cooler, 288W/512A (I use this for 24/7 as it's good balance between performance and thermals)
3. Water cooler, fully unrestrained at 4096W/512A

I bet a lot of people with dinky AIOs pick #3 thinking it'll be fine as "they're on water anyway"

 

[nguyen]

For my i9-13900KS MSI offers three modes on my Z690 Ace:

1. Intel stock, listed as "boxed cooler", at 253W/320A
2. Tower air cooler, 288W/512A (I use this for 24/7 as it's good balance between performance and thermals)
3. Water cooler, fully unrestrained at 4096W/512A

I bet a lot of people with dinky AIOs pick #3 thinking it'll be fine as "they're on water anyway"

From what I read anything above 253W may damage the chips, even when temp is kept low with delid and direct die cooling.
So yeah, better use the Boxed Cooler setting.
Heck I set 150W TDP since the first day I got the 13700K

 

[efikkan]

From what I read anything above 253W may damage the chips, even when temp is kept low with delid and direct die cooling.
So yeah, better use the Boxed Cooler setting.
Heck I set 150W TDP since the first day I got the 13700K

Even 253W beyond 56 seconds (with sustained load) will likely shorten the chip's lifespan.
What kind of sustained clock speeds do you achieve at a 150W power limit?

I'm leaning more and more towards "HEDT" for productive works, even though I don't need many cores. Xeon-W tops out at 4.8 GHz. The IO and memory bandwidth is compelling, but I wish they could reach ~5.5 GHz at stock, then the choice would have been much easier. (things may change though with Zen 5, Arrow Lake and upcoming Xeons)

 

[Dr. Dro]

From what I read anything above 253W may damage the chips, even when temp is kept low with delid and direct die cooling.
So yeah, better use the Boxed Cooler setting.
Heck I set 150W TDP since the first day I got the 13700K

Don't think this is significant, what causes electromigration isn't wattage, it's the current. In a possible oversimplification of the concept, higher voltages mean higher current, and higher temperatures mean the material becomes more flexible. That's what's Black's equation is meant to calculate

Black's equation - Wikipedia

en.wikipedia.org

Remember that Intel's stock values are extremely conservative. 253 W is not excessive for these chips. It'll probably be able to take 400 W constantly without any problems.

 

[dgianstefani]

Don't think this is significant, what causes electromigration isn't wattage, it's the current. In a possible oversimplification of the concept, higher voltages mean higher current, and higher temperatures mean the material becomes more flexible. That's what's Black's equation is meant to calculate

Black's equation - Wikipedia

en.wikipedia.org

Remember that Intel's stock values are extremely conservative. 253 W is not excessive for these chips. It'll probably be able to take 400 W constantly without any problems.

Yes, it's the voltage that's the issue. High wattage is fine as long as it's cooled.

 

[EsliteMoby]

Does this effect laptop CPU as well? 13980hx is pretty much the same chip as the desktop 13900K.

 

[Macro Device]

Does this effect laptop CPU as well? 13980hx is pretty much the same chip as the desktop 13900K.

No. You will burn your pants down way before this issue kicks in.

A happy i5-12400F (CPU @ unlimited wattage but with no OC + RAM @ XMP-3200 C17) user dialling here. Everything is rock solid. Perhaps a wee slow by someone's standards but as my only significant CPU load is gaming and I'm a 60 FPS gamer, 12400F will be overqualified for a while.

Wondering how hard my CPU will be destroyed by i5-19400 or whatever they will be called. Agree with everyone who states the mobo makers must never default to insane OC on their motherboards.

 

[nguyen]

Even 253W beyond 56 seconds (with sustained load) will likely shorten the chip's lifespan.
What kind of sustained clock speeds do you achieve at a 150W power limit?

I'm leaning more and more towards "HEDT" for productive works, even though I don't need many cores. Xeon-W tops out at 4.8 GHz. The IO and memory bandwidth is compelling, but I wish they could reach ~5.5 GHz at stock, then the choice would have been much easier. (things may change though with Zen 5, Arrow Lake and upcoming Xeons)

Well sustained clocks are dependent on what type of workload, for me I only play games at 4K so 5.4-5.5ghz are sustainnable, for Cinebench24 it's 4.6ghz P-cores.

 

[Waldorf]

was asking myself, how long until intel has to do 'the same" as amd (manual settings),
instead of AUTO,
seeing how virtually every brand had different ideas on PBO limits for ryzen,
or my Gb Aorus running at 101 BCLK with way too much soc V, using auto.

 

[chrcoluk]

We need a statement from Intel.

Im curious to see what boards are affected. I've not built with the 14th gen, but the 12th and 13th gen builds have been rock solid.

Fingers crossed here, my 13700k for what I have used it for has been brilliant, still undervolted. No instability at all.

The CPU is configured to run efficiently as possible, I spent some time on power profiles optimising how it adjusts clocks, uses cores etc. and day to day have HT disabled (not in bios but dynamically via windows power settings).

Motherboard manufacturers try to win benchmarks by raising various voltages and power limits above the default. Stock should mean stock.

Plus the odd vendor *asrock cough* setting tjmax to 120C on Auto.

Doesn't surprise me at all since they're basically factory-overclocked to their maximum and Intel themselves have said that 100C temps are totally normal.


I don't want my PC to be an unstable heater with a fire hazard power connector.

Spoiler

Shouldn't answer to a troll but that was just a too tempting bait

If it ends up being temp related (cause of voltage degradation), it would be interesting. They have an employee on a popular youtube channel telling everyone 100C is no problem for day to day usage.

This is my new strategy. In games I'm averaging ~50 W on the 13600KF and ~130 W on the undervolted 4070. And I play at a capped 4K60 (usually with upscaling), so utilization is almost never maxed out.

And I don't buy games at launch anymore. They run much better after a few months worth of patches, and you get a nice discount. If a game is too heavy for my config, I'll play it after the next upgrade.

Yeah typically I have about 200-300W total draw across entire UPS whilst gaming, this includes monitor and some non PC devices. So I am running no where near limits, and this isnt by accident, I remember posting on here a screenshot of what happened on the dune spice wars map screen when I switched to my low power GPU profile and it knocked almost 100W of the power draw of GPU without any performance loss. To me, undervolting and power limiting (voltage curve optimising on GPU) is standard process now, the manufacturers just dont care on this side anymore.

 

[Voodoo Rufus]

I'm definitely going to flash to the newest BIOS this evening and check the CPU power settings. My errors are black screen crashes necessitating a hard reboot.

Well, the BIOS changes didn't do anything. Running less power with MCE disabled but no change in stability.

I started to investigate with the basics on the hardware side and found that the PCIE riser wasn't seated fully. Doh....

Now it's 100% stable, but getting some artifacting on some game menus. No crashes though. More investigation needed but it's a start.