Question DEGRADING Raptor lake CPUs

[Kocicak]

I noticed some reports about degrading i9 13900K and KF processors.

I experienced this problem myself, when I ran it at 6 GHz, light load (3 threads of Cinebench), at acceptable temperature and non extreme voltage. After only few minutes it crashed, and then it could not run even at stock setting without bumping the voltage a bit.

I was thinking about the cause for this and I believe the problem is, that people do not appreciate, how high these frequencies are and that the real comfortable frequency limit of these CPUs is probably at something like 5500 or 5600 MHz. These CPUs are made on a same process (possibly improved somehow) on which Alder lake CPUs were made. See the frequencies 12900KS runs at. The frequency improvement of the new process tweak may not be so high as some people presume.

Those 13900K CPUs are probably highly binned to be able to find those which contain some cores which can reliably run at 5800 MHz. Some of the 13900K probably have little/no OC reserve left and pushing them will cause them to degrade/break.

The conclusion for me is that the best you can do to your 13900K or 13900KF is to disable the 5800 MHz peak, which will allow you to offset the voltage lower, and then set all core maximal frequency to some comfortable level, I guess the maximum level could be 5600 MHz. With lowered voltage this frequency should be gentler to the processor than running it at original 5500 MHz at higher voltage. You can also run it at lower frequencies, allowing for even higher voltage drop, but then the CPU is slowly loosing its sense (unless you want some high efficiency CPU intended for heavy multithread loads).

Running it with some power consumption limit dependent on your cooling solution to keep the CPU at sensible temperature will help too for sure.

 

[del42sa]

I'm not sure we know that...

Depends on the workload, doesn't it?

All in all, although some apps were noticeably hurt by the patch, on average 1% down is only a slight drop. It cannot be said that it would not have a negative effect on performance, but on average it is not noticeable

 

[Kocicak]

I was testing out my 12700K in Intel XTU, trying to achieve higher ST score in CPU-Z. I think it went up to 851 with maybe 5.2 or 5.3 GHz. Afterwards, the RX 6800 developed glitching (weird texture flashing) in Quake 2 RTX, even with a restart. It took some time for the glitching to go away. Someone said that PCI Express connectivity is affected when the processor goes above 5 GHz. I believe that now.

If true, thay must have somehow fixed this in Raptor lake CPUs.

All in all, although some apps were noticeably hurt by the patch, on average 1% down is only a slight drop. It cannot be said that it would not have a negative effect on performance, but on average it is not noticeable

Instead of doing the right thing, Intel is desperately trying to maintain the frequencies and the "Stock performance", because if they lowered the frequencies as they should, they would create a different product with lower performance than what the customers expected when they bought the chips.

Intel will need to figure how to deal with this lower performing product, because keeping the breakneck frequencies is simply not possible.

These CPUs are so nice, cool and efficient with say 5/4 GHz frequency limits, EVERYTHING BAD about these chips would solve just by decreasing the frequencies. The fix is not complicated at all, it is extremely easy!

But Intel wants to keep those awful high frequencies like stubborn dumb donkeys.

 

[igor_kavinski]

If true, thay must have somehow fixed this in Raptor lake CPUs.

Don't think so. You would need to do what I did in Intel XTU and then run Q2RTX to confirm.

Here's the settings I told XTU to try to achieve:

P-cores
1 core: 5.4 GHz
2 core: 5.3 GHz
3 core: 5.2 GHz
4 core: 5.1 GHz
5 core: 5.0 GHz
6 core: 4.9 GHz
7 core: 4.9 GHz
8 core: 4.9 GHz

E-cores
1 core: 4.2 GHz
2 core: 4.1 GHz
3 core: 4.0 GHz
4 core: 3.9 GHz

PL1=PL2=275W

IccMax= 295A

TVB Enabled

DDR5-7000 CL34

I think the ring bus or whatever overheated and it took some time (more than 10 minutes) for it to get back to normal for the graphical glitching to go away.

 

[Hulk]

If true, thay must have somehow fixed this in Raptor lake CPUs.



Instead of doing the right thing, Intel is desperately trying to maintain the frequencies and the "Stock performance", because if they lowered the frequencies as they should, they would create a different product with lower performance than what the customers expected when they bought the chips.

Intel will need to figure how to deal with this lower performing product, because trying to keep the breakneck frequencies is simply not possible.

These CPUs are so nice, cool and efficient with say 5/4 GHz frequency limits, EVERYTHING BAD about these chips would solve just by decreasing the frequencies. The fix is not complicated at all, it is extremely easy!

But Intel wants to keep those awful high frequencies like stubborn dumb donkeys.

Part of the problem is knowing exactly what stock frequencies are.
For example, for the 14900K the all core P frequency is specified by Intel as 5.6GHz assuming adequate cooling. 5.8GHz or even 6GHz turbo for 1 or two cores is also in the spec.

Problem #1 - Massive voltage is required for most parts to achieve 5.8GHz and even more for 6GHz. This voltage is generally on the order of 1.5 or more volts, which I believe is simply too much for a CPU to be long-lived on the current process for Raptor Lake. Even if 5.8 or 6.0 GHz is hit now and then by a few cores, this (over) voltage is supplied to all cores. Not good for longevity and does virtually nothing for real world performance.

Problem #2 - 5.6GHz all core under heavy load with HT enabled also requires quite a bit of voltage for most Raptor silicon, probably around 1.4V, which I think it too much if you expect a long life from your chip. 5.6GHz IS okay I think from some Raptors with really good cooling but it depends on the voltage you need to make this happen in a stable manner.

The "fix" for Intel is to go back in time and eliminate 1/2 core boost and reduce all-core spec to about 5.4GHz with HT on. As I have written previously my 14900K runs well at 5.5GHz, no HT. HT on would require more voltage and return virtually no more performance as I can rely on the 16 E cores rather than the logical cores.

There is nothing wrong with Raptor silicon. It was over binned and over specified.

Or add some tiny print to the specs stating something like "Intel only guarantees ARK frequencies when the processor is operating below 30C."

Just kidding of course.

 

[igor_kavinski]

As I have written previously my 14900K runs well at 5.5GHz, no HT. HT on would require more voltage and return virtually no more performance as I can rely on the 16 E cores rather than the logical cores.

What frequency/voltage do your E-cores run at?

 

[Thunder 57]

There is nothing wrong with Raptor silicon. It was over binned and over specified.

Then why did they include 65W CPUs? Surely they don't boost to a level where they need extra voltage?

 

[coercitiv]

There is nothing wrong with Raptor silicon. It was over binned and over specified.

... and for about one year it was subject to oxidation, which is the perfect embodiment of "something wrong". So we'd have to say there's nothing wrong with Raptor silicon NOW, which implies you still have trust in a company that "failed" to tell customers they were buying products prone to accelerated aging.

Then why did they include 65W CPUs? Surely they don't boost to a level where they need extra voltage?

The included 65W TDP CPUs were the ones boosting higher than the 13600K and 14600K. Also, the slowest 65W TDP CPU in the list, the 13700, has a PL2 of 219W

 

[Hulk]

Part of the problem is knowing exactly what stock frequencies are.
For example, for the 14900K the all core P frequency is specified by Intel as 5.6GHz assuming adequate cooling. 5.8GHz or even 6GHz turbo for 1 or two cores is also in the spec.

Problem #1 - Massive voltage is required for most parts to achieve 5.8GHz and even more for 6GHz. This voltage is generally on the order of 1.5 or more volts, which I believe is simply too much for a CPU to be long-lived on the current process for Raptor Lake. Even if 5.8 or 6.0 GHz is hit now and then by a few cores, this (over) voltage is supplied to all cores. Not good for longevity and does virtually nothing for real world performance.

Problem #2 - 5.6GHz all core under heavy load with HT enabled also requires quite a bit of voltage for most Raptor silicon, probably around 1.4V, which I think it too much if you expect a long life from your chip. 5.6GHz IS okay I think from some Raptors with really good cooling but it depends on the voltage you need to make this happen in a stable manner.

The "fix" for Intel is to go back in time and eliminate 1/2 core boost and reduce all-core spec to about 5.4GHz with HT on. As I have written previously my 14900K runs well at 5.5GHz, no HT. HT on would require more voltage and return virtually no more performance as I can rely on the 16 E cores rather than the logical cores.

There is nothing wrong with Raptor silicon. It was over binned and over specified.

Or add some tiny print to the specs stating something like "Intel only guarantees ARK frequencies when the processor is operating below 30C."

Just kidding of course.

4.2, I don't know the voltage.

 

[Hulk]

Then why did they include 65W CPUs? Surely they don't boost to a level where they need extra voltage?

Perhaps they are also over volting?

I'm not saying something isn't going on here but I do know that many people have been and are slamming massive voltage into these chips to push them way beyond stock. There is now way for Intel to discern valid failures from forced failures and many people will take advantage. I'm trying to remain fair.

My 13900K degraded on auto BIOS settings. My fault? Asus's fault? Intel's fault? Combination of all three? Intel made it right for me with a full refund.

 

[evident]

What's the best way to underclock / undervolt the cpu? I wouldn't mind giving up 5-10% of performance or even more if it means i'll have a stable NAS / home server appliance. I dont need to squeak out every last bit of performance in my use case.

 

[igor_kavinski]

There is now way for Intel to discern valid failures from forced failures and many people will take advantage. I'm trying to remain fair.

But Intel failed at protecting people from themselves. We aren't hearing about people killing their Ryzens while trying to eek the maximum performance out of them.

Intel has dropped the ball so badly that even game servers with low clocked DDR5 have been degrading for more than a year and everyone involved (server vendors, data center technicians, end users, game developers) have been pulling their hair like crazy trying to isolate the issue and Intel just kept quiet, hoping they would give up and lower their settings to almost "Celeron" levels to not let their investment go to complete waste. Well, they were wrong. People were wayyy too patient with them and the dam broke and now they deserve everything that's happening to them. It was their job to make sure that their CPUs were fit for purpose. These issues didn't happen with Alder Lake. They messed up with Raptor Lake and instead of announcing a recall or releasing a public advisory that they internally discovered some issues and these are the guidelines everyone should follow, they just kept quiet and even worse, resorted to denying RMAs for months upon months! They conducted themselves in the worst possible way. For a company of their stature, it was just plain shameful behavior.

 

[Hulk]

But Intel failed at protecting people from themselves. We aren't hearing about people killing their Ryzens while trying to eek the maximum performance out of them.

Intel has dropped the ball so badly that even game servers with low clocked DDR5 have been degrading for more than a year and everyone involved (server vendors, data center technicians, end users, game developers) have been pulling their hair like crazy trying to isolate the issue and Intel just kept quiet, hoping they would give up and lower their settings to almost "Celeron" levels to not let their investment go to complete waste. Well, they were wrong. People were wayyy too patient with them and the dam broke and now they deserve everything that's happening to them. It was their job to make sure that their CPUs were fit for purpose. These issues didn't happen with Alder Lake. They messed up with Raptor Lake and instead of announcing a recall or releasing a public advisory that they internally discovered some issues and these are the guidelines everyone should follow, they just kept quiet and even worse, resorted to denying RMAs for months upon months! They conducted themselves in the worst possible way. For a company of their stature, it was just plain shameful behavior.

Good points, I have to admit.

 

[Kocicak]

These CPUs are so nice, cool and efficient with say 5/4 GHz frequency limits, EVERYTHING BAD about these chips would solve just by decreasing the frequencies. The fix is not complicated at all, it is extremely easy!

One local retailer sells 20 core 14700K (with 5 year warranty ???) now for less than AMD asks for 8 core 9700X, it is so tempting... If I was building a new PC (and did not have 14900K already), I would have jumped on it.

 

[coercitiv]

If I was building a new PC (and did not have 14900K already), I would have jumped on it.

After all, they're very fast, efficient and reliable CPUs, aren't they?

 

[Ranulf]

Well, in theory if you run it at 5.0-5.2ghz it should be stable. Should be...will be...I know it will be fine. Just fine. We're fine here, how are you?

 

[Rigg]

Well, in theory if you run it at 5.0-5.2ghz it should be stable. Should be...will be...I know it will be fine. Just fine. We're fine here, how are you?

Boring conversation.

 

[Kocicak]

After all, they're very fast, efficient and reliable CPUs, aren't they?

They are, once you set them that way. They are very nice CPUs, you just cannot allow Intel to destroy what they made.

 

[VirtualLarry]

https://www.newegg.com/promotions/nepro/23-1322/index.html?cm_sp=Head_Navigation-_-Under_Search_Bar-_-CPU+Trade+In&icid=776887

Trade in your bad 13th/14th Gen "KS" CPUs to Newegg for trade-in credit.

This does seem slightly unethical, however.

You can look at that two ways: Newegg, a major retailer of online electronic goods, gets made whole by the mfg (Intel) for all of the "bad stock" that the two of them cull from the market (with Intel maybe kicking in some marketing $$$ for the program), or that Newegg won't get any recourse from Intel, but they have the legal ability in the T&Cs to claw back any credit or monies received for the supposedly-faulty hardware from you.

 

[igor_kavinski]

Trade in your bad 13th/14th Gen "KS" CPUs to Newegg for trade-in credit.

Don't think it'll be that easy. What if their internal benchmark suite is good enough to break a weak/degraded CPU?

 

[lakedude]

Trade in your bad 13th/14th Gen "KS" CPUs to Newegg for trade-in credit.

Trade in credit, not full value...

 

[yottabit]

The included 65W TDP CPUs were the ones boosting higher than the 13600K and 14600K. Also, the slowest 65W TDP CPU in the list, the 13700, has a PL2 of 219W

A 219 Watt, 65 Watt CPU. What a time to be alive

But Intel failed at protecting people from themselves. We aren't hearing about people killing their Ryzens while trying to eek the maximum performance out of them.

Intel has dropped the ball so badly that even game servers with low clocked DDR5 have been degrading for more than a year and everyone involved (server vendors, data center technicians, end users, game developers) have been pulling their hair like crazy trying to isolate the issue and Intel just kept quiet, hoping they would give up and lower their settings to almost "Celeron" levels to not let their investment go to complete waste. Well, they were wrong. People were wayyy too patient with them and the dam broke and now they deserve everything that's happening to them. It was their job to make sure that their CPUs were fit for purpose. These issues didn't happen with Alder Lake. They messed up with Raptor Lake and instead of announcing a recall or releasing a public advisory that they internally discovered some issues and these are the guidelines everyone should follow, they just kept quiet and even worse, resorted to denying RMAs for months upon months! They conducted themselves in the worst possible way. For a company of their stature, it was just plain shameful behavior.

Yeah but AMD didn’t stop that one guy from cracking his 9950x IOD in half trying to de-lid it so…. #bothsides

So is the consensus 14900k is stable-ish at 5.5 GHz with HT disabled? And with the new microcode update. I’ll admit the Zen5 disappointment has me considering Raptor Lake, maybe buying somebody’s CPU or rig at firesale prices and tuning it for stability. Probably smarter to wait for Arrow Lake I guess.

 

[igor_kavinski]

So is the consensus 14900k is stable-ish at 5.5 GHz with HT disabled? And with the new microcode update. I’ll admit the Zen5 disappointment has me considering Raptor Lake, maybe buying somebody’s CPU or rig at firesale prices and tuning it for stability. Probably smarter to wait for Arrow Lake I guess.

9800X3D will still be better. Hopefully, if gaming is your only use case.

If you just want something for CB R23/24 room heating during winter and don't care about AVX-512, yeah, better wait for Arrow Lake.

 

[Kocicak]

So is the consensus 14900k is stable-ish at 5.5 GHz with HT disabled?

I think 5.5 GHz is way too much.

Probably smarter to wait for Arrow Lake I guess.

I think Arrow lake will bring more features than a raw computing power, possibly better 1 thread and gaming performance.

Better performance can be decreased simply by lower frequencies. I doubt Arrow lake will be clocked so high as Raptor lake.

 

[coercitiv]

So is the consensus 14900k is stable-ish at 5.5 GHz with HT disabled? And with the new microcode update. I’ll admit the Zen5 disappointment has me considering Raptor Lake, maybe buying somebody’s CPU or rig at firesale prices and tuning it for stability. Probably smarter to wait for Arrow Lake I guess.

Raptor Lake is a no-no at this point. Even if you don't want any Zen, waiting for Arrow Lake is the rational thing to do. Based on what I've seen, ARL should have a compelling offering in the i7 and i5 segments. It will regress in max clocks for the i9 equivalent, but the i7 and i5 equivalent look like they'll hold the clocks vs. 14600K/14700K and provide a decent performance uplift. The i5 equivalent in particular may behave very nicely in it's price bracket.

That being said, waiting for ARL also allows you to let the dust settle after this Zen 5 launch. Let's see how Zen 5 pricing evolves until then, and maybe take a look at 9800X3D as well. AMD's cost structure for Zen 5 allows them to be more flexible than Intel with ARL.

 

[igor_kavinski]

I think 5.5 GHz is way too much.

Hulk has no issue, with HT off. Either that's totally safe or he ended up with a golden sample.

Though he IS using DDR4 so there isn't enough bandwidth for the cores to get swamped with data so they are probably resting a lot between memory requests.

@Hulk , in case you are curious how much you might be losing (I don't know what LGA1700 DDR4 usually scores) but my 12700K with DDR5-7000 is hitting something like 9300-9400 in CPU-Z MT.