News Puget says its Intel chips failures are lower than Ryzen failures — retailer releases failure rate data, cites conservative power settings

Dr3ams

Respectable
Sep 29, 2021
226
211
1,960
If Puget's statistics are true, then where's the AMD customer posts and outrage then.

Since 1995, I've never had an AMD or Intel CPU fail on me. Had a few GPUs, hard drives and PSUs break though.
 

Pierce2623

Prominent
Dec 3, 2023
243
207
460
My office had Puget CAD/CAE workstations with a consumer level Intel 13900k chip and two rtx a4000s. Their settings perform significantly below just slapping a 13900k on an Asus z790 in multithreaded workloads. In fact, they don’t even stay at PL2 indefinitely.
 

Gururu

Proper
Jan 4, 2024
116
76
170
This is most certainly a case where you would expect enthusiasts to be the most vocal even though they are an arguably insignificant portion of sales. The CPU issue is much less a problem than Intel’s failure to stay lean and mean in the market. They are top heavy and I worry about the quality of engineers vs. yesteryear. BUT they are an American company, perhaps our flagship for microprocessors and if they go down, I doubt AMD or any other will have the resources or interest to carry that weight. I am hoping they can survive this, not the way IBM did, but the way Apple did.
 
  • Like
Reactions: graham006

sdedalus83

Distinguished
Apr 2, 2007
23
0
18,510
My office had Puget CAD/CAE workstations with a consumer level Intel 13900k chip and two rtx a4000s. Their settings perform significantly below just slapping a 13900k on an Asus z790 in multithreaded workloads. In fact, they don’t even stay at PL2 indefinitely.
Don’t they run review systems exclusively at base clock speeds?
 

graham006

Distinguished
May 3, 2009
19
18
18,515
High-end PC builder Puget Systems say that its data indicate a lower than expected failure rate for its Intel builds, compared to what’s widely reported.

Puget says its Intel chips failures are lower than Ryzen failures — retailer releases failure rate data, cites conservative power settings : Read more
I have had two of the Intel CPUs with supposed problems. (reports saying that all 13th and 14th gen CPUs with TDP 65W and higher _could_ be affected). The first CPU was a core i5-13500 in an ASUS ROG STRIX ITX motherboard. I never overclocked the CPU but did overclock RAM to 5600. Ran this CPU for 1 year with no issues and sold to a friend who currently has it and says it's fine. I upgraded that CPU to a Core i7-14700. Again no overclocking except for RAM. I have applied all BIOS updates provided by ASUS to this system. Again, rock solid performance... After hearing all the news about CPU instability I decided to run some stress tests after updating to most recent BIOS. My results:

Cinebench R24 - 1 hour stress. No problems. multi-core score was 1579.
CPU-Z stress test. Ran for 1 hour. No issues. CPU temp went as high as 100C for a bit but leveled out in the high 60s (C).

I did a RESNET50 training on CIFAR data set on CPU (normally would do this on GPU) and that worked fine as well and stressed the CPU for about 50 minutes. Again no issues.

I think most of the problem is people are pushing their CPUs way beyond what they are designed to do. Don't overclock. If you must, overclock your memory. It's a lot cheaper to replace.
 
  • Like
Reactions: NinoPino

YSCCC

Prominent
Dec 10, 2022
96
42
560
My office had Puget CAD/CAE workstations with a consumer level Intel 13900k chip and two rtx a4000s. Their settings perform significantly below just slapping a 13900k on an Asus z790 in multithreaded workloads. In fact, they don’t even stay at PL2 indefinitely.
This is really interesting, may I ask did you get a metric like Cinebench R23/2024 for a reference point? when they stated that they go on the conservative side and the percentage is at what they reported, that what performance of their system is, if it is stable an reliable at the cost of say, 10% or more copmared to the extreme profile is now, it is not really a fair comparison for consumers, if they release the 13900k getting 30k R23 I bet they will get near as many sales as they did.

Also funny is the total number of systems sent out for the % comparison, if both stayed reasonably low, 10 computers out of 100 failing randomly maybe due to shipment or other parts defect will be 10%, but 10 out of 500 will be 2.5%
 

Ogotai

Reputable
Feb 2, 2021
339
222
5,060
What is the source for that information? Over 50% failure rate? Are you an AMD employee?
not to mention all the other posts about imtels 13th and 14th gens crashing and the instability issues they have
 

TheHerald

Prominent
Feb 15, 2024
698
211
760
I have had two of the Intel CPUs with supposed problems. (reports saying that all 13th and 14th gen CPUs with TDP 65W and higher _could_ be affected). The first CPU was a core i5-13500 in an ASUS ROG STRIX ITX motherboard. I never overclocked the CPU but did overclock RAM to 5600. Ran this CPU for 1 year with no issues and sold to a friend who currently has it and says it's fine. I upgraded that CPU to a Core i7-14700. Again no overclocking except for RAM. I have applied all BIOS updates provided by ASUS to this system. Again, rock solid performance... After hearing all the news about CPU instability I decided to run some stress tests after updating to most recent BIOS. My results:

Cinebench R24 - 1 hour stress. No problems. multi-core score was 1579.
CPU-Z stress test. Ran for 1 hour. No issues. CPU temp went as high as 100C for a bit but leveled out in the high 60s (C).

I did a RESNET50 training on CIFAR data set on CPU (normally would do this on GPU) and that worked fine as well and stressed the CPU for about 50 minutes. Again no issues.

I think most of the problem is people are pushing their CPUs way beyond what they are designed to do. Don't overclock. If you must, overclock your memory. It's a lot cheaper to replace.
Most people won't have any issues - obviously. Anecdotes of 2 cpus are not evidence.

It's clear - even by puget's numbers - that failure rates on 13th and 14th gen are higher than they should be.
 
I think most of the problem is people are pushing their CPUs way beyond what they are designed to do. Don't overclock. If you must, overclock your memory. It's a lot cheaper to replace.
Except that these are companies that demand reliability at stock settings yet are seeing 50% failure rates. Hedge funds and investment firms do not overclock because they rely on algorithms to make buy and sell orders on a millisecond scale.

Also, overclocking memory is overclocking your CPU. The memory controller is on the CPU and its frequency changes with the ram’s frequency, and you can very much blow it with too high voltage or current for stability. Even leaving the mem controller voltage alone but allowing too high of a RAM DDR voltage can have bad effects on the memory controller.
 

wakuwaku

Distinguished
Nov 17, 2008
26
12
18,535
You know, pugets CPU failure rates seems incredibly high for all CPUs, In my experience, it's extremely unusual for a CPU to fail. I have in fact never experienced a cpu fail despite my experience with many hundred machines. CPU failure measured in percentages? I don't for a second believe they're measuring what we believe they are.
many hundreds? they deal with thousands of pcs. come back when you reach their level please.

also there is a decent amount of failures that happen before they send them out to customers (in the shop). Never got a lemon CPU before? Good for you.

I bought a 7950x earlier this year that was a lemon. Totally didn't boot out of the box. Sent it back to the shop, they tested it in their own system and confirmed it was a lemon. Sent it back to AMD, got my new CPU back in ~2 weeks and its now working happily doing its thing.
 

TheHerald

Prominent
Feb 15, 2024
698
211
760
This is really interesting, may I ask did you get a metric like Cinebench R23/2024 for a reference point? when they stated that they go on the conservative side and the percentage is at what they reported, that what performance of their system is, if it is stable an reliable at the cost of say, 10% or more copmared to the extreme profile is now, it is not really a fair comparison for consumers, if they release the 13900k getting 30k R23 I bet they will get near as many sales as they did.
The numbers are there for all to see, they publish reviews with their settings.

Yes, the 13900k scores around 30-31k (10 minute test)with the settings they ship them with, but on the same note, the 7950x scores 33k. The difference is 10% in performance, but the zen 4 parts have twice the failure rate.

Also their workstations aren't really targeted at rendering workloads so these numbers are kinda useless. They are targeted at photoshop / premiere / lightroom etc, tasks that the 13900k (with the aforementioned settings) is faster than the 7950x anyways.

Here is their review

 

YSCCC

Prominent
Dec 10, 2022
96
42
560
The numbers are there for all to see, they publish reviews with their settings.

Yes, the 13900k scores around 30-31k (10 minute test)with the settings they ship them with, but on the same note, the 7950x scores 33k. The difference is 10% in performance, but the zen 4 parts have twice the failure rate.

Also their workstations aren't really targeted at rendering workloads so these numbers are kinda useless. They are targeted at photoshop / premiere / lightroom etc, tasks that the 13900k (with the aforementioned settings) is faster than the 7950x anyways.

Here is their review

Well, as said, the base numbers of systems sold and failures arn't nearly comparable in their lineup, but both still well within acceptable line, what is interesting is, IF (and a big IF) it takes this low performance for the 13900k or 14900k, they are not living up to what they are advertised to do in consumer level. Cases closed, this also makes sense for that review sites can no longer recommend intel for some time, since nobody know how conservative in every power matrix is finally safe for long term, and how low the overaall performance will be after the fix
 

YSCCC

Prominent
Dec 10, 2022
96
42
560
Summary: Puget has its own special BIOS that is more conservative than default settings. Intel 13th and 14th chips are still failing at a higher rate than 12th gen.
To be fair though, they will likely have the same conservative bios for 12th gen, so being the same tick tok cycle it just makes sense with the lower clocked ones lives better
 
  • Like
Reactions: ThomasKinsley

TheHerald

Prominent
Feb 15, 2024
698
211
760
Well, as said, the base numbers of systems sold and failures arn't nearly comparable in their lineup, but both still well within acceptable line, what is interesting is, IF (and a big IF) it takes this low performance for the 13900k or 14900k, they are not living up to what they are advertised to do in consumer level. Cases closed, this also makes sense for that review sites can no longer recommend intel for some time, since nobody know how conservative in every power matrix is finally safe for long term, and how low the overaall performance will be after the fix
But neither does AMD, since even with this low performance for the 7950x it still fails at over twice the rate of Intel, right?