Intel says their 13th/14th Gen instabilities are from 'elevated operating voltages stemming from a microcode algorithm'

[porina]

We're really lacking any indication of how many CPUs out there might be at risk. With very much a gamer hat on, I turn to Steam Hardware Survey again. Unfortunately they don't do it by model, but by clock and cores. I'll focus on cores.

 

Steam Hardware Survey - core count share - June 2024

 

24 cores 1.78%
14900

13900

24 cores is easy, as it includes the '900 parts. It could also include server/workstation parts but we can make the assumption that will not be significant in a gaming environment.

 

20 cores 0.76%
14700

Just the 14700 variations here. 

 

16 cores 3.75%
13700
12900

+many others

I think we'll have to drop 16 cores as a data point since there will be many CPUs with that count, including 7950 and 5950 for example. Only 13700 family will be potentially affected.

 

14 cores 4.15%
14600
14500
13600 (non-K Alder Lake)
13500 (Alder Lake)

Again this gets a little messy since the lower 13th gen were rebadged Alder Lake. 

 

10 cores 5.99%
14400
13400 (Alder Lake)
12600K
10900
10850

Another messy one as there were a few 10 core offerings with only 14th gen being Raptor Lake.

 

 

 

Now I done this, I'm not sure if it helps much. The main interesting point is that 13900+14900 is about 2x 14700. I would have expected 14700 to sell far more than '900. Guess gamers do go to the top models. Also I had forgotten that lower end 13th gen (13600 non-K and lower) were Alder lake refreshes so presumably would not be affected.

 

Would welcome any thoughts on this. Have I missed anything?

[Kisai]

35 minutes ago, porina said:

 

Now I done this, I'm not sure if it helps much. The main interesting point is that 13900+14900 is about 2x 14700. I would have expected 14700 to sell far more than '900. Guess gamers do go to the top models. Also I had forgotten that lower end 13th gen (13600 non-K and lower) were Alder lake refreshes so presumably would not be affected.

 

Would welcome any thoughts on this. Have I missed anything?

The parts that were failing in those servers were like 3/4ths just 13900K/14900K and KS parts.

 

I'd assume that the Turbo boost issue has more to do with voltage values in a table somewhere that might be missing rows, so unless the CPU is being forced to run at "Intel Default" it's probably using a value that applies to the 12th gen on the 13th and 14th gen model.

 

Interesting to note what Intel says is safe:

https://community.intel.com/t5/Gaming-on-Intel-Processors-with/CPU-Voltage/td-p/1442329

Quote

The Operating Voltage for the 13th Generation Intel® Core™ Processors will be up to a maximum of 1.72V, in this case, it is ok if the voltages are going around 1.2V to 1.5V.

 

https://www.intel.com/content/www/us/en/support/articles/000092345/processors.html

Quote

Like I'm sure the blame has to be shared between Intel and the third party motherboard/server manufacturers for trying to squeeze performance in ways that shouldn't be. Like I don't trust the third party MB manufacturers to ever use "default" settings from Intel, and I don't trust Intel (or AMD, because AMD has been historically terrible about this) to have enough of a safety buffer. Like the reason the "K/KS" parts exist in the first place is because you are paying more to be able to use more of that safety buffer at the cost of more energy use. 

 

[dogwitch]

bit of  drop news if you miss it. wendell dig some digging on this.

is mobo manf and asus love doing this.

is force certain chips with a under a  max pull draw of  less then 100 watts.

 to be pushed  by default to north of 300 watts. in boost alg on mobos.

[RejZoR]

17 hours ago, porina said:

We're really lacking any indication of how many CPUs out there might be at risk. With very much a gamer hat on, I turn to Steam Hardware Survey again. Unfortunately they don't do it by model, but by clock and cores. I'll focus on cores.

 

Steam Hardware Survey - core count share - June 2024

 

24 cores 1.78%
14900

13900

24 cores is easy, as it includes the '900 parts. It could also include server/workstation parts but we can make the assumption that will not be significant in a gaming environment.

 

20 cores 0.76%
14700

Just the 14700 variations here. 

 

16 cores 3.75%
13700
12900

+many others

I think we'll have to drop 16 cores as a data point since there will be many CPUs with that count, including 7950 and 5950 for example. Only 13700 family will be potentially affected.

 

14 cores 4.15%
14600
14500
13600 (non-K Alder Lake)
13500 (Alder Lake)

Again this gets a little messy since the lower 13th gen were rebadged Alder Lake. 

 

10 cores 5.99%
14400
13400 (Alder Lake)
12600K
10900
10850

Another messy one as there were a few 10 core offerings with only 14th gen being Raptor Lake.

 

 

 

Now I done this, I'm not sure if it helps much. The main interesting point is that 13900+14900 is about 2x 14700. I would have expected 14700 to sell far more than '900. Guess gamers do go to the top models. Also I had forgotten that lower end 13th gen (13600 non-K and lower) were Alder lake refreshes so presumably would not be affected.

 

Would welcome any thoughts on this. Have I missed anything?

I usually go with top end CPU and higher end motherboard and basically not change it until I buy whole new platform. I usually change 2-3 graphic cards in this time. I've been doing this for basically 25 years. Only time I've broke this rule was upgrade from Core 2 Duo E4300 to E5200 and relatively recent upgrade from 5800X to 5800X3D.

 

If you're mostly gamer, CPU's don't really change as much as graphic cards do and even those have been stagnating in recent years in terms of being a requirement to play new AAA games. Remember the old days where almost every major game release almost required you to buy new graphic card or suffer dramatically worse graphics or framerate. That hasn't been the case for a long time now as new games come out and I can play them at max settings easily.

[sounds]

The Verge reading between the lines of the article on Toms Hardware, and just saying it. It's notable because now a news outlet is saying what we've all been saying for a while now. They've put their name behind it. Well... it is just The Verge so I don't know that they have a lot of reputation left to lose. 

 

"There is no fix for Intel’s crashing 13th and 14th Gen CPUs — any damage is permanent"

 

https://www.theverge.com/2024/7/26/24206529/intel-13th-14th-gen-crashing-instability-cpu-voltage-q-a

 

[Moonzy]

I assume this doesn't affect 12th gen?

[Bitter]

2 hours ago, Moonzy said:

I assume this doesn't affect 12th gen?

Doesn't affect them yet.

[starsmine]

6 minutes ago, Bitter said:

Doesn't affect them yet.

yet? It just does not effect them. They are on the same node. However, they dont have the same power domains. they dont have the same eTVB, they dont have the same microcode. 

If we find 12th gen to have issues at some point, it will NOT be the same issue. 

[Bitter]

43 minutes ago, starsmine said:

yet? It just does not effect them. They are on the same node. However, they dont have the same power domains. they dont have the same eTVB, they dont have the same microcode. 

If we find 12th gen to have issues at some point, it will NOT be the same issue. 

Well assuming there isn't also an issue with the memory controller which may also be a problem since some CPU's regain stability with lower clocked memory. 12th uses essentially the same memory controller.

[saltycaramel]

Are 13/14th gen CPUs with 35W base TDP (e.g. i5-14400T) safe?

 

Quote

Intel Core 13th and 14th Generation desktop processors with 65W or higher base power – including K/KF/KS and 65W non-K variants – could be affected by the elevated voltages issue. However, this does not mean that all processors listed are (or will be) impacted by the elevated voltages issue.

 

- Intel via The Verge

 

[porina]

6 hours ago, starsmine said:

If we find 12th gen to have issues at some point, it will NOT be the same issue. 

This. There may be multiple things feeding into some 13/14 gen being unstable.

 

Currently we know of:

Microcode with excessive voltage for 13/14 gen

Oxidation for some early 13 gen

Over-ambitious mobo settings

 

Some lower models of 13th gen were rebadged Alder Lake, so might be exempt from the microcode thing.

 

5 hours ago, Bitter said:

Well assuming there isn't also an issue with the memory controller which may also be a problem since some CPU's regain stability with lower clocked memory. 12th uses essentially the same memory controller.

XMP/EXPO is never guaranteed by the CPU. You see all over this forum people who buy combos that work fine for most others but not for them. It is only technically a problem if it doesn't work at CPU published supported speeds at standard timings which vary with DPC and other factors.

 

If, big if, there was a problem spanning 12/13/14 gen with memory controller, at a high level you might expect 12th gen to show it more by now given it has been out longer. We're getting into unfounded speculation without more data.

 

 

[05032-Mendicant-Bias]

I have a 13700F bought in early 2023, so within the affected range.

 

My system is rock solid, by far the best computer I built. I don't remember when the last hard crash was. Viktoria 3 and Cyberpunk sometime do crash to desktop, but I doubt it's the CPU fault there.

 

Intel provides a tool to verify stability of the CPU, I ran it before and after

Spoiler

Before BIOS update

After Bios Update

If someone from LTT labs for some reason can make use of the logs, here they are:

Spoiler

--- IPDT64 - Revision: 4.1.9.41
--- IPDT64 - Start Time: 2024-07-27 10:11:04

CPU1
Genuine Intel CPU Test
Module Version: 1.0.26.64b.W
Start Time: Sat Jul 27 10:11:04 2024
Test Result - PASS
Expected: GenuineIntel
Detected: GenuineIntel
End Time: Sat Jul 27 10:11:04 2024
Total Time: in seconds: 0
--------------------------------------------------------------------
CPU1
CPU Brand String Test
Module Version: 1.0.28.64b.W
Start Time: Sat Jul 27 10:11:04 2024
Test Result - PASS
Expected: Intel(R) Core(TM)
Detected: 13th Gen Intel(R) Core(TM) i7-13700F
Intel(R) Core(TM) processor detected..!!..
End Time: Sat Jul 27 10:11:04 2024
Total Time: in seconds: 0
--------------------------------------------------------------------
CPU1
Cache Test
Module Version: 1.0.24.64b.W
Start Time: Sat Jul 27 10:11:04 2024
Test Result - PASS

 --- Reading Cache Size ---  

- Detected L1 Data Cache Size --> 48
- Detected L1 Inst Cache Size --> 32

- Detected L2 Cache Size --> 2048
- Detected L3 Cache Size --> 30720

Cache Size Test Passed!!!

End Time: Sat Jul 27 10:11:04 2024
Total Time: in seconds: 0
--------------------------------------------------------------------
CPU1
MMXSSE Test
Module Version: 1.0.30.64b.W
Start Time: Sat Jul 27 10:11:04 2024
Test Result - PASS

DetectUtils64 DLL Version - 1.1.9

 --- Determining MMX - SSE capabilities ---
..MMX is supported on this CPU..
..SSE is supported on this CPU..
..SSE2 is supported on this CPU..
..SSE3 is supported on this CPU..
..SSSE3 is supported on this CPU..
..SSE4.1 is supported on this CPU..
..SSE4.2 is supported on this CPU..

MMX Test Result --- PASS
SSE Test Result --- PASS
SSE2 Test Result --- PASS
SSE3 Test Result --- PASS
SSSE3 Test Result --- PASS
SSE4.1 Test Result --- PASS
SSE4.2 Test Result --- PASS

End Time: Sat Jul 27 10:11:04 2024
Total Time: in seconds: 0
--------------------------------------------------------------------
CPU1
IMC Test
Module Version: 1.0.26.64b.W
Start Time: Sat Jul 27 10:11:04 2024
Test Result - PASS
--- Reading Memory Size ---
Detected Memory Size is --> 32.00GB
--- Subtest - Memory Size Test Passed!!!  ---
--- Integrated Memory Controller Stress Test ---
Memory to be allocated = 1048576 bytes
Memory Allocated.
Test 1 Ones and Zeros Moving Inversions write operations - Passed
Test 1 Ones and Zeros Moving Inversions verification operations - Passed
Test 2 32Bits Sliding Ones write operations - Passed
Test 2 32Bits Sliding Ones verification operations - Passed
Test 3 32Bits Sliding Zero write operations - Passed
Test 3 32Bits Sliding Zero verification operations - Passed
Memory Deallocated.
--- Subtest - Memory Stress Test Passed!!!  ---
Integrated Memory Controller Test Passed!!!
End Time: Sat Jul 27 10:11:05 2024
Total Time: in seconds: 1
--------------------------------------------------------------------
CPU1
Parallel_PrimeNum Test
Module Version: 1.0.0.13
Start Time: 2024-07-27 10:11:05
Test Result - PASS
 
Module Math_PrimeNum.exe Completed - Pass
 
 
Prime Number Generation Test
Module Version: 1.0.29.64b.W
Start Time: Sat Jul 27 10:11:05 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
Test Result - PASS
Operation Per Second: 3052114
Error: 0
--- Prime Number Generation Test Passed!!!---
End Time: Sat Jul 27 10:11:50 2024
Total Time: in seconds: 45
 
Parallel_PrimeNum
End Time: 2024-07-27 10:11:50
--------------------------------------------------------------------
CPU1
Parallel_FP Test
Module Version: 1.0.0.13
Start Time: 2024-07-27 10:11:50
Test Result - PASS
 
Module AVX.exe Completed - Pass
Module Math_FP.exe Completed - Pass
 
 
AVX Test
Module Version: 2.0.31.64b.W
Start Time: Sat Jul 27 10:11:50 2024
Test Result - PASS

DetectUtils64 DLL Version - 1.1.9

  --- CPU Features Detection ---
..AVX is supported by this CPU..
..AVX2 is supported by this CPU..
..AVX512BW is NOT supported by this CPU..
..AVX512CD is NOT supported by this CPU..
..AVX512DQ is NOT supported by this CPU..
..AVX512ER is NOT supported by this CPU..
..AVX512F is NOT supported by this CPU..
..AVX512IFMA52 is NOT supported by this CPU..
..AVX512PF is NOT supported by this CPU..
..AVX512VBMI is NOT supported by this CPU..
..AVX512VL is NOT supported by this CPU..
..AES is supported by this CPU..
..PCLMULQDQ is supported by this CPU..

..AVX is supported by this Operating System..

Most Advanced AVX Feature Detected.. AVX2
AVX2 Test Result --- PASS
AES Test Result --- PASS
PCLMULQDQ Test Result --- PASS

End Time: Sat Jul 27 10:12:35 2024
Total Time: in seconds: 45
 
 
Floating Point Test
Module Version: 1.0.28.64b.W
Start Time: Sat Jul 27 10:11:50 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
FMA3 supported
Test Result - PASS
Million Floating Points per Second, MFLOPS: 6.045
Error: 0
--- Floating Point Test Passed!!!---
End Time: Sat Jul 27 10:12:35 2024
Total Time: in seconds: 45
 
Parallel_FP
End Time: 2024-07-27 10:12:35
--------------------------------------------------------------------
CPU1
Parallel_Math Test
Module Version: 1.0.0.13
Start Time: 2024-07-27 10:12:35
Test Result - PASS
 
Module Math_PrimeNum.exe Completed - Pass
Module FMA3.exe Completed - Pass
 
 
Prime Number Generation Test
Module Version: 1.0.29.64b.W
Start Time: Sat Jul 27 10:12:35 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
Test Result - PASS
Operation Per Second: 2661330
Error: 0
--- Prime Number Generation Test Passed!!!---
End Time: Sat Jul 27 10:13:20 2024
Total Time: in seconds: 45
 
 
FMA3 Test
Module Version: 1.0.29.64b.W
Start Time: Sat Jul 27 10:12:35 2024
Test Result - PASS

DetectUtils64 DLL Version - 1.1.9

  --- CPU Features Detection ---
..FMA3 is supported by this CPU..
..FMA3 is supported by this Operating System..

FMA3 Test Result --- PASS

End Time: Sat Jul 27 10:13:20 2024
Total Time: in seconds: 45
 
Parallel_Math
End Time: 2024-07-27 10:13:20
--------------------------------------------------------------------
CPU1
Parallel_GPUStressW Test
Module Version: 1.0.0.13
Start Time: 2024-07-27 10:13:20
Test Result - PASS
 
Module Math_PrimeNum.exe Completed - Pass
Module Math_FP.exe Completed - Pass
 
 
Prime Number Generation Test
Module Version: 1.0.29.64b.W
Start Time: Sat Jul 27 10:13:20 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
Test Result - PASS
Operation Per Second: 2661251
Error: 0
--- Prime Number Generation Test Passed!!!---
End Time: Sat Jul 27 10:13:50 2024
Total Time: in seconds: 30
 
 
Floating Point Test
Module Version: 1.0.28.64b.W
Start Time: Sat Jul 27 10:13:20 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
FMA3 supported
Test Result - PASS
Million Floating Points per Second, MFLOPS: 51.675
Error: 0
--- Floating Point Test Passed!!!---
End Time: Sat Jul 27 10:13:50 2024
Total Time: in seconds: 30
 
Parallel_GPUStressW
End Time: 2024-07-27 10:13:50
--------------------------------------------------------------------
CPU1
DGEMM Stress Test
Module Version: 1.0.17.64b.W
Start Time: Sat Jul 27 10:13:50 2024
Test Result - PASS

DetectUtils64 DLL Version - 1.1.9

  --- CPU Features Detection ---
..AVX is supported by this Operating System..

Most Advanced AVX Feature Detected.. AVX2
DGEMM AVX2 Test Result --- PASS

End Time: Sat Jul 27 10:14:50 2024
Total Time: in seconds: 60
--------------------------------------------------------------------
CPU1
Frequency Check
Module Version: 1.0.7.64b.W
Start Time: Sat Jul 27 10:14:50 2024
Test Result - NO COMPARE OPTION USED
Processor Name: 13th Gen Intel(R) Core(TM) i7-13700F
Measured Processor Frequency: 2.111632
End Time: Sat Jul 27 10:14:51 2024
Total Time: in seconds: 1
--------------------------------------------------------------------

System Information
------------------
Processor Name: 13th Gen Intel(R) Core(TM) i7-13700F
Processor Information: Family 6 Model B7 Stepping 1
Number of Physical Cores: 16
Number of Logical Cores: 24
Operating System: Microsoft Windows 11 Pro 64-bit
Graphics Information: NVIDIA GeForce RTX 3080
--------------------------------------------------------------------

--- IPDT64 - End Time: 2024-07-27 10:14:51
--- IPDT64 - Result: Pass
--------------------------------------------------------------------
--- IPDT64 - Revision: 4.1.9.41
--- IPDT64 - Start Time: 2024-07-27 10:53:56

CPU1
Genuine Intel CPU Test
Module Version: 1.0.26.64b.W
Start Time: Sat Jul 27 10:53:56 2024
Test Result - PASS
Expected: GenuineIntel
Detected: GenuineIntel
End Time: Sat Jul 27 10:53:56 2024
Total Time: in seconds: 0
--------------------------------------------------------------------
CPU1
CPU Brand String Test
Module Version: 1.0.28.64b.W
Start Time: Sat Jul 27 10:53:56 2024
Test Result - PASS
Expected: Intel(R) Core(TM)
Detected: 13th Gen Intel(R) Core(TM) i7-13700F
Intel(R) Core(TM) processor detected..!!..
End Time: Sat Jul 27 10:53:56 2024
Total Time: in seconds: 0
--------------------------------------------------------------------
CPU1
Cache Test
Module Version: 1.0.24.64b.W
Start Time: Sat Jul 27 10:53:56 2024
Test Result - PASS

 --- Reading Cache Size ---  

- Detected L1 Data Cache Size --> 48
- Detected L1 Inst Cache Size --> 32

- Detected L2 Cache Size --> 2048
- Detected L3 Cache Size --> 30720

Cache Size Test Passed!!!

End Time: Sat Jul 27 10:53:56 2024
Total Time: in seconds: 0
--------------------------------------------------------------------
CPU1
MMXSSE Test
Module Version: 1.0.30.64b.W
Start Time: Sat Jul 27 10:53:56 2024
Test Result - PASS

DetectUtils64 DLL Version - 1.1.9

 --- Determining MMX - SSE capabilities ---
..MMX is supported on this CPU..
..SSE is supported on this CPU..
..SSE2 is supported on this CPU..
..SSE3 is supported on this CPU..
..SSSE3 is supported on this CPU..
..SSE4.1 is supported on this CPU..
..SSE4.2 is supported on this CPU..

MMX Test Result --- PASS
SSE Test Result --- PASS
SSE2 Test Result --- PASS
SSE3 Test Result --- PASS
SSSE3 Test Result --- PASS
SSE4.1 Test Result --- PASS
SSE4.2 Test Result --- PASS

End Time: Sat Jul 27 10:53:56 2024
Total Time: in seconds: 0
--------------------------------------------------------------------
CPU1
IMC Test
Module Version: 1.0.26.64b.W
Start Time: Sat Jul 27 10:53:56 2024
Test Result - PASS
--- Reading Memory Size ---
Detected Memory Size is --> 32.00GB
--- Subtest - Memory Size Test Passed!!!  ---
--- Integrated Memory Controller Stress Test ---
Memory to be allocated = 1048576 bytes
Memory Allocated.
Test 1 Ones and Zeros Moving Inversions write operations - Passed
Test 1 Ones and Zeros Moving Inversions verification operations - Passed
Test 2 32Bits Sliding Ones write operations - Passed
Test 2 32Bits Sliding Ones verification operations - Passed
Test 3 32Bits Sliding Zero write operations - Passed
Test 3 32Bits Sliding Zero verification operations - Passed
Memory Deallocated.
--- Subtest - Memory Stress Test Passed!!!  ---
Integrated Memory Controller Test Passed!!!
End Time: Sat Jul 27 10:53:56 2024
Total Time: in seconds: 0
--------------------------------------------------------------------
CPU1
Parallel_PrimeNum Test
Module Version: 1.0.0.13
Start Time: 2024-07-27 10:53:56
Test Result - PASS
 
Module Math_PrimeNum.exe Completed - Pass
 
 
Prime Number Generation Test
Module Version: 1.0.29.64b.W
Start Time: Sat Jul 27 10:53:56 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
Test Result - PASS
Operation Per Second: 2113135
Error: 0
--- Prime Number Generation Test Passed!!!---
End Time: Sat Jul 27 10:54:41 2024
Total Time: in seconds: 45
 
Parallel_PrimeNum
End Time: 2024-07-27 10:54:41
--------------------------------------------------------------------
CPU1
Parallel_FP Test
Module Version: 1.0.0.13
Start Time: 2024-07-27 10:54:41
Test Result - PASS
 
Module AVX.exe Completed - Pass
Module Math_FP.exe Completed - Pass
 
 
AVX Test
Module Version: 2.0.31.64b.W
Start Time: Sat Jul 27 10:54:41 2024
Test Result - PASS

DetectUtils64 DLL Version - 1.1.9

  --- CPU Features Detection ---
..AVX is supported by this CPU..
..AVX2 is supported by this CPU..
..AVX512BW is NOT supported by this CPU..
..AVX512CD is NOT supported by this CPU..
..AVX512DQ is NOT supported by this CPU..
..AVX512ER is NOT supported by this CPU..
..AVX512F is NOT supported by this CPU..
..AVX512IFMA52 is NOT supported by this CPU..
..AVX512PF is NOT supported by this CPU..
..AVX512VBMI is NOT supported by this CPU..
..AVX512VL is NOT supported by this CPU..
..AES is supported by this CPU..
..PCLMULQDQ is supported by this CPU..

..AVX is supported by this Operating System..

Most Advanced AVX Feature Detected.. AVX2
AVX2 Test Result --- PASS
AES Test Result --- PASS
PCLMULQDQ Test Result --- PASS

End Time: Sat Jul 27 10:55:26 2024
Total Time: in seconds: 45
 
 
Floating Point Test
Module Version: 1.0.28.64b.W
Start Time: Sat Jul 27 10:54:41 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
FMA3 supported
Test Result - PASS
Million Floating Points per Second, MFLOPS: 5.33
Error: 0
--- Floating Point Test Passed!!!---
End Time: Sat Jul 27 10:55:26 2024
Total Time: in seconds: 45
 
Parallel_FP
End Time: 2024-07-27 10:55:26
--------------------------------------------------------------------
CPU1
Parallel_Math Test
Module Version: 1.0.0.13
Start Time: 2024-07-27 10:55:26
Test Result - PASS
 
Module Math_PrimeNum.exe Completed - Pass
Module FMA3.exe Completed - Pass
 
 
Prime Number Generation Test
Module Version: 1.0.29.64b.W
Start Time: Sat Jul 27 10:55:26 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
Test Result - PASS
Operation Per Second: 1878744
Error: 0
--- Prime Number Generation Test Passed!!!---
End Time: Sat Jul 27 10:56:11 2024
Total Time: in seconds: 45
 
 
FMA3 Test
Module Version: 1.0.29.64b.W
Start Time: Sat Jul 27 10:55:26 2024
Test Result - PASS

DetectUtils64 DLL Version - 1.1.9

  --- CPU Features Detection ---
..FMA3 is supported by this CPU..
..FMA3 is supported by this Operating System..

FMA3 Test Result --- PASS

End Time: Sat Jul 27 10:56:11 2024
Total Time: in seconds: 45
 
Parallel_Math
End Time: 2024-07-27 10:56:11
--------------------------------------------------------------------
CPU1
Parallel_GPUStressW Test
Module Version: 1.0.0.13
Start Time: 2024-07-27 10:56:11
Test Result - PASS
 
Module Math_PrimeNum.exe Completed - Pass
Module Math_FP.exe Completed - Pass
 
 
Prime Number Generation Test
Module Version: 1.0.29.64b.W
Start Time: Sat Jul 27 10:56:11 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
Test Result - PASS
Operation Per Second: 1878663
Error: 0
--- Prime Number Generation Test Passed!!!---
End Time: Sat Jul 27 10:56:41 2024
Total Time: in seconds: 30
 
 
Floating Point Test
Module Version: 1.0.28.64b.W
Start Time: Sat Jul 27 10:56:11 2024
DetectUtils64 DLL Version - 1.1.9
AVX is supported in your OS
Max AVX supported = AVX2
FMA3 supported
Test Result - PASS
Million Floating Points per Second, MFLOPS: 44.59
Error: 0
--- Floating Point Test Passed!!!---
End Time: Sat Jul 27 10:56:41 2024
Total Time: in seconds: 30
 
Parallel_GPUStressW
End Time: 2024-07-27 10:56:41
--------------------------------------------------------------------
CPU1
DGEMM Stress Test
Module Version: 1.0.17.64b.W
Start Time: Sat Jul 27 10:56:41 2024
Test Result - PASS

DetectUtils64 DLL Version - 1.1.9

  --- CPU Features Detection ---
..AVX is supported by this Operating System..

Most Advanced AVX Feature Detected.. AVX2
DGEMM AVX2 Test Result --- PASS

End Time: Sat Jul 27 10:57:41 2024
Total Time: in seconds: 60
--------------------------------------------------------------------
CPU1
Frequency Check
Module Version: 1.0.7.64b.W
Start Time: Sat Jul 27 10:57:41 2024
Test Result - NO COMPARE OPTION USED
Processor Name: 13th Gen Intel(R) Core(TM) i7-13700F
Measured Processor Frequency: 2.111750
End Time: Sat Jul 27 10:57:42 2024
Total Time: in seconds: 1
--------------------------------------------------------------------

System Information
------------------
Processor Name: 13th Gen Intel(R) Core(TM) i7-13700F
Processor Information: Family 6 Model B7 Stepping 1
Number of Physical Cores: 16
Number of Logical Cores: 24
Operating System: Microsoft Windows 11 Pro 64-bit
Graphics Information: NVIDIA GeForce RTX 3080
--------------------------------------------------------------------

--- IPDT64 - End Time: 2024-07-27 10:57:42
--- IPDT64 - Result: Pass
--------------------------------------------------------------------

 

 

Updated the BIOS of my MOBO Gigabyte B670 Gaming X F1 -> F11

Spoiler

I have a "Gigabyte B670 Gaming X" it turns out I had an old F1 driver, I never updated since building the computer. The latest is a F11 driver, I updated and turned back on XMP profile

BEFORE:

 

UPDATE:

 

AFTER:

 

 

Intel also has a processor ID tool

Spoiler

 

I run Cinebench 2024 before and after the BIOS update and the results are painful... multicore I went from 1561 to 1066, singlecore sayed the same at 113.

Spoiler

Before:

After

 

As far as I understand there will be a microcode update in august that I need to look for.

 

Something that is not clear to me: How do I check the CPU microcode version in windows? I found an instruction for linux, but not for windows, I think the microcode updated along with the bios, but I don't know how to check if that happened. Anyone has any suggestions?

 

 

 

 

[Vishera]

Seems to me like the voltages applied during TVB for workloads utilizing a small number of core degrade the ring - which is the interconnect that connects pretty much everything in the CPU.

So it's degradation from high voltage.

My rule of thumb is that 1.35V is always safe for daily use for modern CPUs and anything above that is not guaranteed to be safe.

[porina]

27 minutes ago, 05032-Mendicant-Bias said:

Intel provides a tool to verify stability of the CPU

While it provides some confidence I'm guessing it can't test all scenarios. I wonder if they'll release something that'll specifically cover the latest instability.

 

27 minutes ago, 05032-Mendicant-Bias said:

Something that is not clear to me: How do I check the CPU microcode version in windows? I found an instruction for linux, but not for windows, I think the microcode updated along with the bios, but I don't know how to check if that happened. Anyone has any suggestions?

https://www.intel.com/content/www/us/en/support/articles/000055672/processors.html

You have it already with the ID tool. Now, what versions are what is the follow up question.

 

Note the OS can also load updated microcode to the CPU. Microcode updates are not persistent, so it is preferable to be done in bios as it'll apply no matter what you boot to. Still, the OS path can help those who never update bios, or if there is no bios update available for long out of date mobos. MS are known to deploy microcode.

[Glazarus]

On 7/23/2024 at 8:40 PM, Tetras said:

Since your board is a workstation board, does your board not use very conservative settings for your CPU anyway?

Is this true? Meaning, it already keeps my CPU in a safe voltage range, even without the new microcode? If so, how do I check and confirm this? 
 

This would mean that depending on your motherboard type, you might be safe from these degradation problems. Would it be safe to consider that it’s then better to not upgrade bios for now and wait for the actual fix? Or is it still better to upgrade, even if my workstation motherboard would keep the CPU within safe voltages?

 

Some reports about server use of these processors seem to indicate that safer motherboards weren’t guarding against these failures?

[Tetras]

48 minutes ago, Glazarus said:

Is this true? Meaning, it already keeps my CPU in a safe voltage range, even without the new microcode? If so, how do I check and confirm this? 
 

This would mean that depending on your motherboard type, you might be safe from these degradation problems. Would it be safe to consider that it’s then better to not upgrade bios for now and wait for the actual fix? Or is it still better to upgrade, even if my workstation motherboard would keep the CPU within safe voltages?

 

Some reports about server use of these processors seem to indicate that safer motherboards weren’t guarding against these failures?

Since we don't know exactly what the voltage bug in the algorithm is and how/why these CPUs are being exposed to the higher voltage, I don't think it is possible to say that any motherboard is 100% safe, but Intel are saying that conservative settings are safer than what motherboard makers were using by default, so that's why we're guessing that it helps.

 

In the reddit thread where Intel posted about this issue, the rep strongly recommended that you update your BIOS even if it doesn't include specific fixes because they've been making other unspecified improvements.

 

In regards to the use of W motherboards to run servers, the problem with that is we don't really have any information to confirm what settings were being used since the leakers (like Wendell) never told us. There were also a lot of inaccurate reports in the tech media about exactly what W boards are capable of, because they're not really 'servers', you can overclock both the CPU and memory on them because the W680 chipset is just Z690 with ECC support.

[Zodiark1593]

Wonder what are the odds that faulty microcode voltages have also buggered the binning during validation? Meaning that chips that could only hit specific speeds reliably at spec voltages, have made it to higher bins than they were supposed to. 

[Bitter]

So it's a ring bus problem evidently?

It would be not unlikely that 12th and 15th Gen may be affected if they share ring bus which isn't an uncommon thing to happen between related architectures.

[porina]

3 hours ago, Bitter said:

So it's a ring bus problem evidently?

Intel have claimed a root cause. It might cause things to fail, but things failing are a symptom, not the cause.

[Bitter]

1 hour ago, porina said:

Intel have claimed a root cause. It might cause things to fail, but things failing are a symptom, not the cause.

Yes, I understand the fail is the result of a cause. That's literally part of my job . I've read several leaks stating that the reason the issues are so seemingly random and affect so many different things is that the ring bus is being degraded. I think Ring bus still shares voltage with one of the cache levels and I believe also frequency scales with one of the caches. I'm not super up to date on 13/14 gen topology so I don't know exactly which cache and if it runs at core clock or a divider. Degradation of the bus that carries data between different parts of the CPU is pretty bad though.

[Vishera]

5 hours ago, Bitter said:

So it's a ring bus problem evidently?

I think so, seems to me like it's degrading from the voltage applied to it being too high on 13th and 14th gen i5, i7 and i9 CPUs (The i5 are less affected).

6 hours ago, Bitter said:

It would be not unlikely that 12th and 15th Gen may be affected if they share ring bus which isn't an uncommon thing to happen between related architectures.

It doesn't affect 12th gen because the voltage in 12th gen is a lot lower.

[Bitter]

27 minutes ago, Vishera said:

I think so, seems to me like it's degrading from the voltage applied to it being too high on 13th and 14th gen i5, i7 and i9 CPUs (The i5 are less affected).

It doesn't affect 12th gen because the voltage in 12th gen is a lot lower.

Has anyone figured out the damage threshold for voltage?

[Vishera]

2 hours ago, Bitter said:

Has anyone figured out the damage threshold for voltage?

I have no clue, testing for these sort of issues can take months.

[leadeater]

3 hours ago, Bitter said:

Has anyone figured out the damage threshold for voltage?

 

1 hour ago, Vishera said:

I have no clue, testing for these sort of issues can take months.

Nobody other than TSMC, Intel, GloFo etc can actually reliably and accurately test that. The only way to tell is using specialized equipment far outside the realms of affordability for a consumer or reviewer, and even if you could afford it that doesn't mean you can operate it correctly and have enough information to assess it either.

 

These manufacturers already specify the voltages and a lot of that comes from testing the silicon fab process and transistor properties. Different areas and components of the die however have different voltage maximums and those also change with heat as well.

 

There is also a lot of tweaks over time and between generations so it doesn't matter a whole lot if two product generations use the same named process node or similar architectures because they won't actually be exactly the same. A design or process flaw that isn't known about can be present and used across generations though.

 

4 hours ago, Bitter said:

I'm not super up to date on 13/14 gen topology so I don't know exactly which cache and if it runs at core clock or a divider. Degradation of the bus that carries data between different parts of the CPU is pretty bad though.

Ring Bus is part of the LLC/L3 cache. It's all under what is called Uncore by Intel and that has it's own voltage domain which applies to the Ring Bus, LLC and Memory Controller (other stuff too not so  relevant to this discussion).

[starsmine]

7 hours ago, Bitter said:

Has anyone figured out the damage threshold for voltage?

  

7 hours ago, Vishera said:

I think so, seems to me like it's degrading from the voltage applied to it being too high on 13th and 14th gen i5, i7 and i9 CPUs (The i5 are less affected).

It doesn't affect 12th gen because the voltage in 12th gen is a lot lower.

 

All nodes are done under the idea of continuous improvement. 
hence why the node was fine for the 12th, and for a single stepping of 13th, it wasn't. 
And as silicon quality varies an areas, that threshold changes. 
 

 

4 hours ago, leadeater said:

 

Nobody other than TSMC, Intel, GloFo etc can actually reliably and accurately test that. The only way to tell is using specialized equipment far outside the realms of affordability for a consumer or reviewer, and even if you could afford it that doesn't mean you can operate it correctly and have enough information to assess it either.

Hey... there is also Micron, IBM, BAE, Samsung, 

but yes seeing the equipment its what the failure analysis labs are able to do. cutting silicon inside sub .18 micron accuracy to take pictures of individual transistors.