[SOLVED] Cause of low 8,197 Cinebench multi-core performance w/ new i5-12400 build ?

Status
Not open for further replies.
Apr 2, 2022
9
2
15
All of the following components were ordered new about 2 weeks ago.
  • CPU: Intel Core i5-12400F
  • GPU: Nvidia GTX 1070
  • SSD: XPG GAMMIX S70 BLADE 1TB
  • RAM: Corsair Vengeance LPX DDR4 3200 C16 2X16GB (Running on XMP 2.0)
  • MBD: Asrock H670 PG Riptide (BIOS flashed w/ latest version as of 3/30).
  • PSU: CORSAIR RMx Series RM650x 2018 CP-9020178-NA
Put this together about 2 weeks ago and obviously haven't overclocked it (F SKU), but ran a few benchmarks to make sure everything checked out alright.

From the single core tests, it appears to about match what I've seen other sites reports. But on multi-core tests it comes in significantly lower. During the test I've had HWiNFO and HWMonitor running at different times and the clock speeds seem to be steady at 3,990 Mhz and there don't appear to be any indicators of throttling. I've also run Cinebench as "Realtime" priority via Task Manager and ensured there's no other competing process for CPU w/ no different results.

Any ideas on what might be causing this score?

Userbenchmark: https://www.userbenchmark.com/UserRun/51674424

Cinebench R23 Multi-Core Score: 8,197

HWiNFO during Cinebench test (started and stopped during the 10m test): View: https://imgur.com/a/cyvEybj

HWMonitor: View: https://imgur.com/a/FB2sDaD

PassMark CPU Mark

View: https://imgur.com/a/s3pCsUb


Link to normal CPU Mark for 12400F: https://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i5-12400F&id=4681
 
Last edited:

Lutfij

Titan
Moderator
Welcome to the forums, newcomer!

Are your rams populating slots A2 and B2? Also, you forgot to mention the make and model of your PSU(and it's age). Mind stating where you sourced the installer for Windows 11?
 
Apr 2, 2022
9
2
15
Welcome to the forums, newcomer!

Are your rams populating slots A2 and B2? Also, you forgot to mention the make and model of your PSU(and it's age). Mind stating where you sourced the installer for Windows 11?


Hi and thank you.

Yes, the RAM sticks are occupying A2 and B2.
PSU is a 2 week old CORSAIR RMx Series RM650x 2018 CP-9020178-NA
Windows 11 was installed via a USB Install sources from the Media Creation Tool downloaded from Microsoft
 
@oNSXo
Review sites have reported 12400F Cinebench R23 scores in the low 12000 point range.

https://overclock3d.net/reviews/cpu_mainboard/intel_core_i5-12400f_ddr4_vs_ddr5_review/8

@reapzstar had the exact same problem you are having. Monitoring software like HWiNFO reported that the CPU was running at full speed for the entire benchmark but for some reason, his R23 scores were also way below what you would expect.


He was never able to solve this mystery. Perhaps Intel released a microcode update that has dramatically reduced Cinebench performance. Try finding some HWiNFO screenshots in an early review to see if the microcode version has changed compared to the version you are running now.

Here is an example from my 10850K. You need to find a similar screenshot for a 12400F.

TcrWpRR.png


The 3 most common microcode versions for the 12400F are 0x12, 0x15 and 0x18. I have seen some terrible CPU-Z multi-thread results while on 0x15. See what microcode version your CPU is using and try updating the BIOS if you are not running the most recent version.
 
Last edited:
Apr 2, 2022
9
2
15
@oNSXo
Review sites have reported 12400F Cinebench R23 scores in the low 12000 point range.

https://overclock3d.net/reviews/cpu_mainboard/intel_core_i5-12400f_ddr4_vs_ddr5_review/8

@reapzstar had the exact same problem you are having. Monitoring software like HWiNFO reported that the CPU was running at full speed for the entire benchmark but for some reason, his R23 scores were also way below what you would expect.


He was never able to solve this mystery. Perhaps Intel released a microcode update that has dramatically reduced Cinebench performance. Try finding some HWiNFO screenshots in an early review to see if the microcode version has changed compared to the version you are running now.

Here is an example from my 10850K. You need to find a similar screenshot for a 12400F.

TcrWpRR.png


The 3 most common microcode versions for the 12400F are 0x12, 0x15 and 0x18. I have seen some terrible CPU-Z multi-thread results while on 0x15. See what microcode version your CPU is using and try updating the BIOS if you are not running the most recent version.

So, one more interesting thing that makes me think some type of throttling again (but that would seem strange since the core speeds aren't changing).

I just booted into an Ubuntu LiveCD, let the machine set for 30 minutes and ran 5 consecutive CPU Mark runs.

View: https://imgur.com/a/iOQcDes
(SS of first and last)

A lot of the scores stay the same; however, the ones that drop by >10% are Sorting, Compression and SSE. All in all these beat my CPU Mark scores on Windows (above), which was 16,689.

Scores (First to Last):
  1. 19,074
  2. 18,131
  3. 17,625
  4. 17,601
  5. 17,473
Here's a shot from HWiNFO's CPU Specs. Like I mentioned, Asrock put out a new BIOS just a few days ago and in the CHANGELOG even mentioned an updated microcode, and it looks like mines running 0x1F. I have experimented w/ reflashing older BIOS back from Nov/Dec just to see if something was up w/ the latest, but I still experienced the same scores.

View: https://imgur.com/a/bvzPYSz
 
@oNSXo
Try running the CPU-Z benchmark. Multi thread results from one 12400F to the next seem to vary significantly during this test.

I also wonder if it might be chipset related. A 4000 point variation in R23 is not normal.
 
Apr 2, 2022
9
2
15
@oNSXo
Try running the CPU-Z benchmark. Multi thread results from one 12400F to the next seem to vary significantly during this test.

I also wonder if it might be chipset related. A 4000 point variation in R23 is not normal.

Here's the results of my CPU-Z: https://valid.x86.fr/pa4f2i (4,447 Multi-thread).

I found another CPU-Z result of someone else that has my same MB and they are getting results above 4,940 (https://valid.x86.fr/d2txfd).
 
Last edited:
Apr 2, 2022
9
2
15
@oNSXo
Try running the CPU-Z benchmark. Multi thread results from one 12400F to the next seem to vary significantly during this test.

I also wonder if it might be chipset related. A 4000 point variation in R23 is not normal.
One other weird thing on those CPU-Z results that differs from others I've seen is that it says:

Multiplier
Current: 40 / Min: 8 / Max: 25

Other ones I've seen all show
Multiplier
Current: 40 / Min: 4 / Max: 44

Now, the "Current" value matches at 40.... it just seems that mine saying Max: 25 is odd.
 
UBM shows the ram running at 2133.
The ram is OK. It is 2133 ram that is running at 3200 MHz.

6YMntK8.png


@oNSXo
I have seen CPU-Z incorrectly report the Min and Max data before. The current speed is correct. Your CPU is not throttling, at least not externally. All monitoring software shows it running at full speed.

I was curious so I adjusted my 10850K to 6 cores - 12 threads and slowed it down to 4000 MHz for a fair comparison. It scored 8231 in Cinebench R23. A 12th Gen should be scoring a lot better than that. Almost every R23 score I can find for the 12400F is 12000 points or better. I cannot think of any reasonable explanation why this is happening.

VZ7omIR.png
 
Apr 2, 2022
9
2
15
The ram is OK. It is 2133 ram that is running at 3200 MHz.

6YMntK8.png


@oNSXo
I have seen CPU-Z incorrectly report the Min and Max data before. The current speed is correct. Your CPU is not throttling, at least not externally. All monitoring software shows it running at full speed.

I was curious so I adjusted my 10850K to 6 cores - 12 threads and slowed it down to 4000 MHz for a fair comparison. It scored 8231 in Cinebench R23. A 12th Gen should be scoring a lot better than that. Almost every R23 score I can find for the 12400F is 12000 points or better. I cannot think of any reasonable explanation why this is happening.

VZ7omIR.png

I can't say I've ever heard of a defective CPU or MB that didn't present itself in not working (either freezing, crashing, completely inoperable, etc), but is there any chance that either one of them may be defective. I hate to just start switching things out especially if there's no issue, but it's puzzling as to what's going on here (and I've only got about ~2 weeks to exchange parts, if that's the route I should take).
 
it's puzzling
It sure is. I have never come across an issue like this before where there is a huge drop in the Cinebench score with no obvious signs of any throttling. Cinebench testing tends to be very consistent if your computer is running at a consistent speed.

I hate to just start switching things
I agree. Hard to justify swapping parts when only your Cinebench score is way below average. You and @reapzstar both have Asrock boards but his board has a different chipset, B660M. If I was going to swap parts, I might try swapping the motherboard first to one from a different manufacturer.

If you ever get this figured out, definitely come back here and post an update.
 
If your CPU is holding steady near 4 GHz during testing when constrained at 65W, one must wonder what clock speeds are hit when others run the 12400F on a Z690 board, enable MCE, disable boost duration limits/power limits...
 

reapzstar

Distinguished
Jan 31, 2016
148
6
18,695
Hey!

I had(sort of still have) the same problem.
What I did to make it a little better was I bought some proper thermal paste(Kryonaut from Thermal Grizzly), and made sure I mounted the screws tight enough.
My temps are a LOT better, maximum it gets to under Cinebench 23 is 54 and now I get around ~10.900, which is still like a 1000 off of the 12.000 people are getting on YouTube, but a lot better than the 8.900 and 9.500 I was getting before.
I don't think this is what made the difference, since on the video I saw of a guy testing the CPU, he used a H610 mobo and the stock cooler, and still got 12.000 points Link
I'm assuming there might be something wrong with the ASRock boards, since I noticed you rock it with an ASRock mobo as well.
Can't really think of anything else, and it is really frustrating after a while, since visually everything appears to be flawless, but the numbers just don't match...
 
  • Like
Reactions: uWebb429
All of the following components were ordered new about 2 weeks ago.
  • CPU: Intel Core i5-12400F
  • GPU: Nvidia GTX 1070
  • SSD: XPG GAMMIX S70 BLADE 1TB
  • RAM: Corsair Vengeance LPX DDR4 3200 C16 2X16GB (Running on XMP 2.0)
  • MBD: Asrock H670 PG Riptide (BIOS flashed w/ latest version as of 3/30).
  • PSU: CORSAIR RMx Series RM650x 2018 CP-9020178-NA
Put this together about 2 weeks ago and obviously haven't overclocked it (F SKU), but ran a few benchmarks to make sure everything checked out alright.

From the single core tests, it appears to about match what I've seen other sites reports. But on multi-core tests it comes in significantly lower. During the test I've had HWiNFO and HWMonitor running at different times and the clock speeds seem to be steady at 3,990 Mhz and there don't appear to be any indicators of throttling. I've also run Cinebench as "Realtime" priority via Task Manager and ensured there's no other competing process for CPU w/ no different results.

Any ideas on what might be causing this score?

Userbenchmark: https://www.userbenchmark.com/UserRun/51674424

Cinebench R23 Multi-Core Score: 8,197

HWiNFO during Cinebench test (started and stopped during the 10m test): View: https://imgur.com/a/cyvEybj

HWMonitor: View: https://imgur.com/a/FB2sDaD

PassMark CPU Mark

View: https://imgur.com/a/s3pCsUb


Link to normal CPU Mark for 12400F: https://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i5-12400F&id=4681
Quick test.
Turn off C states in the bios and rerun the benchmarks.
Any diff?

Yup I miss read the UBM ram speed.
 
Apr 2, 2022
9
2
15
Quick test.
Turn off C states in the bios and rerun the benchmarks.
Any diff?

Yup I miss read the UBM ram speed.

Hey, sorry, the forum stopped e-mailing me for some reason for replies.

Anyway, I turned off C-states and "CPU Thermal Throttling" and no difference. Just clocked in at 8,158 pts, no other BG processes or anything.
 
Apr 2, 2022
9
2
15
Hey!

I had(sort of still have) the same problem.
What I did to make it a little better was I bought some proper thermal paste(Kryonaut from Thermal Grizzly), and made sure I mounted the screws tight enough.
My temps are a LOT better, maximum it gets to under Cinebench 23 is 54 and now I get around ~10.900, which is still like a 1000 off of the 12.000 people are getting on YouTube, but a lot better than the 8.900 and 9.500 I was getting before.
I don't think this is what made the difference, since on the video I saw of a guy testing the CPU, he used a H610 mobo and the stock cooler, and still got 12.000 points Link
I'm assuming there might be something wrong with the ASRock boards, since I noticed you rock it with an ASRock mobo as well.
Can't really think of anything else, and it is really frustrating after a while, since visually everything appears to be flawless, but the numbers just don't match...

But, I don't even think my temps are bad? They top out at 70C. While it might be nice to run a bit cooler, to my understanding there's not any throttling going on til we hit 100C.
 
Apr 2, 2022
9
2
15
Hey!

I had(sort of still have) the same problem.
What I did to make it a little better was I bought some proper thermal paste(Kryonaut from Thermal Grizzly), and made sure I mounted the screws tight enough.
My temps are a LOT better, maximum it gets to under Cinebench 23 is 54 and now I get around ~10.900, which is still like a 1000 off of the 12.000 people are getting on YouTube, but a lot better than the 8.900 and 9.500 I was getting before.
I don't think this is what made the difference, since on the video I saw of a guy testing the CPU, he used a H610 mobo and the stock cooler, and still got 12.000 points Link
I'm assuming there might be something wrong with the ASRock boards, since I noticed you rock it with an ASRock mobo as well.
Can't really think of anything else, and it is really frustrating after a while, since visually everything appears to be flawless, but the numbers just don't match...

Hey again!

I have good news for me and hopefully, you too.

I started flipping more BIOS switches this morning after I requested an RMA already (of course), but anyway, now my Cinebench multi-core score is 12,154!

It's the "CPU Vcore Compensation" option under OC Tweaker. I saw this previously, but I kept it on Auto because I assumed it didn't apply to an F SKU.... wrong! I even adjusted Base Frequency Boost to 180W before while keeping Vcore at Auto....still no effect. But kicking Vcore compensation to "Level 5" (whatever that means... I tried "Level 3" and that did nothing), brings my Cinebench score up to the 12,000s.

Sure enough, core clocks still stay at 3,990 Mhz, but it's definitely using ~86W of power now? Weird.... but I'll take it.

FYI @uWebb429


View: https://imgur.com/a/8siS0sb


View: https://imgur.com/a/mEp1Iy7
 
Status
Not open for further replies.