[SOLVED] i9-11900F - - - low single-core Cinebench 2024 result, multi-core seems OK ?

[radosuaf]

I just bought this one used and benchmarking it I have the system as in the signature (ASUS APE2 enabled) and my Cinebech single core result is pretty poor...

CPU-Z looks somewhat similar (this is 11900F vs 11900K, but multi-core is 1.4% slower whereas single-core is 21% slower):

Is it possible that one or some of the cores don't work full-speed? Is there a way to test a particular core or do those run all the cores, but just one at a given time?

 

[radosuaf]

Clips came today finally... 851 multi-core . Single-core still at 99.

Asus are asking for details. Maybe they will do something?

 

[uWebb429]

Maybe they will do something?

Asus is better than most companies at releasing BIOS updates.

 

[radosuaf]

Let's do hope they fix this.

 

[radosuaf]

Got the answer from Asus today:
"Please be kindly informed that we used the latest official BIOS 2001 with 11900F and tested Cinebench 2024 with a single-core score of 100, and also tested CPU-Z with a single-core score of 698, which is much higher than the screenshot you provided.
During the Cinebench 2024 single-core test, HWinfo shows that the processor frequency can reach up to 5 GHz.

To summarize - we believe that the processor is faulty and suggest contacting Intel to resolve the issue."


Don't know what to think...

Funny thing is that reference score for 11900K in CPU-Z is 695 and on various sites it's even lower:
https://valid.x86.fr/bench/1

https://www.cpu-monkey.com/en/cpu_benchmark-cpu_z_benchmark_17_single_core

Would a 11900F score higher? Somewhat suspicious.

 

[uWebb429]

Somewhat suspicious.

An 11900K can run single core tasks like Cinebench faster than an 11900F. It is Intel rated to be 100 MHz faster. When installed in the same computer with the same operating system and same background tasks running, a 11900K should be faster compared to a 11900F.

https://ark.intel.com/content/www/u...1900k-processor-16m-cache-up-to-5-30-ghz.html

https://ark.intel.com/content/www/u...1900f-processor-16m-cache-up-to-5-20-ghz.html

I would not use CPU-Z as a single core test because it fully loads two threads which fully loads and activates two separate cores. Results can vary depending on what operating system is being used and how many background tasks are running. The Cinebench single core test is a true single core test. Cinebench results tend to be very consistent. Their 100 point score is right in line with your 99 or 100 point scores when your CPU is not throttling. Without sitting with the Asus tech, it is impossible to know what BIOS settings they used.

During the Cinebench 2024 single-core test, HWinfo shows that the processor frequency can reach up to 5 GHz.

You have already seen during your testing that HWiNFO does not accurately track the CPU MHz during single core testing, especially when a CPU is throttling. Your CPU at default specs was significantly throttling. It was not until you started using ThrottleStop that the reduced MHz became obvious.

the processor is faulty

Intel individually sets the voltage curve for each CPU. Your CPU combined with your motherboard is likely setting the CPU voltage high while the 11900F that Asus tested might be setting the voltage at an average or lower than average value. This varies depending on what VID voltage table was programmed by Intel into the CPU. High default voltage is causing your CPU to VMAX and EDP throttle during single core Cinebench testing.

I would not say the processor is faulty. It is more of very poor quality than outright faulty. It is more than three years old so I am not sure if Intel would be willing to replace it or not. Probably not. There is no guarantee that a replacement processor will be any better.

Intel does not guarantee that the maximum turbo frequency will be achieved or maintained during Cinebench or any other test. It is more of a theoretical maximum.

Max Turbo Frequency is the maximum single-core frequency at which the processor is capable of operating using Intel® Turbo Boost Technology and, if present, Intel® Turbo Boost Max Technology 3.0 and Intel® Thermal Velocity Boost. Frequency is typically measured in gigahertz (GHz), or billion cycles per second.

If it was my CPU, I would continue using it as is with the reduced loadline values and any other ThrottleStop tricks that might be necessary so it can run single core tasks at its full rated speed. After a few minor tweaks, your CPU is capable of running as it was designed to run.

I always buy K series CPUs even if I do not intend to overclock. Intel bins every CPU they produce. Their best CPUs that can run reliably at low voltage are likely to become K series processors. CPUs that need more voltage to run reliably become non K series CPUs.

 

[radosuaf]

Intel individually sets the voltage curve for each CPU. Your CPU combined with your motherboard is likely setting the CPU voltage high while the 11900F that Asus tested might be setting the voltage at an average or lower than average value. This varies depending on what VID voltage table was programmed by Intel into the CPU. High default voltage is causing your CPU to VMAX and EDP throttle during single core Cinebench testing.

I would not say the processor is faulty. It is more of very poor quality than outright faulty. It is more than three years old so I am not sure if Intel would be willing to replace it or not. Probably not. There is no guarantee that a replacement processor will be any better.

I always buy K series CPUs even if I do not intend to overclock. Intel bins every CPU they produce. Their best CPUs that can run reliably at low voltage are likely to become K series processors. CPUs that need more voltage to run reliably become non K series CPUs.

Ahhh... Now you made me look for a 11900KF . I know it does not make much sense financially and it's 3-5% performance increase at best, but it's fun to test new hardware and used ones are at decent prices now... Is silicon quality of K and KF the same? Ks are a bit too expensive and I don't care about iGPU.

I will also check if changing MCE from Auto to Enabled helps with EDP throttling.

 

[uWebb429]

silicon quality of K and KF the same?

Intel tries to sell as much of what they produce as they can. If there are any problems, Intel would rather disable a feature or two and sell a CPU as a lower end model than have to throw it in the garbage and not get anything for it. They have used this binning process since the beginning of time.

CPUs with a defective iGPU can be sold as KF processors. The iGPU might be bad but the CPU cores should be similar in quality between K and KF. Honestly, no one knows for sure. You might learn something if you can afford to buy a 100 or a 1000 of each and you did endless testing but that is not very practical.

Buying an 11900K or 11900KF might not solve your problem. When measured at the same MHz, the average peak VID voltage should be less compared to the average peak 11900F VID voltage. You have to keep in mind that the 11900K and 11900KF run 100 MHz faster. The VID voltage for many 10th and 11th Gen processors goes up significantly when going from 5.2 GHz to 5.3 GHz. You could still run into VMAX or EDP throttling issues during a single core Cinebench test.

it does not make much sense

I come to the same conclusion every time I think about replacing my 10850K with an 11900K. The only difference I would ever notice is that my wallet would be a little lighter. If you ever buy an 11900KF, let me know how testing goes. I know you want one.

 

[radosuaf]

The iGPU might be bad but the CPU cores should be similar in quality between K and KF.

Buying an 11900K or 11900KF might not solve your problem.

I know you want one.

I do hope so. Difference in price is not justifiable for someone who wants to use discrete graphics card.

I will do if I manage to find something at a decent price. That was the case with swapping i7-11700 for i9-11900F. I found one very cheap. I should be able to sell the i7 for more than I bouth the i9 for. It's a bit tempting to buy a new one and do the tests - if the CPU is visibly faster, then leave it, and if not - return. Will have to think about that .

I asked Asus to provide me the BIOS settings they used and they replied they'll send it, but pretty sure they will tell "Optimized Defaults".

EDIT: OK, so I decided to try to lower AC/DC Load Line... The lowest I could go and finish Cinebench is 0.19. Now checking if this one's stable.

 

[radosuaf]

Have I just bought a 11900KF?! Maybe... Will post some info when I get it and sit down to testing...

ASUS sent me info on their BIOS settings. Of course they wrote they used defaults.

 

[radosuaf]

Sooo... I actually got a 11900K . Long story, but it's sitting in my PC now. Unfortunately, results are slightly disappointing.
Let's start with the results:

Increases are like 2-3%, actually. I did the first test with default settings, CPU during the multi-core test in CB was an absolute heater, reaching temps north of 80C. EDP, PL1 and TVB limits reached. Result was 856 or so. I went down to 0.25 IA AC/DC Load Line, but got BSOD during the test. In the end 0.28 is the lowest stable with the above results. No limits reached during the single-core test, actually temps really low, multi-core TVB throttling all the time (no EDP throttling, though ), probably would need to change the cooler to get higher results, but see no point, really, as during normal workloads temps are absolutely fine.
11900K is definitely not worth the premium you need to pay over the 11900F, but in the end it's not THAT much more (when you buy used, that is) if you chase for absolutely highest performance.

 

[uWebb429]

multi-core TVB throttling all the time

Clear the ThrottleStop FIVR Thermal Velocity Boost box to get rid of TVB throttling.

If a Loadline setting of 0.25 gives you a BSOD, I would set this to 0.40 or 0.50. It is always best to use a setting that is not on the knife's edge of stability.

If ThrottleStop Limit Reasons shows PL1 lighting up red during Cinebench, I would increase the turbo power limits.

reaching temps north of 80C

80°C used to be a big number. Users are not quite so nervous these days. Intel's latest 13th and 14th Gen CPUs run at over 90°C for hours at a time. As long as a CPU is stable there is nothing to worry about. Intel CPUs do a great job of looking after themselves. They will automatically thermal throttle if they ever need to.

 

[radosuaf]

If a Loadline setting of 0.25 gives you a BSOD, I would set this to 0.40 or 0.50. It is always best to use a setting that is not on the knife's edge of stability.

I'm not doing any critical work, if it crashes, then it crashes . And any crash will result in increasing the value, so it will be OK at some point . More voltage means more power, more heat, more noise, so I'd have it lowest possible.

If ThrottleStop Limit Reasons shows PL1 lighting up red during Cinebench, I would increase the turbo power limits.

Power limits are at max already, but PL1 stopped showing up as soon as I decreased AC/DC Load Line - must hve been set at 200W, now power peaks at around 186W.

As long as a CPU is stable there is nothing to worry about.

The only reason to worry is the heat & the noise . My fans are set to spin at 100% from 70C, so the P12 Max in the rear gets quite noisy. I know up to 105C is fine with 11th gen.

 

[radosuaf]

We're getting there...

Even got 895, so 900 is within reach... This is roughly 12600K level of performance. Important thing - you need to set Adaptive Boost to Enabled instead of Auto. All cores are boosting to 50.25 GHz, so close enough to 5.1 GHz.

 

[radosuaf]

900 breached . This program is a godsend!

To summarize - changes from the defaults (only relating to CPU settings):
IA AC Load Line: [Auto] -> [0.3]
IA DC Load Line: [Auto] -> [0.3]
Intel Adaptive Boost Technology: [Auto] -> [Enabled]
PL1 Power Limit -> [Auto] -> [250]
PL1 Power Limit -> [Auto] -> [300]

The CPU's hitting maximum ~260W in Cinebench, in gaming it's half of it, 130W max.