• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Your Geekbench 4 results

CHADBOGA

CHADBOGA

Platinum Member
To my surprise, a search didn't bring up a Geekbench 4 results thread, so with a new and exciting CPU launch almost upon us, I thought it could be a fun thread for people to post their Geekbench 4 results for everyone to see, and once the Ryzen 3000 series is out, the results will be really interesting.

As far as I know, Geekbench 4 doesn't favour one CPU maker over the other(if I am wrong about this, I'm sure you will let me know 😛).

I've only chosen Geekbench 4 as it is very easy to run, doesn't take long and gives both Single Core and Multi-core results in one go. You can download the free version here.

As this is not meant to be a super serious thread, I'm not recommending people go to extraordinary lengths to provide screenshots of system settings, but of course feel free to do so anyway.

In my next post where I provide my results, I'll demonstrate the modest lengths I'm going to.🙂
 
Windows 10 is updated to the latest version, so I presume all security patches have been applied and I have not bothered to seek out any Bios updates, if they even exist for my motherboard and/or CPU.

My Ivy Bridge i5 3570K is overclocked to 4.0Ghz

Single Core = 4141

Multi Core = 12529

tvqfGJI.jpg


OphdWiC.jpg
 
Last edited:
Not wanting to threadcrap but does geekbench test desktop CPUs properly. Last time I used it it hardly stressed my CPU at all. The single threaded test showed hardly any change from idle usage and the multithreaded just gave few spikes up to about 60% usage but not fully stressing the CPU.
 
WelshBloke said:
Not wanting to threadcrap but does geekbench test desktop CPUs properly. Last time I used it it hardly stressed my CPU at all. The single threaded test showed hardly any change from idle usage and the multithreaded just gave few spikes up to about 60% usage but not fully stressing the CPU.
Click to expand...

I stresses the CPU for sure, biut it does pause the test peridically in order to not throttle the CPU. So you probably checked the CPU load with an application that average the load during a certain time.
 
WelshBloke said:
Not wanting to threadcrap but does geekbench test desktop CPUs properly. Last time I used it it hardly stressed my CPU at all. The single threaded test showed hardly any change from idle usage and the multithreaded just gave few spikes up to about 60% usage but not fully stressing the CPU.
Click to expand...
I rarely saw 100% usage., I think the test is very flawed.
 
Markfw said:
I rarely saw 100% usage., I think the test is very flawed.
Click to expand...
The most likely reason for that is not that the test is flawed, but rather that the load is bursty enough that it's not detected as 100% due to the sampling period used when measuring the CPU load. The main problem with short tests like these is that they usually don't uncover thermal throttling, but that mainly applies to low-power platforms.
 
teejee said:
I stresses the CPU for sure, biut it does pause the test peridically in order to not throttle the CPU. So you probably checked the CPU load with an application that average the load during a certain time.
Click to expand...
I was just watching the graph in the windows task manager thing.
Surely to actually test the CPU you'd need to give it a task that wasnt momentary?
 
WelshBloke said:
I was just watching the graph in the windows task manager thing.
Surely to actually test the CPU you'd need to give it a task that wasnt momentary?
Click to expand...
Depends on what you're trying to measure. Even a comparatively short load will be sufficient to establish a meaningful result, as long as the CPU has spun up to its expected working frequency. You obviously need to be aware of the fact that the CPU will run mainly at max turbo frequencies in a test like this, so it might not be the best idea for establishing performance for long-running tasks.
 
Brunnis said:
Depends on what you're trying to measure. Even a comparatively short load will be sufficient to establish a meaningful result, as long as the CPU has spun up to its expected working frequency. You obviously need to be aware of the fact that the CPU will run mainly at max turbo frequencies in a test like this, so it might not be the best idea for establishing performance for long-running tasks.
Click to expand...
But am I likely to notice a difference in the tasks that Geekbench is measuring in the real world? If they are that short anyway I'm not going to notice if one is a fraction of a second faster than another as opposed to waiting for a video to encode.
 
WelshBloke said:
But am I likely to notice a difference in the tasks that Geekbench is measuring in the real world? If they are that short anyway I'm not going to notice if one is a fraction of a second faster than another as opposed to waiting for a video to encode.
Click to expand...
Yes, a lot of the workloads that are tested are of the type that scale. So, even if Geekbench just runs them for a short time, they may run a good deal longer when used in practice (e.g. compression, compilation, etc.). There's also the fact that Geekbench 4 gives a composite score that probably does at least a reasonable job of characterizing the general performance of a CPU. However, as always, if you're interested in a very particular use case, you need to test that (exact) use case.
 
Markfw said:
Well, I am not going to signup to publish, but here is my Result, 2970wx, win 10 latest patches@stock:
Single 4468 Multi 31911
Click to expand...

Markfw said:
I rarely saw 100% usage., I think the test is very flawed.
Click to expand...
Yeah, I agree that geekbench is not the best. It also seems to over-exaggerate SIMD performance.

Here is the result of a 16-core Zen 2 3950X engineering sample running at 4.3 Ghz(hat tip to @exquisitechar) compared to @Markfw 2970wx running at 4.2Ghz:

http://browser.geekbench.com/v4/cpu/compare/13495867?baseline=13496396

Single 5868 Multi 61072.

Now obviously the 16 core zen 2 is good, but I don't think it's that good. IMO not enough to justify 2x multi-threaded the score of a 24 Core Zen+ threadripper.
 
I'm not sure that 2970WX score is even correct.

Check the 2600X score against it and there is no way that it is reporting the multi-core score correctly.

edit: it certainly isn't an all-core 4.2GHz score, and it's base is just 2GHz.
 
multi results on geekbench are not reliable with node interleaving off vs. on (at least on my 48 core Opteron). 2x difference in performance results.
 
My 2970wx is stock. I wonder if some of the results over overclocked and stuck ? when I get my 2990wx up, and OC'ed maybe we can see.
 
w10 1903 (scheduler fixes)
R7 1700 @ 3.9GHz all cores
DDR4-3400 14-15-14-28 1T
C6H, BIOS 7002 (Zen2 AGESA 0.0.7.2a)
Performance Bias set to GB3/AIDA (L2/L3 latency similar to Zen+)

https://browser.geekbench.com/v4/cpu/13503467

eGXOBk3.png

ST: 4818
  • Crypto Score 6483
  • Integer Score 4258
  • Floating Point Score 4239
  • Memory Score 6529
MT: 29800
  • Crypto Score 22721
  • Integer Score 37497
  • Floating Point Score 34227
  • Memory Score 7610
-----------------------------------------------------------------------------------------------

Same setup on a live USB boot of Ubuntu 19.04 x64 (Kernel 5.0)

https://browser.geekbench.com/v4/cpu/13503705

Blza8ek.png

ST: 5031
  • Crypto Score 6532
  • Integer Score 4891
  • Floating Point Score 4660
  • Memory Score 5528
MT: 32795
  • Crypto Score 23900
  • Integer Score 42222
  • Floating Point Score 37079
  • Memory Score 7386
-----------------------------------------------------------------------------------------------

And a comparison link between both.

https://browser.geekbench.com/v4/cpu/compare/13503705?baseline=13503467
 
Last edited:
I think, that you should at least list the individual integer and floating point scores instead of the overall scores, because they are quite skewed due to memory and krypto/hash performance. This statement holds in particular for newer CPUs, which have significantly enhanced crypro scores due to additional fixed funtion HW units.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top