A CPU summary finally? CPU Holomark scores! Now with GPUs!

A CPU summary finally??? CPU Holomark scores
Now with GPUs

Rhino 6, all done on 64bit version.

Okay so it has been asked for a while, so I thought hey why not do it.

There is a lot of people than want to know what difference CPUs have over each other.

This is a list of CPU scores for Holomark 2.60 (incomplete) and 2.61.

What I would ask is for people to keep posting their Holomark scores in the thread below.
Shout out to @Holo for this.

[Edit] I figured people would be more interested in the GPUs rather than the individual timings for each CPU tests, so I’ve swapped these.
Also it’s ordered by CPU score.

Red indicates re-run tests on the same machine.

Q - Identical CPUs have drastically different scores?
A - Different GPUs causing bottle necks, overclocked cores or something else, or first time run on a new machine; I can’t pinpoint each case.
Q - Identical systems improve scores on second test?
A - Is it the first time you’ve run this test on your machine? I’ve also noticed some differences by running my systems though CPU+GPU stress tests then launching Holomark and getting similar results to having run Holomark twice. Warming up the system seems to help performance, don’t just take a score from a system you’ve literally just turned on.
Q - My CPU has a high turbo/boost than another CPU, but scores lower?
A - Turbo clocking is not sustained, hence why overclocking is increasing the base clock of a CPU. Base clocks and cores are important here not turbo.

And here is V2.6 just for reference and so you can see that there isn’t much difference between AA levels.

I will try and update my tables as frequently as possible.

[Edit] Although people used to say Rhino was primarily single core utilization. Any CPU with good (not necessarily high) core count + high base clock speeds, which is fairly recent, will perform fine.
Just don’t skimp on your other components.


Does holomark recognize overclocking? Seems unlikely that none of these cpus are overclocked unless I’m misunderstanding what the “clock speed” vs “max clock” fields mean?

‘Clock speed’ is base, though some maybe higher if people have overclocked their base.
As far as ‘Max clock’, I can’t be 100% sure, but I would assume that the CPU test does not run long enough for the boost clock to kick in.

If you consider there are 8 CPU tests, they are relatively short for each and there would be a small, almost unnoticeable pause between them for us. So even if the cores ‘turbo’ up, there’s no reason to keep a sustained boost from it’s perspective between pauses.
(this is purely a guess on my part from what I know)

Probably a more important factor is that in the real world a CPU never consistently boosts for long periods of time, which is why you overclock your base to get it closer to the ‘turbo’.

Hi Nicholas,

I’m going to be building my 1st. rig. Not sure exactly how to read this but it looks like the i7-8700K is the way to go. Right now I can get either the i7-8700k or the R7-3700x for $200. Which way should I go? I am in the AEC field.


Hi @Pegasus82,

I should first say that you’re right, this table should not be read peanuts for peanuts, as we only have very few results from a lot of the same model CPUs and it would be nice to have a 3700x on the table.
Whether it’s true or not, the GPUs seem to be having an effect on CPU scores, as well as re-runs yielding different results.
Also a 3700x for $200 damn that’s good.

I’m not a fan boy of either red or blue, but AMD in 2019 definitely is a strong contender.

I’m going to guess that you work with large files both in storage and number of parts, with high tolerances.
If we’re just talking about CPUs I personally would have 3700x.

You might expect the 3700x to perform similar to the 9900k, but I can in no way say this.
What I will say is that I prefer the 3700x for it’s AM4 socket, which has more CPU upgrade options; as well as the higher RAM speed support.

That being said I don’t know anything about the rest of your workflow or software. I’m taking a stab in the dark and guessing you won’t run long important/final simulations on this machine, given the CPUs; in which case you won’t need ECC memory. Neither of these support ECC, I just wanted to check.

Don’t neglect the rest of you’re build, I don’t know exactly what you work with, so I’m going of the information I have. However RAM and GPU are still important, your PC is only as good as it’s weakest part.

Sorry for the long reply, I never like just saying X is better than Y without explaining and I wouldn’t take anyone’s advice if they didn’t justify it.

Hi @nicholasm785,

This is really great info. Thanks for the quick response. I use rhino, sketch up, adobe suite (photoshop, inDesign), AutoCad, ArcGIS (large data sets apply here), and would use any other design software necessary for a project in reality, but those are what I use most. While single core performance has been the most important, it seems the new version of Rhino and Photoshop are starting to utilize up to 8 cores (no more than that). Considering the Nvidia RTX 3080 is right around the corner and boasts significant improvements in rasterization I will opt for a 2060 super or possibly 2070 regular as Nvidia is going to reintroduce the 2070 regular into the market and adjust pricing. That should last me through the intro of the next gen cards and I’ll buy when prices settle.

The benchmarks on the Ryzen 3950x seem to be more positive than expected. Especially since the 3950x is “binned” and thermals are performing better than expected I will stick with the AM4 socket and go with the 3700x immediately. I’ll look to a 3950x upgrade in the future.

For the rest of the build I’m going with 32gb of Ram and making sure the air flow is good. Thanks again for the info.