Discussion What is the best CPU benchmark?

[FlameTail]

Title says it all. Need I elaborate ?

In my opinion the best benchmark is the one that gives a good all round overview of a CPU's performance across a wide range of use scenarios. Of course, you are free to have your own definition of what the best CPU benchmark is, and if so, please post below!

So, what is the best CPU benchmark?

 

[Kocicak]

I do not get it, people have different use scenarios and therefore they pay attention to the benchmark/s that match/es their use scenario the best. These is no need for ONE all embracing benchmark.

 

[FlameTail]

One point i wanted to raise, in making this thread, is that for some reason, Apple M chips seem to underperform in Cinebench.

Just have a look at this.

Intel Core i9 12900HK vs Apple M2: performance comparison

We compared Intel Core i9 12900HK (2.5 GHz) against Apple M2 (3.5 GHz) in games and benchmarks. Find out which CPU has better performance.

nanoreview.net

In the Single Core test, the M2 outperforms the 12900HK in Geekbench, but in Cinebench the 12900HK is ahead by a very significant margin, so much so that the numbers are almost flipped compared to Geekbench.

What explains this disparity? I have heard that Cinebench is badly optimised for ARM architecture. Is this true?

 

[Storm-Chaser]

-Not my favorite, but geekbench is pretty good
-AIDA64 has a battery of CPU tests:

-Passmark performance test 10.2
-CPU-z measures single core and mutli core performance

 

[Mopetar]

Wouldn't surprise me if it were true. Up until Apple used their own SoCs in their computers, there wasn't a market for running Cinebench on ARM that I'm aware of.

Geekbench has been used in phone testing for ages. Hardly surprising an ARM-based CPU would do well there.

 

[nicalandia]

Geekbench is about the best as it's a combination of many tasks.

 

[nicalandia]

One point i wanted to raise, in making this thread, is that for some reason, Apple M chips seem to underperform in Cinebench.

Here is the 7700X vs The M2 Max

MacPro7,1 vs Mac14,6 - Geekbench

 

[FlameTail]

Also, Geekbench single/multi scores almost mirror their SPEC counterparts don't they?

For example if a CPU is 20% faster in SPEC, you can expect a similar ~20% score increase in Geekbench

 

[Hulk]

I like benchmarks that return repeatable results, free and portable (no installer), and are ubiquitous, meaning results are everywhere. The work done during the benchmark isn't so important in that if there are many results available I can relate the score to the overall strength of a processor.

For these reason I tend to look for CB R23 results. It's one of the only ones I know "by heart" if it's a good score. ie 2000+ ST is really good.

 

[Thala]

One point i wanted to raise, in making this thread, is that for some reason, Apple M chips seem to underperform in Cinebench.

It is not just Apple but all ARM64 SoCs - because the raytracing kernel is implemented using the Intel Embree library, which uses SSE/AVX SIMD implementation to speed-up computation. Issue is, that the ARM64 implementation is just a simple SSE-2-NEON wrapper. So technically the implementation is still calling SSE(2) intrinsics - which are implemented using NEON. Some intrinsics are taking more than 10 instructions on ARM - while taking a single instruction on x86-64.

In conclusion, you should stay away from Cinebench when it comes to comparing different CPU architectures.

 

[FlameTail]

It is not just Apple but all ARM64 SoCs - because the raytracing kernel is implemented using the Intel Embree library, which uses SSE/AVX SIMD implementation to speed-up computation. Issue is, that the ARM64 implementation is just a simple SSE-2-NEON wrapper. So technically the implementation is still calling SSE(2) intrinsics - which are implemented using NEON. Some intrinsics are taking more than 10 instructions on ARM - while taking a single instruction on x86-64.

In conclusion, you should stay away from Cinebench when it comes to comparing different CPU architectures.

Oh yeah, this is the answer i was looking for.

 

[lopri]

I like y-cruncher.

 

[Shmee]

3DMark has a decent CPU benchmark.

 

[DrMrLordX]

So, what is the best CPU benchmark?

Ideally, something open source so people can see what's under the hood. Also so as to make it easier to update it for newer microarchitectures/ISAs.

 

[FlameTail]

Behold. Let this thread rise from the dead!

 

[FlameTail]

Geekbench needs to have two Multi-core benchmark modes. One in the GB6 style and one in the GB5 style.

Then it'll be the perfect benchmark.

 

[FlameTail]

In conclusion, you should stay away from Cinebench when it comes to comparing different CPU architectures.

Since then, things have changed eh?

We got Cinebench 2024- now beefier and better optimized for ARM processors, and Geekbench 6- which changed the Multi-core scaling from Geekbench 5.

 

[SarahKerrigan]

SPECint_speed. It has its flaws, but it's extremely well-characterized, it's portable-ish, and there is no better single selection of code for characterizing "normal" single-threaded integer perf on real applications.

 

[DrMrLordX]

Behold. Let this thread rise from the dead!

Ugh, no necros please.

 

[Nothingness]

SPECint_speed. It has its flaws, but it's extremely well-characterized, it's portable-ish, and there is no better single selection of code for characterizing "normal" single-threaded integer perf on real applications.

Agreed. The main drawback is that its code size is on the small side compared to some applications. It also lacks run time code generation benchmark (due to portability being a prerequisite).

But yeah, not perfect but good (with the caveat that one should look at scores of subtests of course ).

Geekbench is not that bad contrary to what some believe. Its source is available to some customers, the benchmark selection is discussed by several companies (though I guess contrary to SPEC, the choice in end belongs to Primate Labs). But I still prefer SPEC.

 

[FlameTail]

 

[HutchinsonJC]

In my mind the only correct answer is: The one that demonstrates the CPU's performance for YOUR intended use.

In my view many of the all-in-one attempts are great as an all-around score, but that score is NOT going to provide the appropriate information that I'd use to make recommendations to someone assuming I understand exactly what they primarily intend to do with the machine.

If someone says they will primarily be doing transcodes and encodes and maybe weekend gaming where your understanding is that they make money from the transcodes and encodes, you don't recommend the same CPU as the next person who says that they primarily game, surf the net, and occasionally (a couple times a month) do some 1 to 2hr source duration transcodes.

And with that, it's easy to get into how people have two completely different rigs with two different intended goals.

 

[APU_Fusion]

This is already being debated in the Zen 5 thread 😁

 

[Markfw]

This is already being debated in the Zen 5 thread 😁

But this is the correct place for that discussion. There its OT. It has been reported.

 

[APU_Fusion]

But this is the correct place for that discussion. There its OT. It has been reported.

Oh i was making joke. Should have put /s there. My bad