Question Geekbench 6 released and calibrated against Core i7-12700

[igor_kavinski]

Geekbench 6 - Geekbench Blog

www.geekbench.com

Weird choice of baseline CPU and even weird is that the baseline score is 2500.

i7-12700 does hardly 2000 in GB5 with the fastest DDR5.

EDIT: Gideon's browser extension/add-on: https://forums.anandtech.com/thread...d-against-core-i7-12700.2610597/post-41452115

Helps you make sense of the results if you are left scratching your head about why a certain test is fast on a certain CPU.

Super useful!

Sample screenshot:

 

[Doug S]

I can prove this theory wrong with this. GB also added SVE support, something only CPUs from ARM support so far.

But don't you see, that's just misdirection and cover for the huge payoff from Apple. John Poole wrote that SVE support while he was waiting at the marina for the new yacht he bought with all that dirty Apple money to be delivered!

 

[okoroezenwa]

AMD is also favoured in Geekbench due to GB supporting a number of AVX512 extensions.

Oh definitely. I was just curious why Igor didn’t mention them but mentioned Arm when the reply of yours he quoted had both. It’s very curious.

 

[igor_kavinski]

Oh definitely. I was just curious why Igor didn’t mention them but mentioned Arm when the reply of yours he quoted had both. It’s very curious.

Coz AMD doesn't have cash lying around for these things. Besides, their Zen 5 performance slides show that they are happy with their CPU performance in GB5!

 

[poke01]

Coz AMD doesn't have cash lying around for these things. Besides, their Zen 5 performance slides show that they are happy with their CPU performance in GB5!

AMD is worth $100 billion more than ARM is…

AMD’s revenue was $22 billion in 2023 and ARM’s revenue was $2.6 billion.


It’s all cherry-picked benchmarks, they used a base M3, not even M3 Pro to compare multi-thread performance but they use an Intel Ultra 9 185H when using Cinebench 2024.

 

[igor_kavinski]

AMD’s revenue was $22 billion in 2023 and ARM’s revenue was $2.6 billion.

I said "cash lying around". You do know that AMD has quite the reputation as a penny pincher? If they wanted, they could've hurt Intel really, really bad with much more competitive products. But they prefer to co-exist instead of declaring all-out war and I believe AMD will regret this "weakness" of theirs.

 

[thunng8]

I said "cash lying around". You do know that AMD has quite the reputation as a penny pincher? If they wanted, they could've hurt Intel really, really bad with much more competitive products. But they prefer to co-exist instead of declaring all-out war and I believe AMD will regret this "weakness" of theirs.

I think this has turned into a comedy thread

 

[poke01]

I said "cash lying around". You do know that AMD has quite the reputation as a penny pincher? If they wanted, they could've hurt Intel really, really bad with much more competitive products. But they prefer to co-exist instead of declaring all-out war and I believe AMD will regret this "weakness" of theirs.

Your now reaching so far that’s it’s beyond just being biased to AMD.

AMD is already hurting Intel with their X3D SKUs and Eypc’s. Of course, Intel is being kept afloat by the US Gov and its OEMs like Dell and MS.

AMD is already more favourable to investors than Intel currently and the balance sheet shows that Intel is bleeding.

——

Let’s stick to the topic of this thread and leave out speculation. Unless you have receipts that Apple, AMD, Intel or ARM is paying Primate Labs, the discussion should about GB6 and its updates and benchmarks/improvements.

 

[okoroezenwa]

Coz AMD doesn't have cash lying around for these things.

L O L. AMD is too cash strapped to engage in a little benchmark tweaking for payment. I see.

Besides, their Zen 5 performance slides show that they are happy with their CPU performance in GB5!

So they’ve been paying since GB5, got it. 👍

 

[TwistedAndy]

Cinebench R23 and to a certain extent 2024 are very optimized for the Intel CPUs. Not a fair benchmark for ARM and even AMD.

Cinebench R23 and 2024 tend to show good results on all platforms, including AMD, Intel, Apple, and even Qualcomm.

AMD is also favoured in Geekbench due to GB supporting a number of AVX512 extensions.

According to the Chips & Cheese research, the AVX-512 usage in Cinebench R24 is close to zero:

From running Cinebench under Intel’s Software Development Emulator, Cinebench 2024 heavily leverages the AVX extension. Like libx264 video encoding, scalar integer instructions still play a major role. Both libx264 and Cinebench contrast with Y-Cruncher, which is dominated by AVX-512 instructions. AVX-512 is used in Cinebench 2024, but in such low amounts that it’s irrelevant.

In general, Chips & Cheese considers Cinebench R24 as a more balanced benchmark than SPEC 2017:

Maxon designed Cinebench to measure rendering performance, but we can look at how similar Cinebench is to other workloads too. It’s not representative of gaming performance, where the frontend is challenged by a ton of branches and a large instruction footprint. It’s also not completely bandwidth or compute bound, setting it apart from the likes of Y-Cruncher. Video encoding is a closer comparison, though video encoders tend to emphasize vector performance more than Cinebench does.

By itself, Cinebench 2024 is a moderate IPC benchmark with a sizeable instruction and data footprint. Code spills into L2, but the instruction stream is easier to follow than what we saw in games. Decoupled branch predictors can thus keep the frontend fed even in the face of L1i misses. On the data side, Cinebench 2024 spills out of L3 and requires a modest amount of DRAM bandwidth. High scheduler capacity across integer and FP operations help keep more memory operations in flight in the face of DRAM latency. In that sense, Cinebench 2024 can be seen as Cinebench R23 with more emphasis on DRAM performance. When hitting the execution units, Cinebench 2024 uses scalar and 128-bit packed floating point operations. Wider vector execution units are not useful. Scalar integer performance plays an important role in keeping the FP execution units fed.

In the end, Cinebench 2024 poses decent challenges to the frontend and backend. It has a more realistic instruction footprint than SPEC2017, which has no subtest with more than 12 L1i MPKI. Maxon has also addressed Cinebench R15 and R23’s small data-side footprint, which could be mostly contained by a 8 MB last level cache. High core count systems could be constrained by memory bandwidth, which happens across a lot of other well-threaded applications. These characteristics make Cinebench a decent benchmark. There’s area for improvement though. It could be a better stress test if it more heavily leverages vector execution. Hopefully the next version of Cinebench is better vectorized.

A link to the detailed review: Cinebench 2024: Reviewing the Benchmark

 

[Bencher]

Someone needs to probably correct me on this but after running the gb6 test on a stock 12900k, I scored a bit over 19k with average power draw of 19w according to hwinfo. Sure it spiked higher (150w max) but the average across the whole test was 19.

Seeing how the m4 is much newer and much larger chip, aren't the results it gets insanely disappointing? Even when I disable half my cores (both p and e) I still get a score of 13-14k. How many transistors are the cores of the m4 made up of?

 

[Nothingness]

Someone needs to probably correct me on this but after running the gb6 test on a stock 12900k, I scored a bit over 19k with average power draw of 19w according to hwinfo. Sure it spiked higher (150w max) but the average across the whole test was 19.

Seeing how the m4 is much newer and much larger chip, aren't the results it gets insanely disappointing? Even when I disable half my cores (both p and e) I still get a score of 13-14k. How many transistors are the cores of the m4 made up of?

The M4 has 4P + 6E cores. By disabling half of your cores you have 4 with Hyper threading + 8 so that's still more than the M4.

EDIT: Forgot to remind that GB6 MT scaling is highly very different depending on the subtests you look at.

 

[poke01]

Cinebench R23 and 2024 tend to show good results on all platforms, including AMD, Intel, Apple, and even Qualcomm.

Nope, Cinebench R23 is heavily biased towards Intel.

 

[poke01]

Cinebench R24

This is the best Cinebench

 

[Bencher]

The M4 has 4P + 6E cores. By disabling half of your cores you have 4 with Hyper threading + 8 so that's still more than the M4.

EDIT: Forgot to remind that GB6 MT scaling is highly very different depending on the subtests you look at.

The number of cores isnt really relevant though. If you wanna compare arm with x86 isn't the transistor count more appropriate than number of cores?

And no, my test was 4+4, I disabled half my p and half my e cores.

 

[TwistedAndy]

Nope, Cinebench R23 is heavily biased towards Intel.

Actually, AMD has constantly featured different versions of Cinebench in its presentations since Zen 2. Also, Qualcomm featured results of this benchmark during the X Elite announcement.

 

[Nothingness]

The number of cores isnt really relevant though. If you wanna compare arm with x86 isn't the transistor count more appropriate than number of cores?

You start your comparison with benchmark results. Before drawing any conclusion, you should first make sure the benchmark results are comparable. Then you'll be able to compare perf/area.

And no, my test was 4+4, I disabled half my p and half my e cores.

Oops I mistakenly read 14900k, my apologies. The hyper threading point is still valid though which might make the comparison unfair (though I have no clue how much GB6 can benefit from HT).

To simplify the task why don't you start with ST scores? Of course area comparison will be more difficult as you'll have to factor cache hierarchy differences.

 

[Nothingness]

Nope, Cinebench R23 is heavily biased towards Intel.

Yeah anyone comparing Arm vs x86 using Cinebench R23 has no clue about what he's doing. Or a bias.

And it takes two seconds and half a brain to realize that:

Cinebench R23 Scores [Updated Results]

Cinebench R23 is the newest instalment of the most popular CPU-Rendering Benchmark Cinebench. We have the Score Results for all modern Processors.

www.cgdirector.com

Cinebench 2024 Scores [CPU & GPU] (Updated Results)

The new Version of Cinebench, 2024, benchmarks CPUs and GPUs in the popular Redshift render engine. We benchmarked all modern CPUs and GPUs and list them in our score overview.

www.cgdirector.com

 

[igor_kavinski]

Let’s stick to the topic of this thread and leave out speculation. Unless you have receipts that Apple, AMD, Intel or ARM is paying Primate Labs, the discussion should about GB6 and its updates and benchmarks/improvements.

Agree. I said what I needed to say and probably said what some have been thinking about. It's out there for everyone to read and that was the purpose all along.

Thank you folks, for a civil discussion.

 

[poke01]

Actually, AMD has constantly featured different versions of Cinebench in its presentations since Zen 2. Also, Qualcomm featured results of this benchmark during the X Elite announcement.

I did say Cinebench R23 is bad. The 2024 version is good which both Qualcomm and AMD featured in their keynotes.

 

[igor_kavinski]

And no, my test was 4+4, I disabled half my p and half my e cores.

Can you please post a comparison of your score with M4? It may not look that good for 12900K but it would still be interesting!

 

[FlameTail]

So how does ARM SVE affect GB6 scores?

When GB6.3 and M4 released, everybody and their grandmothers were talking about SME, and how it inflates the score in GB6.

We have had ARM Cortex cores with SVE for years now. Why has no one done an investigation yet?

 

[igor_kavinski]

We have hard ARM Cortex cores with SVE for years now. Why has no one does an investigation yet?

Great question!

 

[Nothingness]

So how does ARM SVE affect GB6 scores?

When GB6.3 and M4 released, everybody and their grandmothers were talking about SME and how it inflates the score in GB6.

We have hard ARM Cortex cores with SVE for years now. Why has no one does an investigation yet?

Qualcomm made sure you can't use SVE on their Arm-based chips by disabling it in the firmware. With that in mind, good luck finding any result in GB DB which really makes use of SVE.

 

[FlameTail]

Qualcomm isn't the only ones making ARM CPU chips.

How about Tensor, Exynos or Dimensity?

 

[Bencher]

Can you please post a comparison of your score with M4? It may not look that good for 12900K but it would still be interesting!

Sure, this is with 4+4

https://browser.geekbench.com/v6/cpu/6156635

I have one with all cores running, scores 19k,ill try to find that too.

Eg1. Average power draw for the 19k score with the full 8+8 config was 19watts. Shouldnt the m4 absolutely destroy the 12900k since it's a much bigger chip?