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:

 

[511]

Lunar Lake getting destroyed by a phone chip: https://browser.geekbench.com/v6/cpu/compare/13274616?baseline=13140044

Strix Point is able to counter the multicore onslaught thanks to SMT but appears defenseless in ST: https://browser.geekbench.com/v6/cpu/compare/13274616?baseline=13267938

CRAP!

Use linux as phones uses linux kernel and as MS_AT posted phones can't sustain them without good cooling or browsing your phone in Antarctica

 

[igor_kavinski]

Igor I would have thought that after so many years on these forums you would know better and wouldn't make a sensation out of comparing geekbench performance between android and Windows 11

The Oryon V3 architecture in upcoming Snapdragon X2 laptops is supposed to be even better.

 

[poke01]

We can also take a look at the M3 which is clocked at 4.0GHz too and this 8 Elite 2 is run at 4Ghz for this bench. So we can do a clock comparison.

If we exclude non-SME 1t subtests it’s pretty even.

samsung SM-S947U vs Mac15,12 - Geekbench

So has Qualcomm managed to catch up to 2023 Apple ARM architecture?

Idk, now it doesn’t seem that impressive. They got Gerard Williams too.

Compared to the A18 Pro which has SME, Apple beats Qualcomm at 4GHz.

iPhone 16 Pro Max vs samsung SM-S947U - Geekbench

I wonder what the core power consumption at 4.7GHz will be, no mainstream phone will sustain that I think.

Also, this is likely tested on a device that is likely a step before mass production.

 

[Jan Olšan]

This is nothing new, really. Apples M4 Max, beats top Zen 5 desktop in ST Geekbench by ~20%.

Where?

official charts say 3385 or 3400 depending on which score you look versus 3912

 

[poke01]

Where?

official charts say 3385 or 3400 depending on which score you look versus 3912

Geekbench 6.3+ has SME which is hardware accelerated on M4+ SoC.

Geekbench 5 has the M4 Max tied with 9950X3D in 1t. (Yes I used linux)

 

[gdansk]

So has Qualcomm managed to catch up to 2023 Apple ARM architecture?

We presume these are not final clocks, no?
At this level of difference doesn't the question become who can scale frequency faster: Apple or Qualcomm?

 

[gdansk]

Where?

official charts say 3385 or 3400 depending on which score you look versus 3912

I happen to have both 9950X3D and M4 Pro.
Here's my M4 Pro v. 9950X3D. Both stock of course. An M4 Max might be a bit faster, but let's say around 12%.

Spoiler

and again but without SME

 

[Josh128]

Where?

official charts say 3385 or 3400 depending on which score you look versus 3912

I spitballed ~20%, not perfect, but is close enough to 15%-16%. Lets put it this way, its the difference between Zen 4 and Zen 5, basically a whole generation of difference.

 

[Jan Olšan]

I spitballed ~20%, not perfect, but is close enough to 15%-16%. Lets put it this way, its the difference between Zen 4 and Zen 5, basically a whole generation of difference.

I could spitball 10% then, would you also be content with that?

Not sure those official Geekbench results are well chosen anyway, since the scores usually have spread and kinda feels the m4 result is closer to the top of the spread and the 9950X to the lower end. That's the general problem with these claims. Somewhat inaccurate spitball there, slightly unfair scores, suddenly big deal.

I don't like it specifically in the context of Apple because their fan volunteer marketers love to overstate stuff. M1 having 5W TDP, "all week on battery charge", that sort of superlative poetry when numbers tell a more sober story :/

 

[511]

I spitballed ~20%, not perfect, but is close enough to 15%-16%. Lets put it this way, its the difference between Zen 4 and Zen 5, basically a whole generation of difference.

Geekbench include SME btw so it's not a fair comparison

 

[511]

as for historical significant i want a 4004 microprocessor lol

 

[Doug S]

So has Qualcomm managed to catch up to 2023 Apple ARM architecture?

Idk, now it doesn’t seem that impressive. They got Gerard Williams too.

How are you dismissing that as if it isn't an accomplishment? Look how badly Qualcomm fell on their face the previous time they tried to do their own cores, which were slower and more power hungry than ARM's lackluster A7x cores.

If this is a 4 GHz part that will be going to 4.7 GHz maybe they designed a bit less for IPC and a bit more for clock rate? Though I'm taking a wait and see on what GB reporting it was running around 4 GHz during that run actually means. Was that because it was some early part not running at full speed? Seems kinda late in the game for it to fall that short. Was it because GB's bursty nature doesn't get it to boost to its highest speed? If so I expect some OEMs to "adjust" its behavior to improve its benchmarks because of course they will. Or is GB's frequency reporting wrong? I mean, it doesn't even identify armv9 correctly and that is WAY easier than reporting frequency correctly on an unreleased SoC.

 

[poke01]

Not sure those official Geekbench results are well chosen anyway, since the scores usually have spread and kinda feels the m4 result is closer to the top of the spread and the 9950X to the lower end. That's the general problem with these claims. Somewhat inaccurate spitball there, slightly unfair scores, suddenly big deal.

You got Windows skewing the results. This is why looking at individual scores and subtests is important.

I don't like it specifically in the context of Apple because their fan volunteer marketers love to overstate stuff. M1 having 5W TDP, "all week on battery charge", that sort of superlative poetry when numbers tell a more sober story :/

Funny cause the same thing happened with Lunar Lake. Turns out Intel just optimised to to have good battery life during video playback and under load or any other application other than video playback it falls short due to Lion Cove.

I remember people here and elsewhere saying Lunar Lake won’t boost to 30 watts but 17 watts will be the maximum for the all core load. It happens in every fan echo chamber.

 

[poke01]

How are you dismissing that as if it isn't an accomplishment? Look how badly Qualcomm fell on their face the previous time they tried to do their own cores, which were slower and more power hungry than ARM's lackluster A7x cores

This time I’m dismissing it because they specifically brought a company to fix that. Mind you the company they brought wasn’t exactly in a bad shape so I expect more than matching last years Apple cores.

Again, it’s just one result and it is a strange one.

 

[Geddagod]

How are you dismissing that as if it isn't an accomplishment? Look how badly Qualcomm fell on their face the previous time they tried to do their own cores, which were slower and more power hungry than ARM's lackluster A7x cores.

If this is a 4 GHz part that will be going to 4.7 GHz maybe they designed a bit less for IPC and a bit more for clock rate? Though I'm taking a wait and see on what GB reporting it was running around 4 GHz during that run actually means. Was that because it was some early part not running at full speed? Seems kinda late in the game for it to fall that short. Was it because GB's bursty nature doesn't get it to boost to its highest speed? If so I expect some OEMs to "adjust" its behavior to improve its benchmarks because of course they will. Or is GB's frequency reporting wrong? I mean, it doesn't even identify armv9 correctly and that is WAY easier than reporting frequency correctly on an unreleased SoC.

Lowkey if I was Qcomm and my expensive cpu team acquisition can't differentiate themselves well from the stock ARM stuff, since that didn't really happen this gen with the x925, I would start getting mad lol.

If this is a 4 GHz part that will be going to 4.7 GHz maybe they designed a bit less for IPC and a bit more for clock rate?

I mean the IPC uplift still seems larger than the Fmax uplift. If its rumored to hit 4.7GHz, that's like a 5% improvement over last gen. The IPC improvement is like ~12%.

 

[511]

Lowkey if I was Qcomm and my expensive cpu team acquisition can't differentiate themselves well from the stock ARM stuff, since that didn't really happen this gen with the x925, I would start getting mad lol.

I mean the IPC uplift still seems larger than the Fmax uplift. If its rumored to hit 4.7GHz, that's like a 5% improvement over last gen. The IPC improvement is like ~12%.

Don't forget the SME accelerating one benchmark

 

[Geddagod]

Don't forget the SME accelerating one benchmark

With SME it's like ~17%, 12% was the figure i saw on twitter without sme

 

[511]

With SME it's like ~17%, 12% was the figure i saw on twitter without sme

Can you share the Link I doubt 12%

 

[Geddagod]

Can you share the Link I doubt 12%

https://twitter.com/x/status/1955231604764447087

 

[511]

https://twitter.com/x/status/1955231604764447087

The entire gain is almost from clock speed except for 3-4% from IPC

 

[Geddagod]

The entire gain is almost from clock speed except for 3-4% from IPC

The chip in question clocked lower than previous gen one, in this test, due to it being an es.
If what you said is true, perf gain would be <10% (if 8 elite gen 2 hits 4.7GHz).

 

[Doug S]

https://twitter.com/x/status/1955231604764447087

Wow those numbers are terrible. Almost all of the INT gain is from objection detection, which is the one I remember as being most affected by SME. Somehow the test I pay most attention to, clang, got worse! How the heck does that happen?!

 

[gdansk]

Wow those numbers are terrible. Almost all of the INT gain is from objection detection, which is the one I remember as being most affected by SME. Somehow the test I pay most attention to, clang, got worse! How the heck does that happen?!

Did you miss the clock rates are not equal? I ask because I had missed it earlier too.

 

[Doug S]

Lowkey if I was Qcomm and my expensive cpu team acquisition can't differentiate themselves well from the stock ARM stuff, since that didn't really happen this gen with the x925, I would start getting mad lol.

To be fair ARM got a LOT better between the time they announced the Nuvia acquisition and time the first fruit of that was released a year ago. There used to be a massive gap between Apple and ARM's designs. I think people don't notice how much it has shrunk because ARM was two years behind Apple a decade ago and they are two years behind now, but "two years" used to be a lot bigger when yearly process gains were bigger and there was more low hanging IPC fruit to grab.

That "crowding" between Apple, Qualcomm and ARM might be because they've all mostly exploited the same ideas.

 

[Doug S]

Did you miss the clock rates are not equal? I ask because I had missed it earlier too.

We still don't know what that "4050" clock rate means, and whether that's accurate since we've all seen GB6 produce wildly wrong frequency numbers on unreleased parts. If it can run at 4.7 GHz why wasn't it tested at that speed? Or is it that it can "run" at that speed, but only in one of those ridiculous phones that has a fan in it? I don't buy the "its an engineering sample" argument. Its not far from release and it is on N3P which is effectively a mature process from day one - TSMC never has yield issues with their iterations. So if it can only run at 4 GHz they can't blame the process on that, it would indicate some shortcomings in design.