Question Geekbench 6 released and calibrated against Core i7-12700

Page 34 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Jul 27, 2020
28,002
19,125
146
Last edited:

Gideon

Platinum Member
Nov 27, 2007
2,030
5,034
136
You are DA MAN!

I think a better way would be to create a browser extension or maybe a quick and dirty Electron app that takes a GB URL and presents the information in your proposed way.

Primate Labs don't want to do anything. Their forums are proof of that. Tons of requests from their users and they just sit on their asses, enjoying their pot of gold.

Thanks for the idea @igor_kavinski!
Remember the per-test ISA annotation mockups I posted?

Well, I finally got around to implementing it as an actual browser extension. It works on both individual result pages and comparison pages (which was an absolute PITA to support as they contain no ISA info themselves).

Current Features:
  • Shows the instructions a CPU supports in a grouped color coded way
  • Shows the instructions the test uses (info from here) and the given CPU supports under each test.
  • Optional tooltips with detailed info:
Kuvatõmmis 2025-05-24 132645.png


Comparison page
1748082739455.png
Single Result page:
screenshot1.png

Info about chosen colors and the rather arbitrary groups for ISA extensions in the Browser extension popup:
Kuvatõmmis 2025-05-24 133500.png

Download links
Currently, the per-test extensions shown are based on the official geekbench documentation, but It seems incomplete and old (no RISC-V info for instance, despite already supporting it)

If anyone has more up-to date info on what test use what extensions, please let me know!

The ISA grouping/coloring is also somewhat arbitrary.

Suggestions for alternative ISA grouping/coloring schemes as well as bug reports are very welcome!
 
Last edited:
Jul 27, 2020
28,002
19,125
146
  • Like
Reactions: Gideon

Gideon

Platinum Member
Nov 27, 2007
2,030
5,034
136
Question: How do I turn on the optional tooltips?
There is a browser extension popup (as with orher plugins, normally in the top right corner)

There, under the settings tab you enable it, press "save" and they should appear on hover.

Currently I only enabled them in the results section not in the "System Information" table (as they are sort of experimental) but that can be changed in the future
 
  • Like
Reactions: igor_kavinski

Gideon

Platinum Member
Nov 27, 2007
2,030
5,034
136
Photo Library and Object Detection use AMX on CPUs that support it.
Fixed in version 0.2.2

1748108297916.png

Also added tooltips to all top badges and enabled them by default (now you have to explicitly disable them in settings to not see them):

Kuvatõmmis 2025-05-24 203731.png

Firefox already has the latest version, but Chrome is still 2 versions behind at 0.2.0 due to really slow review process. Should get there in a day or two.


Oh, and the Tooltip also has a lengthy scrollable "Extensions" tab, to explain the categorization/coloring choices. (It's partially broken in 0.2.0 but fixed in later versions)

1748108583122.png
 

Gideon

Platinum Member
Nov 27, 2007
2,030
5,034
136
OOPS!

Spoke too soon. Photo library test isn't showing AMX for the Sapphire Rapids chip.

View attachment 124294
Version 0.2.2 does fix it. It's just that the update i pushed to Chrome Web Store is still pending due to google's excessively slow (automatic) review process. Even 0.2.1 isn't up yet!

Firefox already has version 0.2.2 (5 minutes after uploading) and there i can see the badge correctly:
Kuvatõmmis 2025-05-24 230609.png
 

Gideon

Platinum Member
Nov 27, 2007
2,030
5,034
136
Geeklens v0.3.0 released.

Download links (only for initial install, updates are automatic)

1. Finally squashed the AMX bug for good
2. Added inital support for RISC-V instructions. Only in Sytem Info as I have no solid info which benches use what
3. Fixed broken table layout for geekbench results with very long names, like this RISC-V CPU (go visit that result without the plugin, I dear you :D)

[v0.3.0] - 2025-05-26

Added:
  • Added Initial support for RISK-V
Fixed:
  • Properly fixed reporting AMX support for Photo Library
  • Fixed issues with disabling color or removing tooltips not working properly
  • Fix cell name overflow with really long names

Changed:
  • Removed save button from settings, now changes are applied immediately


I Actually fixed the stuff on the 26th, but it took until today for Chrome webstore to finally publish the new version
 

Sgraffite

Senior member
Jul 4, 2001
204
144
116
Is that on battery power? Seems off otherwise because: https://browser.geekbench.com/v6/cpu/compare/13022402?baseline=12968924


120U is better in every way so shouldn't be scoring less than Raptor Lake with fewer cores.
It is plugged in. Not sure why it is losing to the 1315U, as the 120U even has high clock speeds. I ran it in Linux if that matters, not sure if you would expect different results one way or the other between Win/Linux for Geekbench.
 

Josh128

Golden Member
Oct 14, 2022
1,319
1,985
106
Jul 27, 2020
28,002
19,125
146
Well, at least AMD can continue to be the torchbearer of decent x86 MT performance. ARM won't be able to do much in this regard unless they go crazy with P-core spam. Things will only get much better with Zen 6 but yeah, don't think AMD can do an ST clean sweep without hitting 7 GHz.
 
Jul 27, 2020
28,002
19,125
146
Not sure why nothing in x86 world can pull those levels of ST.
Intel's been late trying to come up with a way to mitigate the impact of branching instuctions. APX will be the solution that should reduce power use considerably and avoid branch mispredictions: https://www.intel.com/content/www/u...ical/advanced-performance-extensions-apx.html

Now if x86 had been a consortium of players willing to co-exist instead of arch rivals (rivalry mainly due to Intel's arrogance), APX would've been implemented a decade ago.
 

MS_AT

Senior member
Jul 15, 2024
868
1,762
96
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;)

Now about the chip itself, I get it, Snapdragon 8 Elite 2 can win ST benchmarks, but practically since Android doesn't have any sensible productivity features (official desktop mode is still 1 or 2 years away I am afraid) and even thinner shells makes these wins only usable on burst benchmarks, since the phone will not be able to sustain 10W for any sensible period of time. I mean, does it need 6 E cores and 2 P cores, wouldn't 1 P cores, 4 E cores do as well in most of day to day use. You anyway offload everything to dedicated accelerators. On the other hand, they wouldn't sell it cheaper, so maybe I shouldn't complain that marketing wars give us more cores;)

If they used that SKU in entry laptop market to push WoA or Linux, that would be much more interesting, but well, it is what it is.

APX will be the solution that should reduce power use considerably and avoid branch mispredictions
Conditional move instructions are already in use today, APX will just expand them, and they are not always a net performance win.