Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Geekbench 4 is out. CPU and GPU..

My MVC flashed GTX780 running on Nvidia Web Drivers will not will not run the API Test.

Also, what do the numbers really mean? My single core score is 3% higher and my multi core score is 25% lower than Geekbench 3? When I dyno my car, I expect consistency, and can roughly measure the difference with any modification I make. Not the case here?

Lou
 
flowrider said:
My MVC flashed GTX780 running on Nvidia Web Drivers will not will not run the API Test.

Also, what do the numbers really mean? My single core score is 3% higher and my multi core score is 25% lower than Geekbench 3? When I dyno my car, I expect consistency, and can roughly measure the difference with any modification I make. Not the case here?
Click to expand...
I haven't examined the details of the two benchmarks but it's expected the two results would not be the same. If they were there would be no reason to release the new benchmark.
 
ITguy2016 said:
I haven't examined the details of the two benchmarks but it's expected the two results would not be the same. If they were there would be no reason to release the new benchmark.
Click to expand...
Absolutely true - when a new version of a benchmark suite is released there's little value in comparing results from the old version to the new version.

For example, the Xeon CPUs have more L3 cache than many systems from 15 years ago had RAM. Running benchmarks that fit in cache doesn't give much insight into how the memory system performs. The benchmarks need to evolve and include programs that realistically use all of the processor and memory features.

Of course, that will mean that many recent systems will get poorer numbers on the new version. C'est la vie.
 
Last edited:
^^^^Numbers were the same when going from Geekbench V2 to V3. If they change there is no consistency in measuring one machine to another and measuring the same machine mod to mod.

And why the slight increase in single core score and the massive decrees in multiple core score? Makes no sense to me.

Lou
 
flowrider said:
^^^^Numbers were the same when going from Geekbench V2 to V3. If they change there is no consistency in measuring one machine to another and measuring the same machine mod to mod.

And why the slight increase in single core score and the massive decrees in multiple core score? Makes no sense to me.

Lou
Click to expand...
Just. Stop.

Benchmark suite numbers cannot be compared across different versions. Period.

The reason that there is a new version of the benchmark suite is because technical changes (like larger L3 caches) made the old version less representative of real world performance.

There is "no consistency in measuring one machine to another" across benchmark versions. Nor should there be.
 
flowrider said:
^^^^Numbers were the same when going from Geekbench V2 to V3. If they change there is no consistency in measuring one machine to another and measuring the same machine mod to mod.
Click to expand...
Unlikely as the the two versions, like version 4, use different baseline reference systems:

"Note that Geekbench 3 uses a different baseline system than Geekbench 2, so Geekbench 3 scores are not comparable to Geekbench 2 scores."​

https://www.primatelabs.com/blog/2013/08/geekbench-3/

flowrider said:
And why the slight increase in single core score and the massive decrees in multiple core score? Makes no sense to me.
Click to expand...
There are no increases / decreases in score between the two versions. The number is derived from an increase / decrease from a reference system and not the previous benchmark. They're not comparable.
 
  • Like
Reactions: keysofanxiety, ActionableMango and pastrychef
I think that benchmark is for measuring the maximum performance not real world usage of machines, because of not every body have exectly same type of usage. (e.g. somebody use photoshop which more suite with quad core high frequency cpu, when another use final cut pro which more suite with 12 core lower frequency cpu) But, in case of gb4, primatelabs already think for us what we should need
 
Long-shot request:

Is anyone running both the latest macSierra public beta and latest El Capitan on the same system?

Do you also have an nvidia card?

Can you pretty pretty please run Geekbench 4 and compare the CUDA scores with macSierra's 367 nvidia web drivers and el cap's 346 drivers to see if there's a difference?

Thank you!

EDIT - Totally forgot to add my scores!

5,1 12-core 3.46 GHz x 5690 & flashed Titan-X

Geek bench 4
64-bit Single - 2876
64-bit Multi - 20302
32-bit Single - 2582
32-bit Multi - 18458
OpenCL - does not work - crashes
CUDA - 144858


Geek bench 3
64-bit Single - 2792
64-bit Multi - 31918
32-bit Single - 2579
32-bit Multi - 28607
 
Last edited:
I wish i could test this out on OSX, but since im running a GTX 1080 im not able to see boot osx. I just built this machine to replace my mac pro and i couldn't be happier with the performance. Btw this is with my gimpy 32gb 1333 mac pro ecc ram. Once i get new ram the numbers should boost up some.

Im posting my results for CPU, Open CL, and CUDA. Once Osx gets working gtx drivers ill re-do it but who knows when that'll be lol.

CPU.jpg




cuda.jpg



open cl.jpg
 
Yes X5690 seems a little bit slower than W3690. I had 3092 points Single Core.
 

Attachments

  • upload_2016-9-3_11-36-59.png
    upload_2016-9-3_11-36-59.png
    2 MB · Views: 166
Their single core performance (MBA6,2) is killing my multi core (post #4)...:D

BTW i did notice that, while i do my CPU benchmark, CPU's are idling more than 50% of the time.
Did anyone else notice this, or is it just happening with the older CPUs and their instructions?
 
owbp said:
Their single core performance (MBA6,2) is killing my multi core (post #4)...:D

BTW i did notice that, while i do my CPU benchmark, CPU's are idling more than 50% of the time.
Did anyone else notice this, or is it just happening with the older CPUs and their instructions?
Click to expand...

You mean when GB4 accessing the single core performance? The CPU is mostly idle because only one core is working.
 
h9826790 said:
You mean when GB4 accessing the single core performance? The CPU is mostly idle because only one core is working.
Click to expand...
I thought that was the case, and most probably is, but Activity Monitor is showing at that time that Geekbench is using only about 40% (out of 800% and 100% being one full core).
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back