Benchmark your Computer with the New Geekbench 3

[blackened23]

I want to get a Haswell now for higher benchmark scores. Grr. The voice of reason tells me to wait for Haswell-E next year since it will have all the goodies (SATA express, DDR4, etc).

 

[BallaTheFeared]

/wave hand

This isn't the Hasfail you're looking for.

 

[scannall]

Ran it on my laptop. AMD A8-4500m at 1.9 Ghz.

1282 and 3752

 

[guskline]

On my system 5% faster in 64bit mode.

Here are my results for rig 1 and 2 below both in 32bit and 64 bit

Rig 1 37700k @ 4.5Ghz
32bit
ST/MT = 3985/15974
64bit
ST/MT = 4169/16613

Rig 2 8350 @ 4.6Ghz
32bit
ST/MT = 2492/13410
64bit
ST/MT = 2757/14555

 

[sefsefsefsef]

Thats a much higher score than sefsefsefsef and apparently the same notebook (HP Envy 15). Is this because of different OS (Windows vs Linux)?

For my earlier runs I hadn't closed other running programs or put the laptop in an ideal cooling situation, so I reran the test just now giving the laptop the best shot at a high score that I could, and here's the result:

1846 / 4966 (which is an increase of 6.7% / 1.7% compared to my earlier runs)
http://browser.primatelabs.com/geekbench3/10689

The other Envy 15 seems to not be identical. The motherboards are different, they're running different OSes, and that other guy is running the 64-bit version, which someone earlier mentioned gives higher scores. The amount of RAM is also different, but that probably wouldn't make a difference in the score.

 

[crashtech]

The reality of Haswell is sinking in, therefore I was impelled to turn up the heat. My 32-bit results:

2700K @ 4.9GHz, 4048/16233

http://browser.primatelabs.com/geekbench3/10707

High ambient temps are keeping me away from 5.0 for now.

 

[mikk]

Here are my results for rig 1 and 2 below both in 32bit and 64 bit

Rig 1 37700k @ 4.5Ghz
32bit
ST/MT = 3985/15974
64bit
ST/MT = 4169/16613

Rig 2 8350 @ 4.6Ghz
32bit
ST/MT = 2492/13410
64bit
ST/MT = 2757/14555

Your AMD rig improves more in 64bit. I have a i5-4670 system so I can only speak for Intel. Here my results for i5-4670 @3,6 Ghz, DDR3-2133.

64bit: http://browser.primatelabs.com/geekbench3/9532
32bit: http://browser.primatelabs.com/geekbench3/9539

 

[Jaydip]

So, Balla is still unchallenged ?

 

[bgt]

And even better considering it s compiled with ICC
with numerous "intel only paths"....

Sick really.....thats why I only trust real usage of a CPU. Intel has too much money to scr.w benchmarks.

 

[Arachnotronic]

And even better considering it s compiled with ICC
with numerous "intel only paths"....

Yeah, when Intel wins a benchmark, it's because they're doing illegal compiler optimizations

BTW, didn't Mr. Poole say that GB3 was compiled with MSVC++?

 

[CHADBOGA]

BTW, didn't Mr. Poole say that GB3 was compiled with MSVC++?

Yeah but Mr.Poole only works for the company who made the benchmark, what would he know, compared to people prepared to kill & die for AMD?

 

[Abwx]

Yeah, when Intel wins a benchmark, it's because they're doing illegal compiler optimizations

BTW, didn't Mr. Poole say that GB3 was compiled with MSVC++?

It use GCC for linux , that s explicitely stated , for the Windows
version i dont know but ICC_patch show path that are CPU
manufacturer dependent.

That s not illegal since they did put a disclaimer for their compiler
but it makes no doubt , and this is subtancied by many site , that
optimum paths are reserved for their CPUs , as explicitly stated
by intel in said disclaimer.

Perhaps you should check Anand s tests of Blender using linux
and windows but that s another debate , open a thread if you
want more infos about it...

 

[Zucker2k]

^^^ In other words, you were just blowing hot air, as usual.

 

[blackened23]

It use GCC for linux , that s explicitely stated , for the Windows
version i dont know but ICC_patch show path that are CPU
manufacturer dependent.

That s not illegal since they did put a disclaimer for their compiler
but it makes no doubt , and this is subtancied by many site , that
optimum paths are reserved for their CPUs , as explicitly stated
by intel in said disclaimer.

Perhaps you should check Anand s tests of Blender using linux
and windows but that s another debate , open a thread if you
want more infos about it...

Wow. You have got to be kidding me. Just accept that the AMD is slower in single thread performance and stop looking for dubious reasons for benchmark results. IT is what it is. Slower single thread performance.

 

[Nothingness]

It use GCC for linux , that s explicitely stated , for the Windows
version i dont know but ICC_patch show path that are CPU
manufacturer dependent.

Mr Poole explicitly said VC was used. One guy applied the patch on an Intel CPU another one on an AMD CPU; both got almost the same benchmark score as without the patch. As far as I'm concerned, the case is closed.

 

[2is]

In AMDs defense, they would be faster if they weren't slower... I think that's what he's trying to say.

 

[Abwx]

It seems to me that geekbench is using integer ops and in benchmarks
it is verious obvious that an FX8350 has far more integer throughput
at stocks than a 2500K and it beat even the 2600K and 3770K when
they are at stock frequency , hence , any integer related bench
that dont reproduce this hierarchy can be said to no being equaly
optimized for different cpu vendors.

 

[BallaTheFeared]

I tried to track down Gikaseixas's score but couldn't find it in 3.0, there is a result with that score for a 4.8GHz 8350, but that's using the old version not 3.0.

 

[SlowSpyder]

My FX 9370 should arrive tomorrow (long story, but I couldn't get the 9590). I hope to have a new score soon!

 

[Durp]

So a 4Ghz Haswell is very roughly equal to a 7GHz Vishera in single threaded performance in this benchmark? D:


Seems legit.

 

[FwFred]

http://browser.primatelabs.com/geekbench3/9417

Intel Core Solo T1300 @ 1.66 GHz
ubuntu 12.04

first gen core vs. amd bobcat
http://browser.primatelabs.com/geekbench3/compare/9417?baseline=8797

clovertrail+ vs. amd bobcat c-50 apu
http://browser.primatelabs.com/geekbench3/compare/8957?baseline=8797
don't believe this...

also does anyone expect me to believe this??
http://browser.primatelabs.com/geekbench3/compare/8701?baseline=4252 [exynos 5 octa vs amd a8-4555m]
http://browser.primatelabs.com/geekbench3/compare/5818?baseline=9439 [i3-3217u vs iphone 5]

this is my all time favorites
http://browser.primatelabs.com/geekbench3/compare/728?baseline=2553 [i7-4770k vs exynos 5 dual]
http://browser.primatelabs.com/geekbench3/compare/2631?baseline=2553 [a10-4600m vs exynos 5 dual]

Some of the x86 vs ARM comparisons are obviously just badly configured test runs on the x86 side. The 4770k seems to be a mere fraction of what others are getting. Still, this shows problems with non-regulated and audited, closed-source benchmarks.

Perhaps we all shouldn't take this benchmark so seriously. Too bad nobody is submitting SPECINT 2006 for mobile CPUs :thumbsdown:

 

[Exophase]

It use GCC for linux , that s explicitely stated , for the Windows
version i dont know but ICC_patch show path that are CPU
manufacturer dependent.

That s not illegal since they did put a disclaimer for their compiler
but it makes no doubt , and this is subtancied by many site , that
optimum paths are reserved for their CPUs , as explicitly stated
by intel in said disclaimer.

Perhaps you should check Anand s tests of Blender using linux
and windows but that s another debate , open a thread if you
want more infos about it...

The problem here is that you guys think that the ICC patcher detects a cmp for the Intel vendor string it means it was compiled with ICC. In reality that code could have been independently provided. It's not a huge coincidence at all, since we know Geekbench displays the vendor string. It's not like we're talking about matching an entire function - all the patcher is reporting is finding cmp with pieces of the version string.

 

[AtenRa]

I tried to track down Gikaseixas's score but couldn't find it in 3.0, there is a result with that score for a 4.8GHz 8350, but that's using the old version not 3.0.

FX 8350 @ 4.7

Single - 2346

Multi - 13287

Edit: Not bad for a $190 cpu

2346 in single thread at 4.7GHz ?? you have to be kidding me

3647/13593 singe/multi 4770K @stock 16 gb 1600

2500K @stock, 2780 single 8820 multi. Only using 4GB single channel RAM and it's been a few months since I last fresh installed so that's probably causing a drop.

4770K single thread is 31% faster than 2500K ??

i7-3720QM @ 2.6GHz w/ 16GB RAM

Single Core: 3027
Multicore: 13426

Core i7 3720QM at 2.6GHz is faster than FX8350 at 4.7GHz ?? yea right.

 

[TuxDave]

http://browser.primatelabs.com/geekbench3/9417

Intel Core Solo T1300 @ 1.66 GHz
ubuntu 12.04

first gen core vs. amd bobcat
http://browser.primatelabs.com/geekbench3/compare/9417?baseline=8797

clovertrail+ vs. amd bobcat c-50 apu
http://browser.primatelabs.com/geekbench3/compare/8957?baseline=8797
don't believe this...

also does anyone expect me to believe this??
http://browser.primatelabs.com/geekbench3/compare/8701?baseline=4252 [exynos 5 octa vs amd a8-4555m]
http://browser.primatelabs.com/geekbench3/compare/5818?baseline=9439 [i3-3217u vs iphone 5]

this is my all time favorites
http://browser.primatelabs.com/geekbench3/compare/728?baseline=2553 [i7-4770k vs exynos 5 dual]
http://browser.primatelabs.com/geekbench3/compare/2631?baseline=2553 [a10-4600m vs exynos 5 dual]

Someone please come up with a better graphical representation for that website. If one chip DOMINATES the other chip in a single benchmark, the full width of the bar is set by that ratio (20:1 for example). Then in other benchmarks where the chip doubles the other, the bar is tiny and if you're not paying attention to the numbers you get the impression of "meh, all about the same".

 

[fixbsod]

i7-2700k @ 3.9GHz / 4x4 GB Corsair 1600 9-9-8-24:

3287 / 13073