Dedicated 8570m Scores lower than Onboard 6520G?

[bhvm]

Hello Friends.
I am in a bit of dilemma, and this question comes purely out of curiosity.

I was running Furmark 1.9 (Standard 720P bench) on some of my Machines to get a idea about their Performance. Guess what, 8570m Scored 303, Onboard 6520G scored 400!
To give you an idea, here are the specs of machines-

HP G002AX
AMD A8 6410 @ 2Ghz, Turbo 2.4 (Beema) (Onbiard R5 Gfx disabled)
Dedicated 8570m, 384 GCN shaders, 650mhz core, 900mhz memory
4GB Ram
1TB HDD
Furmark 1.9- 303.

Acer 5560
AMD A6-3420m @ 1.5 ghz. LLano
Onboard 6520g, 320 VLIW shaders, 400mhz core. No dedicated mem, 666mhz system ram,
8GB Ram
1TB HDD
Furmark 1.9- 400.

As you can see, the 8570 appears to have a clear advantage, mathematically. The CPU, GPU shaders, Clock speeds, everything!. Of course, both Machines were plugged in, Power profile set to HIGH PERFORMANCE and ATI power play to MAXIMIZE PERFORMANCE.

Please guide regarding the anomaly.

 

[caswow]

using furmark for performance determination. i actually dont know if you are really serious D:

 

[smithkt]

Furmark is generally used as a burn in stress test program. It will typically cause your GPU to heat up significantly which could be triggering some kind of thermal throttling on the 8570.

Furmark should not be used as a benchmark tool to compare performance. Try the Unigine benchmarks. Valley or Heaven would be better for performance comparison.

 

[LTC8K6]

See if the 8570m is throttling during Furmark.

GPU-Z might tell you.

 

[bhvm]

using furmark for performance determination. i actually dont know if you are really serious D:

I actually found Furmark most reliable. Its 100% GPU bound benchmark. I could underclock my CPU right down to 1Ghz and still gain almost same scores. Its very useful for comparing GPUs across Systems

See if the 8570m is throttling during Furmark.

GPU-Z might tell you.

It does not appear to be throttling. I've OC'd the GPU from stock 650/900 to 722/1000 and it appears to stay there.


One More anomaly is that Certain Versions of GPU-Z reports 256 shaders. Older versions report 384 shaders. Everest reports 320 shaders. Whats the truth?

 

[TurtleCrusher]

Furmark is not reliable for performance testing at all. Both major manufactures have nerfed furmark burn in to look better in reviews in regards to power consumption and temperature. Clocks will remain constant but the algorithms are simplified, reducing power draw and compute intensity.

Use Heaven benchmark instead.

 

[bhvm]

Ill try other benchmark and let you know. Any other benchmark which is tiny and portable?
Thanks.

 

[therealnickdanger]

Unigine Valley is decent and small-ish... I don't know about portable, however. I think it needs to be installed.

 

[Deders]

Agreed that Furmark isn't the best way of determining overall performance, It doesn't use half as many features as modern games will.

It can be great for quickly seeing how a GPU and cooling system behave in high heat situations, as well as worst case scenario power draw, but even then I prefer using one of MSI Kombuster's Tessellation based tests, especially since some drivers limit your GPU's clockspeeds when they detect Furmark.

 

[dark zero]

Hello Friends.
I am in a bit of dilemma, and this question comes purely out of curiosity.

I was running Furmark 1.9 (Standard 720P bench) on some of my Machines to get a idea about their Performance. Guess what, 8570m Scored 303, Onboard 6520G scored 400!
To give you an idea, here are the specs of machines-

HP G002AX
AMD A8 6410 @ 2Ghz, Turbo 2.4 (Beema) (Onbiard R5 Gfx disabled)
Dedicated 8570m, 384 GCN shaders, 650mhz core, 900mhz memory
4GB Ram
1TB HDD
Furmark 1.9- 303.

Acer 5560
AMD A6-3420m @ 1.5 ghz. LLano
Onboard 6520g, 320 VLIW shaders, 400mhz core. No dedicated mem, 666mhz system ram,
8GB Ram
1TB HDD
Furmark 1.9- 400.

As you can see, the 8570 appears to have a clear advantage, mathematically. The CPU, GPU shaders, Clock speeds, everything!. Of course, both Machines were plugged in, Power profile set to HIGH PERFORMANCE and ATI power play to MAXIMIZE PERFORMANCE.

Please guide regarding the anomaly.

The reason in bold. Llano was a very decent uArch, even far better than CAT in every aspect. Only Bristol and Stoney Ridge seems that they managed to catch them up.