Stock voltage is 1.425 and mines at 1.440 not much of an increase VRMs are cool to the touch.

agent_x007 wrote:

Would you like to test that theory ? […]
Show full quote

Imperious wrote:

My HD3850 ended up in a Asrock 4-coredual-sata with an overclocked E5200@3.4ghz. That definitely maxed out the video card.

Would you like to test that theory ?

3DMark 01 SE HD 3850 mini.png

😎

EDIT : And score after GPU OC...

3DMark 01 SE HD 3850 OC mini.png

You have actually proved my point, as an increase in the GPU core clock made only a minor difference. Your score increase over mine with the Pentium M is all CPU based.
A good example of this is that my Nature score is higher at 443fps.
3dmark06 would be a better test for this, at a certain point more cpu power will make only a minor difference.

Imperious wrote:

You have actually proved my point, as an increase in the GPU core clock made only a minor difference. Your score increase over mine with the Pentium M is all CPU based.
A good example of this is that my Nature score is higher at 443fps.
3dmark06 would be a better test for this, at a certain point more cpu power will make only a minor difference.

My point was to show you how much E5200@3,4GHz, may bottleneck 3850.
I agree, that in newer tests (or in high res.) there probably be no difference at all, between my E5800@4GHz, and E5200@3,4GHz.

Now, moving on to Dothan scores :
If I see correcly, you overclocked 3850 to 850MHz/972MHz => Good for you 😀 (nice OC !)
But here's a thing : If I were to compare it to my non-overclock score (and btw. I underclocked my GPU to real stock clocks of 3850, that is to 668/828) :

"Car" Low - 441,9 vs. 697,5 (+57,84%)
"Car" High - 129,5 vs. 218,8 (+68,95% [!])
"Dragon" Low - 665,7 vs. 962,9 (+44,64%)
"Dragon" High - 312,5 vs. 458,3 (+46,66%)
"Lobby" Low - 378,6 vs. 492,1 (+29,98%)
"Lobby" High - 160,1 vs. 217,3 (+35,73%)
In all these tests (excluding "Nature", which I discuss in a moment 😉), my stock clocked (668/828) card scored at least 30% higher, compared to yours with OC.

As to "Nature" test :
Stock (668/828) vs. OC (850,5/972) :
353,7 vs. 443,4 (+25,36%)
That's with +27,32% GPU, and +17,39% on VRAM.

From my point of view, everything is fine because my score is in line with my OC :
353,7 vs. 429,4 (+21,4%), for 789/1053 OC or +18,11% on GPU and +27,17% on VRAM.

In short : Dothan@2,7GHz, didn't bottlenecked your 3850 in Nature test (it's either that or I got a problem with my chipset/memory performance).
BUT in ALL other tests... yeah, not so much.
I would like to see those E5200 @ 3,4GHz with 3850 numbers now (or send me a PM ?).

I agree with what You are saying. The Nature test is very GPU dependent, the other games tests more CPU dependent.
A Bottleneck, whether it be cpu or gpu, involves many variables that can affect the results.

I wish I could properly publish my old results from the Asrock motherboard, but I turfed it in the bin a couple of years ago due to major cold
start problems, and fortunately I have since proved the HD3850 wasn't the culprit there. If You or anyone has the 3dmark2001se resultbrowser.exe then
I could get the full results back, but certainly I can't find that file on the web anywhere. What I can still do, whilst it's setup still, is do another run at Your underclocked speeds
and compare that. I did use to save all my results in the 3dp files, but without resultbrowser they are impossible to read.
This is as good as I can do at the moment, and Yes, your 4ghz AGP monster kills the Asrock I had. All but one of those Dothan scores is with a 7800gs

Thank you for scores.
I asked on Futuremark forum for resultbrowser2001.
They didn't have it in their archives 🙁

EDIT : This is your lucky day 😀
I found it !
I copied it to google drive.
Here's download link for (resultbrowser2001se.exe - version you need for 3DMark01 SE) : LINK

Thankyou very much for that, I'll make sure to copy that rare program to numerous places for safekeeping. I have attached the Asrock motherboard 3dp files for You to peruse.
Meanwhile, Today I received a very nice HD3870 PCI-e in the Mail, I gave up looking for a decent priced 512MB HD3850 as the AGP card values has affected the PCI-e models.
Anyway, I ran it at 668/900 (can't go lower for the memory in ATI Tray Tools) with my I5-750 @ 4017. The Nature test results are very close to the AGP 3850 at similar speeds.