Aquamark3 GeForce6 scaling Analysis

C

cornelious0_0

[H]F Junkie
Joined
Apr 6, 2003
Messages
12,783
Well, seeing as this run of tests really didn't take that long, I thought I'd let you guys know what I found.

What I did was take my CPU and video card, clock them both down to stock, and run Aquamark3. Then I slowly started to raise the FSB, 5 at a time, recording the GFX, CPU, and Overall scores each time I did. Once I reached the 270fsb mark (where my 2.8C craps out on air cooling) I restarted the testing, and did the same thing with the 6800GT clocked at 425/1100.

My goal here was to try and find out when/where the CPU bottleneck imposed on these cards starts to wittle away. This can be done by simply taking a look at the GFX scores each time I raised the FSB.....as even though it was the CPU I was overclocking, the GFX score continued to rise, and I wanted to see when the gains started to diminish.

All tests were run with Image Settings set to High Quality, and all Optimizations disabled.....using the 66.72 Forceware drivers.

Scores are listed in the forum of GFX | CPU | Overall:

6800GT @ 350/1000

200 FSB - 2.80GHz: 8,181 | 9,027 | 56,310
205 FSB - 2.87GHz: 8,259 | 9,238 | 57,092
210 FSB - 2.94GHz: 8,379 | 9,542 | 58,229
215 FSB - 3.01GHz: 8,445 | 9,768 | 58,971
220 FSB - 3.08GHz: 8,517 | 9,928 | 59,604
225 FSB - 3.15GHz: 8,565 | 10,144 | 60,229
230 FSB - 3.22GHz: 8,625 | 10,317 | 60,834
235 FSB - 3.29GHz: 8,661 | 10,641 | 61,564
240 FSB - 3.36GHz: 8,730 | 10,898 | 62,336
245 FSB - 3.43GHz: 8,757 | 11,068 | 62,759
250 FSB - 3.50GHz: 8,799 | 11,266 | 63,294
255 FSB - 3.57GHz: 8,817 | 11,523 | 63,791
260 FSB - 3.64GHz: 8,844 | 11,691 | 64,172
265 FSB - 3.71GHz: 8,874 | 11,949 | 64,721
270 FSB - 3.78GHz: 8,883 | 12,144 | 65,062

6800GT @ 425/1100

200 FSB - 2.80GHz: 8,886 | 9,058 | 59,625
205 FSB - 2.87GHz: 9,033 | 9,289 | 60,789
210 FSB - 2.94GHz: 9,201 | 9,584 | 62,161
215 FSB - 3.01GHz: 9,321 | 9,751 | 63,079
220 FSB - 3.08GHz: 9,393 | 9,957 | 63,840
225 FSB - 3.15GHz: 9,492 | 10,164 | 64,721
230 FSB - 3.22GHz: 9,588 | 10,412 | 65,664
235 FSB - 3.29GHz: 9,714 | 10,645 | 66,716
240 FSB - 3.36GHz: 9,765 | 10,838 | 67,336
245 FSB - 3.43GHz: 9,873 | 11,088 | 68,316
250 FSB - 3.50GHz: 9,966 | 11,259 | 69,065
255 FSB - 3.57GHz: 10,062 | 11,523 | 70,054
260 FSB - 3.64GHz: 10,110 | 11,635 | 70,483
265 FSB - 3.71GHz: 10,185 | 11,925 | 71,374
270 FSB - 3.78GHz: 10,203 | 12,148 | 71,876

As you can see, when the GT is/was clocked at is default 350/1000, the advances in the GFX score almost came to a stand-still when I hit 245fsb...only increasing 27 points beyond the score at 240fsb. The trend continues through 'till 270fsb, and by then it has started to become even less of a gain, a measly 9 point jump going from 265fsb to 270fsb. With the card clocked at 425/1100 however, the results were much different. For starters, the gains to the GFX score were much more prominent each time I raised the FSB, which is to be expected with the higher clock rates on the card. It was interesting however to see that even the transition from 265fsb to 270fsb still netted me a gain of 18.....signalling that the CPU was just then becoming less of a factor in the GFX score.....a whole 25fsb ( 300MHz) later then when the card was running at its stock clock rates.

I may re-run these tests once I get my 3.2E running on my water cooling that's starting to arrive this week.....as I'd like to see how the numbers look once I hit and/or cross the 4GHz mark. I think I may also be tempted load up Doom3 when I get the 3.2E and run these exact same tests in Doom, to see how performance scales in a real life game, and in an OpenGL environment.
 
PopCorn said:
Great work. You're beginning to impress me young one. ;)
Click to expand...

Thx, now I just gotta try and come up with some way to do this data visually.....because the graph wouldn't be NEARLY as simple as the 3DMark05 one was.....and I'm not quite sure how to set it up. I want to though, so if anyone has any suggestions for a chart/graph format let me know and I'll get it done up. :cool:
 
Line graph..... (instead of bar graph)

Mhz across the bottom, score vertically....

Its actually very simple. It would show a linear increase at first then leveling out near the end of the line.
 
cornelious0_0 said:
Thx, now I just gotta try and come up with some way to do this data visually.....because the graph wouldn't be NEARLY as simple as the 3DMark05 one was.....and I'm not quite sure how to set it up. I want to though, so if anyone has any suggestions for a chart/graph format let me know and I'll get it done up. :cool:
Click to expand...

How 'bout a line graph... FSB/proc speed on one axis and score on the other. Place FSB/proc speed on the X axis and you'll be able to easily see the decrease in performance gain as you move towards the right (higher FSB) where the slope of the line will level off and not rise quite as rapidly. Kinda like disc transfer speed graphs I guess...
 
I guess I could just make the Y axis scale from 8k up to 75k and just use three different color'd lines for the GFX, CPU, and Overall scores......and I'd just make a seperate graph for the Ultra scores.

Makes sense....off I go to get started on it, cya in a bit. :cool:
 
bump for good info... even though it doesnt seem to be getting a lot of hits
 
RM_Bulldog said:
only got 3.7 using 1.518 vcore on my 3.2E with water cooling
Click to expand...

There are still a couple steppings out there for the 3.2E that are basically "bunk" overclockers.....do you know what yours was?
 
can i just say one thing to start with,
cornelious0_0 you are my GOD man ;) .
love the tests mate, really good info, you deserve a hit of my bong :p .
keep up the great work, also is it ok if i use the link to this thread on other sites?, just to spread the work.
PEACE.
 
Warhorse said:
can i just say one thing to start with,
cornelious0_0 you are my GOD man ;) .
love the tests mate, really good info, you deserve a hit of my bong :p .
keep up the great work, also is it ok if i use the link to this thread on other sites?, just to spread the work.
PEACE.
Click to expand...

For sure man....paste that shit up. :p

I finally got around a ridiculously stupid problem i was having with Excel, so I'm gonna try and get the graph up tonight. What it basically is, is a line graph showing the GFX scores from both clocks with the score on the Y axis and the cpu clock on the X.
 
Just wanted to say one more thing, I think I've decided on my next project.....

I have already tested out the scaling of my Aquamark3 scores at these two different video card clocks, so I figure what's next is for me to run 3DMark2001, 3DMark03, and 3DMark05 in the same manner.....to see how each benchmark "deals" with the two video card setups and scaling CPU.

What do you guys think? Interested in the lengthy results? :p
 
RM_Bulldog said:
mine was a SL7e5 pack date 9/08/04
Click to expand...

Well from what I've seen the E5 isnt nearly as nice for overclocking as the KC. Well, what I do know actually is that the KC is THE best stepping there is....so i guess you could look at it that way. :rolleyes:
 
Hmm, just found this, dated 11/oct/04, heres a piece of it,
"For this article I will be taking a GeForce 6800 Ultra graphics card and testing it at a number of processor frequencies. The system will keep exactly the same specifications for all the tests, the Pentium 4 processor clock will just be increased from 2.8GHz through to 3.6GHz. However, I will also include a severely under-clocked 2GHz version. This will be done to show just how important processing power is and how great an effect it can have on graphic card performance. Naturally, there will be very few systems out there utilizing a GeForce 6800 Ultra graphics card with just 2GHz of processing power. The 2GHz configuration has simply been added out of interest, to see just how big of an impact it will have on the graphics cards performance."
http://www.legionhardware.com/html/doc.php?id=349
 
Thx for the link, the results are quite interesting.

I'm very sorry for not getting started on my new project just yet, but it's because I'm still in the middle of another one. This week I will (hopefully) be getting my heatercore in the mail.....and once I get my case painted (again, hopefully in the next week or two) I can have my computer back up and running in the Lian-Li with the water cooling.

I really don't have the patience right now to sit down and drudge through another 5 dozen tests when I know the h2O is getting close to being done.

I hope you guys understand, but expect me to get right to work on the next comparison thread once the water cooling and case are done.....by the end of the month for sure.
 
You must log in or register to reply here.
Back
Top