7800GT Clock steppings

[IceMole]

Has anyone figured out the 7800GT clock steppings yet? I have the same XFX card as the latest review, they were only able to hit 477mhz. Mine however Auto detects at 515, I only have the card about 2 hours so far and I've been testing from 490 up. I'm at 510 right now, just wondering what a good point to shoot for is.

 

[otispunkmeyer]

im really not too sure

plenty of tech sites document this wierd stepping thing with the GTX, wether the steps are the same on the GT i dont know, but the best thing to do i think is to simply test it your self

run the source stress test, or some other fairly sensitive bench mark, and test a good range of frequnecies....you should be able to find the steps your self.

but to be fair...does it really need OC'ing?

 

[otispunkmeyer]

ok here you go

straight from Anand

"instead, jumps occurred between 461MHz/462MHz and 477MHz/478MHz. We will be looking into this more closely in the next 7800 GT article, and hopefully, we can shed some more light on NVIDIA's frequency selection issues."

 

[IceMole]

I was hoping someone went beyond that, I think a guy over at XS is going to do that pretty soon. It doesn't much help since they stopped at 477 and I'm now @ 520 and still going.

 

[Lonyo]

GL Excess might be useful, or 3D Mark01.
Both IIRC have pure pixel throughput tests which in theory should show you were the steps are. In theory being the important part.
I haven't tested it myself (I have a Leadtek 7800GT, haven't fiddled with clocks yet), but in theory it might give a decent indication.

 

[Cookie Monster]

I heard the New Rivatuner will allow unlocking of the masked pipes and VS.
So stay tuned.

 

[Lonyo]

3D Mark01 Fillrate tests:
(Core/memory)
(Single texture)
(Multi texture)

400/1000
3015
6774


432/1000
3030
7231


450/1000
3025
7670


460/1000
3025
7670


463/1000
3025
7670


470/1000
3025
7675


471/1000
3025
7670


472/1000
3030
7670


473/1000
3030
8088


475/1000
3025
8103


479/1000
3020
8098


I tested at random intervals really, up to 450, then I took the AT guideline numbers (461/462 or whatever) and went a bit to each side, but there wasn't a big jump at that level, nor any real jump from 450 to 470MHz on the core. Haven't wanted to go any higher yet.
Leadtek 7800GT (not the one with the funky heatsink).
EDIT: Mine seems to change between 470 and 479 somewhere.

After running some tests with Splinter Cell: Chaos Theory, we found that the frequencies didn't match up with what we saw with the 7800 GTX. We didn't see any jump in framerate between 472MHz/473MHz as we did with the GTX; instead, jumps occurred between 461MHz/462MHz and 477MHz/478MHz. We will be looking into this more closely in the next 7800 GT article, and hopefully, we can shed some more light on NVIDIA's frequency selection issues.

Interesting results for them, because my 472/473 matched up perfectly with their results from the 7800GTX, while their 7800GT results were different. Very odd.

 

[Lonyo]

(And the numbers are low because I had 4xAA and 8xAF on, but running the tests at 472/473 without them on gave the same kind of results, a jump in performance between the two speeds).

 

[otispunkmeyer]

Originally posted by: Lonyo
(And the numbers are low because I had 4xAA and 8xAF on, but running the tests at 472/473 without them on gave the same kind of results, a jump in performance between the two speeds).

nice work lonyo :thumbsup:

looks like OCing these new 7 series cards is going to be a unique experience for different cards.

i am impressed at the OP's card......520 and still going! looks like your on your own buddy. get testing!

 

[IceMole]

I did a quick test every 10mhz from 450 (stock) to 520 (max tested for artifacts so far)

Watch what happens above 480mhz, nothing much below that, it takes leaps above there.


3D Mark01 Fillrate tests:
(Core)
(Single texture)
(Multi texture)


450 (stock)
4776.5
8742.6

460
4781.5
8938.9

470
4791.6
9049.6

480
4801.6
9044.6

490
4886.7
9542.9

500
4811.7
9739.2

510
4816.7
9945.5

520
4816.7
10046.2

 

[Lonyo]

Mine won't do much above about 480MHz on the core, so I can't experiment further, but your results seem different to mine, between 470 and 480 certainly, which means it's probably that the cores overclock uniquely, and there may not be "set" points at which you see the results of speed increases.

How are you testing for artifacts? I'm using rthdribl at the moment, but the speeds I have seem to be my max stable anyway (473/1200) according to the "test" function, so I can't really make them any higher even if I wanted to.

 

[IceMole]

Well when I read XFX picks better chips for their overclocked cards I had no idea I'd be able to go above 500. Not bad for $368 @ zzf! I'll keep testing a little later, seems I hit the single texture wall @ 390, but the multi just keeps going. Greatest leap was at 480-490.

Edit: my rates are greater because the ram is set at 1200 not 1000 like yours. I have yet to push that any higher as well.

 

[otispunkmeyer]

Originally posted by: Lonyo
Mine won't do much above about 480MHz on the core, so I can't experiment further, but your results seem different to mine, between 470 and 480 certainly, which means it's probably that the cores overclock uniquely, and there may not be "set" points at which you see the results of speed increases.

How are you testing for artifacts? I'm using rthdribl at the moment, but the speeds I have seem to be my max stable anyway (473/1200) according to the "test" function, so I can't really make them any higher even if I wanted to.

i believe you can use Ati tool....it wont do any overclocking for you, but i think the built in artifact tester will work on NV cards

 

[IceMole]

I use ATITool v0.25 Beta 7 which allows for Nvidia overclocking. I used the auto overclock but as it passed 520 I freaked out and stopped it. It'll also do the memory which works not as well. I left mine and came back and it was above 1300 and my screen was frozen.