PDA

View Full Version : Overclocking the 7800GTX


zakelwe
07-07-2005, 12:09 AM
If you want to benchmark your card at full speed make note of unwinders comments here

http://www.beyond3d.com/forum/viewtopic.php?t=24615&postdays=0&postorder=asc&start=60

Some updated info about investigation progress:

--------------------------------------
1) G70 has 3 independently clocked domains with no doubts. I've already fully ripped clock detection logic from the driver, so new RT will replace single "Core clock" graph with 3 graphs, representing current clocks of each domain. I still have no strict info about the functional purpose of these domains, most likely they are geometric domain (that clock is currently read by publicly available RT and 3DMark), shader domain and ROP domain. So this naming scheme ("Core clock \ geometric domain", "Core clock \ shader domain" and "Core clock \ ROP domain") is currently used in beta of RT 15.7. However we're still waiting for NV's official comments on it, as well as performing own synthetic testing allowing us to determine domain functional purposes ourselves. So domain naming scheme will probably change in future.
2) Shader/ROP domain are clocked with more primitive PLL comparing to geometric domain allowing per-1MHz clock frequency adjustment, and currently NVIDIA driver is able to adjust clocks of these domains with 27MHz (oscillator frequency) step only. For example, default 430MHz ROP clock is generated as 27MHz * 16, and the next ROP clock the driver is able to set is 459MHz (27MHz * 17). This results in rather interesting effect when overclocking G70: often overclocking will adjust geometric domain clock only, e.g. attempt to set 440MHz instead of default 430MHz will result in generating the same 432MHz for ROP domian clock, but 480MHz (440+40) for geometric domain clock. This effect is clearly visible on new RT's core clock monitor, G70 owners may also verify it with fillrate tests and see that there are no changes in fillrate for such example.
3) Target 3D clock for shader/ROP domain (430MHz) and delta for geometric domain clock (40MHz) are explicitly specified in VGA BIOS in performance table. New RT is able to display this delta in "NVIDIA VGA BIOS information" section of the diagnostic report. The example of such info is shown below:

$ffffffffff ---------------------------------------------------
$ffffffffff NVIDIA VGA BIOS information
$ffffffffff ---------------------------------------------------
$1100000000 Title : GeForce 7800 GTX VGA BIOS
$1100000002 Version : 5.70.02.11.01
$1100000100 BIT version : 1.00
$1100000200 Core clock : 275MHz
$1100000201 Memory clock : 1200MHz
$1100010000 Perf. level 0 : 275MHz/600MHz/1.20V/100%
$1100010001 Perf. level 1 : 415(+35)MHz/600MHz/1.40V/100%
$1100010002 Perf. level 2 : 430(+40)MHz/600MHz/1.40V/100%
$1100020000 VID bitmask : 00000011b
$1100020100 Voltage level 0 : 1.20V, VID 00000000b
$1100020101 Voltage level 1 : 1.30V, VID 00000010b
$1100020102 Voltage level 2 : 1.40V, VID 00000001b

Please pay attention to performance level 1 descriptor. It is so called low power 3D perormance level, and the system uses this performance level as a temporary step when switching from 3D (performance level 2) to 2D (performance level 0). As you see, it has different "Geometric clock delta", and it perfectly matches with temporary 450MHz clock (415+35), which you can see on many clock graphs.
During overclocking the driver uses this BIOS-defined delta and generates closest possible to desired clock for ROP/shader domain and closest possible to (desired clock + delta) for geometric domain. => We'll probably see some BIOS editors in the future, allowing us to alter this delta. Or even independent clock control sliders for each domain (of course, if NV will decide to provide such functionality in their driver).
4) Digit-Life are currently preparing the review summing all this info.
5) Unfortunately I'm leaving the city for vacation soon, so I'll be probably offline since 11th July till 8th of August. So new RT will be publicly launched when I'll return.

Stay tuned
-------------------------------

This means that unless you reach a multiple of 27Mhz you will not be overclocking the ROP or shader domain, just the geometric domain, so your benchmark scores will not be as good as possible until you reach the next multiplier level.

To test this can someone with a 7800 bench 3dmark03 GT4 nature at 435, 445, 455 and 465 ? The increase should not be linear but spike at 465Mhz when the ROP and shader domains raise to the next level.

Regards

Andy

Geforce4ti4200
07-07-2005, 02:32 AM
wow this is strange! But some cards stop a couple MHz before the next jump then you got no choice. Just clock as high as you can and bench away, linear or not its still gonna help!

mesyn191
07-07-2005, 10:28 AM
Wow they must be doing some seriously wierd stuff with the G70 core, it sure doesn't sound like just a refresh of the NV40 with more pipes that alot of people are saying its, or they just decided to allow previously hidden info. to be out in the open...

Ubermann
07-07-2005, 11:49 AM
http://www.bit-tech.net/news/2005/07/07/g70_clock_speed/

"The chip was designed from the ground up to use less power. In doing that, we used a lot of tricks that we learned from doing mobile parts. The clock speeds within the chip are dynamic - if you were watching them with an oscilloscope, you'd see the speeds going up and down all the time, as different parts of the chip come under load."

So why haven't we heard about this feature before?

"We haven't talked about this feature before now because we wanted the technology to speak for itself," says Nvidia's PR Manager Adam Foat. "People noticed its effect - that the 7800 is amazingly quiet and fantastically cool - and that's what we wanted. "

Found info at www.elitebastards.com

brandinb
07-07-2005, 12:43 PM
i think its a big step forward to implement this technology giving us a quiet cool running but huge performing card

the more effecient the better imo

MaxxxRacer
07-07-2005, 01:11 PM
now that is just freeking wierd....

ALT-F13
07-07-2005, 02:39 PM
Ok I'm too lazy to find cracked 3dm03, so did some testing with 2k1 Nature (single Leadtek card, 'cos my SLI setup don't go higher than 485 on gpus). Forceware 80.40, all stock settings, memory clock always 1200 MHz.

MHz - FPS - FPS delta

430 - 239.4
440 - 240.6 - 1.2
450 - 243.2 - 2.6
460 - 245.5 - 2.3
470 - 246.2 - 0.7
480 - 248.6 - 2.4
490 - 249.5 - 0.9
500 - 251.6 - 2.1
510 - 252.7 - 0.9
520 - 253.3 - 0.6

530 is too much for the card.
So, as you can see, the difference is really not linear and FPS delta jumps from 0.6 to 2.6 for each 10 MHz. BUT! Predicted clocks for bumps should be 430+27=457, 457+27=484 and 484+27=511. Instead of this we can see highest bumps (over 2 FPS/10MHz) between 440 and 450, 450 and 460, 470 and 480, 490 and 500 which don't fit in theory :(

Well, I need RT 15.7 beta to check more precisely, hope to get it ASAP :) Will return with report.

Kanavit
07-07-2005, 02:45 PM
this is interesting stuff. keep us informed.

Cybercat
07-07-2005, 05:23 PM
Interesting. It still doesn't explain why so many people can't break 530 though.

Kanavit
07-07-2005, 08:46 PM
didn't Shamino, and Sampsa broken that barrier with bios update?

Cybercat
07-07-2005, 09:32 PM
didn't Shamino, and Sampsa broken that barrier with bios update?
yes, with their own custom BIOS. Normal people don't get that commodity. There could be a lock, or a problem in syncronization (or lack thereof) with other elements of the chip, a throttling issue (put in there purposefully - essentially another type of lock), or....who knows.

HaLDoL
07-08-2005, 03:01 AM
The barrier with watercooled/aircooled 6800u's was also 520-530

ALT-F13
07-08-2005, 03:23 AM
Ok, 15.7 beta is here. The trouble is that I received it for NDA until one specific article will appear in the russian Net with screenshots and stuff. So until then I'll be unable to tell you _everything_ :) But I promise I will ASAP.

And now I start testing :)

Badong
07-08-2005, 05:54 AM
until one specific article will appear in the russian Net with screenshots and stuff.
iXBT's exclusive or where?

LowRun
07-08-2005, 06:58 AM
Ok, 15.7 beta is here. The trouble is that I received it for NDA until one specific article will appear in the russian Net with screenshots and stuff. So until then I'll be unable to tell you _everything_ :) But I promise I will ASAP.

And now I start testing :)

So, does it make it possible to monitor the three clocks at the same time?

mr_gibs
07-08-2005, 09:49 AM
wow that is strange, but that does explain why ocing my card in small incraments yields little change...

harleybro
07-08-2005, 10:52 AM
Not sure about the whole clock thing. Here's links to a few of my scores. No modded bios just the one that came with the card also no vmods just air on some water on others.
540 clock on 03:
http://service.futuremark.com/compare?2k3=4086104
546 on 05 stock cooler on:
http://service.futuremark.com/compare?3dm05=989200
590 on 05 H2O:
http://service.futuremark.com/compare?3dm05=996175

LowRun
07-11-2005, 06:21 AM
So, does it make it possible to monitor the three clocks at the same time?

Well, i'll answer my own question.

Yes it does.

http://www.tt-hardware.com/img/news4/news120705_4dujour.gif

mr_gibs
07-11-2005, 08:10 PM
can someone tell me if there is a way around this???? my card runs fine and has no artifacts whatsoever @497/1.37 but if i raise either the gpu or ram one more notch it fails to pass the test, i would think it would start artifacting before it fails the test, i don't expect much more from the ram but i hope the gpu can do better, especially watercooled... :confused:

Kanavit
07-11-2005, 08:32 PM
Not sure about the whole clock thing. Here's links to a few of my scores. No modded bios just the one that came with the card also no vmods just air on some water on others.
540 clock on 03:
http://service.futuremark.com/compare?2k3=4086104
546 on 05 stock cooler on:
http://service.futuremark.com/compare?3dm05=989200
590 on 05 H2O:
http://service.futuremark.com/compare?3dm05=996175impressive clocks, looks like Harleybro isn't having the same problems as everyone else with his 7800 gtx. :toast:

t-max
07-12-2005, 04:53 AM
But wich cards are you using Harleybro ??, i got 2 x leadtek gtx on my way

Drow
07-12-2005, 05:13 AM
It seems to be a Gainward or a XFX.

Kanavit
07-12-2005, 08:29 AM
But wich cards are you using Harleybro ??, i got 2 x leadtek gtx on my way
it looks like e-vga 7800gtx (http://forums.amd.com/index.php?showtopic=50287), from the stock hsf.

Shadowmage
07-12-2005, 11:41 AM
Wow, interesting. So which clock is the bottleneck?

Gryph
07-12-2005, 12:01 PM
Is the Riva Tuner 1.57 available yet? And does it fix the overclocking issue?

Gryph

texuspete00
07-12-2005, 02:34 PM
yes, with their own custom BIOS. Normal people don't get that commodity. There could be a lock, or a problem in syncronization (or lack thereof) with other elements of the chip, a throttling issue (put in there purposefully - essentially another type of lock), or....who knows.


They're practically Beta testing... don't be so jealous. :p: I feel certain changes might make it in to newer cards, and the rest of us OC'ers will flash.

I do hope they have the cold throttle at normal temps under control for reg folks anyone know? You know when your plans involve watercooling a card and you have to worry about cold throttle... that just ain't cool.

Kinc
07-12-2005, 07:10 PM
Results From "ALT-F13" (I've added the indivudual clocks):
430 - 239.4 - --- (472/432/432)
440 - 240.6 - 1.2 (480/432/432)
450 - 243.2 - 2.6 (490/459/459)
460 - 245.5 - 2.3 (500/459/459)*
470 - 246.2 - 0.7 (510/459/459)
480 - 248.6 - 2.4 (520/486/486)
490 - 249.5 - 0.9 (530/486/486)
500 - 251.6 - 2.1 (540/513/513)
510 - 252.7 - 0.9 (550/513/513)
520 - 253.3 - 0.6 (560/513/513)
530 - ----- - --- (570/540/540) <- Too high ROP/shader

*Delta seems too high here, everything else seems to make sense. My conclution of the results above is that when the geometric, shader and ROP are syncronized (460(+40)/459/459) in the best possible way, the performence is at its peak.

If you want to run all clocks att the same you can force 2D through RT or in bios and overclock from there. If you set the 2D vcore to 1.4V you should be able to reach high perfect syncronos clocks.

Cybercat
07-12-2005, 08:22 PM
don't be so jealous. :p:
WTF?

Kanavit
07-12-2005, 08:37 PM
Results From "ALT-F13" (I've added the indivudual clocks):
430 - 239.4 - --- (472/432/432)
440 - 240.6 - 1.2 (480/432/432)
450 - 243.2 - 2.6 (490/459/459)
460 - 245.5 - 2.3 (500/459/459)*
470 - 246.2 - 0.7 (510/459/459)
480 - 248.6 - 2.4 (520/486/486)
490 - 249.5 - 0.9 (530/486/486)
500 - 251.6 - 2.1 (540/513/513)
510 - 252.7 - 0.9 (550/513/513)
520 - 253.3 - 0.6 (560/513/513)
530 - ----- - --- (570/540/540) <- Too high ROP/shader

*Delta seems too high here, everything else seems to make sense. My conclution of the results above is that when the geometric, shader and ROP are syncronized (460(+40)/459/459) in the best possible way, the performence is at its peak.

If you want to run all clocks att the same you can force 2D through RT or in bios and overclock from there. If you set the 2D vcore to 1.4V you should be able to reach high perfect syncronos clocks.Kinc, what are those figures stand for? can you explain plz? thanks

Kinc
07-13-2005, 12:08 PM
Kinc, what are those figures stand for? can you explain plz? thanks

(550*/513**/513**)

*Geometric clock
**Shader clock
***ROP clock

ROP and Shader clock kan only work at 432, 459, 486, 513, 540, 567 etc. With an increase of 27MHz.

Ex. If u set the 3d clock to 470MHz (geometric, actuall clock is 510MHz) the shader/ROP will run at 459MHz (the nearest clock available)

harleybro
07-13-2005, 08:26 PM
impressive clocks, looks like Harleybro isn't having the same problems as everyone else with his 7800 gtx. :toast:
Nope all's good. It is an EVGA here's a quick 03 run using my 3500+ :D
http://service.futuremark.com/compare?2k3=4112252

mr_gibs
07-13-2005, 08:30 PM
Is the Riva Tuner 1.57 available yet? And does it fix the overclocking issue?

Gryph

idk if we are talking about the same thing but there's version 2.0 here http://download.guru3d.com/rivatuner/

Kanavit
07-13-2005, 08:30 PM
(550*/513**/513**)

*Geometric clock
**Shader clock
***ROP clock

ROP and Shader clock kan only work at 432, 459, 486, 513, 540, 567 etc. With an increase of 27MHz.

Ex. If u set the 3d clock to 470MHz (geometric, actuall clock is 510MHz) the shader/ROP will run at 459MHz (the nearest clock available)
thx Kinc. i think i am beginning to understand a little better now.

Cybercat
07-13-2005, 08:39 PM
What I don't understand is how performance can actually go down from overclocking. Harleybro's 3DMark05 score went down about 100 points when he went from 546 to 590. You think the current version of overclocking software simply isn't overclocking the ROP/shader clocks to coincide with the geometry clock?

harleybro
07-13-2005, 08:46 PM
Almost 200mhz. less on cpu and worse timings. Sry didn't write the info in. :)

Cybercat
07-13-2005, 08:49 PM
Oh, my mistake.

Hicks
07-14-2005, 07:25 AM
Im not sure i quite follow this.

Ive got a Gainward Golden Sample and it seems to run at a massive 520/100 without me doing anything to it? So what relation has this got with this 3 clocks thing????

Kanavit
07-14-2005, 08:35 AM
Im not sure i quite follow this.

Ive got a Gainward Golden Sample and it seems to run at a massive 520/100 without me doing anything to it? So what relation has this got with this 3 clocks thing????
According to Kinc's figures, it looks like these are your 3 clocks.

520 - 253.3 - 0.6 (560/513/513)