Rivatuner 2.04 Released

snaggletooth said:
:) i've been nothing but impressed with the way this card has clocked since installing it at the beginning of september.

i can take the core and memory a little higher too, but it's set 2 'increments' below where i start to have problems in each case.


it'd be awesome to see someone crack into the stable-1700 range with this shader oc... i wonder what it would take.

what would be a good proggy (read: easy benchmark, aside from 3dmark) to use to judge the tangible benefits of the shader clock?

:confused:
Click to expand...

Try this: http://unigine.com/download/

See if the shaders make much of a diff, haven't used that demo/benchmark in a while.
 
I must be missing something here guys. How do you change the ratio on the shader clock. I' running riva 204 and cant figure it out,
Thanks
 
Spirals said:
I must be missing something here guys. How do you change the ratio on the shader clock. I' running riva 204 and cant figure it out,
Thanks
Click to expand...

1. Install 163.67 or .69 beta drivers
2. Install Rivatuner 2.4
3. Click Power User
4. Check box to stop annoying message
5. Open RivaTuner\Nvidia\Overclocking
6. Set your shader ratio in ShaderclockRatio
7. To use Nvidia default ratio enter -1 in ShaderclockRatio
 
ok well since i got home from work and saw all the new stuff released today i have been playing around overclocking tweaking and testing, and i have come accross a couple issues which i am not sure what they mean.

using rivatuner 2.04 163.69 beta drivers

first if i use ati tool to find max clock it scans all the way to 675 and stops there with no errors or artifacts if i then use ati tool to scan for max memory it scans well past 1000 with no errors and not artifacts.

but, if i go into rivatuner and set it 675/1000 i get major artifacts as soon as i fire up ati tool. if i set the core/mem in ati tool to 675/1000 and check for atifacts i won't get any.

so after a lot of setting core/mem in riviatuner then going back to ati to check i have come up with 634/1000 with no artifacts.

but i can set my core/mem to 660/1000 and play games or run 3dm06 without any issues, but will get atifacts if i check. so is ati tool wrong? when i check for atifacts.

second with my core/mem set to 634/1000 ati tool shows it at 634/1000 but rivatuner shows it at core/shader/mem at 621 1512 1000 and i have the shaderratio set to -1

if i bump up my core to 640 rivatuner shows core/shader/mem at 648 1512 1000 but ati tool with show core/mem at 640/1000

i am not sure whats going on which app do i believe? it all just seems weird to me.

thnx for reading my long ass post any thoughts on get max performance would be great,

oh one other thing what ratio should i use to get max performance out of the shaderclock ?

thnx again.
 
so i played around with unigine a little bit and came up with some results...

first i did a stock run, well, stock for my card. 576CORE, 1350SHADER, 837MEMORY. as you'll see, i maxed every detail +4xAA +HDR which are both set in-engine.



avg fps: 34.2 scores: 726



next i did a run at my OC of 648CORE, 1512SHADER(DEFAULT RATIO), 1080MEMORY. again, everything is maxed



avg fps: 39.3 scores: 834. looks about right.




finally, i did a run at my OC+SHADERS@2.5, so 648CORE, 1620SHADER, 1080MEMORY, everything maxed like before.



avg fps: 39.0 scores: 827. bah?


i tried the final run a few more times and came up with very similar results which all hovered around my default shaders result.

anyone else wanna confirm what i've found or debase my findings? looks to me like there isn't much to be gained... well, at least in the unigine bench. :p


i'll try out some other benchies tomorrow to see if i can parallel the results. kinda disappointing at first glance though.
 
Got my Ultra to 690/1728/2302 without artifacting so far with the scan on ATITool though I noticed its very sensitive to temperature even w/ the HR-03 plus. With the fan on high and the temperature cool, no problems, but as soon as it hits a thermal limit, it artifacts. But I haven't tested it too much in depth yet. Might lower the clocks still, might be a core vs. memory issue. Could not get 1784 shader to be stable no matter the multiplier though.
 
snaggletooth said:
so i played around with unigine a little bit and came up with some results...

first i did a stock run, well, stock for my card. 576CORE, 1350SHADER, 837MEMORY. as you'll see, i maxed every detail +4xAA +HDR which are both set in-engine.



avg fps: 34.2 scores: 726



next i did a run at my OC of 648CORE, 1512SHADER(DEFAULT RATIO), 1080MEMORY. again, everything is maxed



avg fps: 39.3 scores: 834. looks about right.




finally, i did a run at my OC+SHADERS@2.5, so 648CORE, 1620SHADER, 1080MEMORY, everything maxed like before.



avg fps: 39.0 scores: 827. bah?


i tried the final run a few more times and came up with very similar results which all hovered around my default shaders result.

anyone else wanna confirm what i've found or debase my findings? looks to me like there isn't much to be gained... well, at least in the unigine bench. :p


i'll try out some other benchies tomorrow to see if i can parallel the results. kinda disappointing at first glance though.
Click to expand...

Could be that there's a "sweet spot" in the ratio of the core speed to the shader speed, where faster shaders end up in a logjam past a certain point trying to sync up with the bus to the other units. This might explain why nVidia was in no particular hurry to enable independent speed control for the shaders--that there would be no benefit outside of a limited range. I look forward to more game tests!
 
Commander Suzdal said:
Could be that there's a "sweet spot" in the ratio of the core speed to the shader speed, where faster shaders end up in a logjam past a certain point trying to sync up with the bus to the other units. This might explain why nVidia was in no particular hurry to enable independent speed control for the shaders--that there would be no benefit outside of a limited range. I look forward to more game tests!
Click to expand...


i was thinking along the same lines. more tests will show gooder conclusions tho :p
 
:confused: RivaTuner's behaving oddly for me... it sometimes is adjusting the clock when I change it, and showing it in the hardware monitor, but even when it does, it isn't adjusting the shader clock. I haven't changed the shader setting, just left it at -1 for default ratio. Any ideas? Vista x64 here.
 
GoldenTiger said:
:confused: RivaTuner's behaving oddly for me... it sometimes is adjusting the clock when I change it, and showing it in the hardware monitor, but even when it does, it isn't adjusting the shader clock. I haven't changed the shader setting, just left it at -1 for default ratio. Any ideas? Vista x64 here.
Click to expand...



be running the atitool fuzzy cube when you're switchin clocks and it will show up in your riva hardware monitor.

2.33 seems to be *default* for me as well, as far as the shader ratio is concerned. both -1 and 2.33 work. there's a range of ratios that will create the default shader but i heard that ratio over on the guru3d forums and it works.

edit: if anyone would be so kind as to elaborate on the very basic mathematics used to calculate the shader scores, i'm sure many 8x00 users would be thrilled
 
KamelRed said:
1. Install 163.67 or .69 beta drivers
2. Install Rivatuner 2.4
3. Click Power User
4. Check box to stop annoying message
5. Open RivaTuner\Nvidia\Overclocking
6. Set your shader ratio in ShaderclockRatio
7. To use Nvidia default ratio enter -1 in ShaderclockRatio
Click to expand...

Thank you. having trouble now with getting riva to detect default clocks, that allows me to oc at all. Thanks for the help though, will keep trying.
 
so, just muckin around some more with the shaders and i got to some benchin with the HL2LC stress test.

first off, here are the settings i used for all the tests


__________________________

round 1, stock. 576CORE.1350SHADERS.837MEMORY



avg fps: 113.54
---
...pitted against a shader ratio of 2.5, so 576CORE.1458SHADERS.837MEMORY



avg fps: 115.59 not bad, 2fps increase with the shader ratio set to 2.5.
____________________________


round 2, heavy OC shaders stock. 648CORE.1512SHADERS.1080MEMORY



avg fps: 128.17 this OC really helps out in source based games with a noticable increase in performance, leaving the shader ratio alone.
------

...now let's OC the shaders with a ratio of 2.5 again. 648CORE.1620SHADERS.1080MEMORY



avg fps: 129.11 at the already high OC i have going, bumping the shaders up to 1620 yeilded a net return of 1 fps. i'd call that within the realm of error one way or the other, but these results are completely repeatable; the shader OC'd number is always higher by at least 1fps at this high OC.
________________________

so although not a huge benefit, the shader ratio adjustment does help eek out a couple more fps with the source engine.


edit: i think i'm coming to realize that the actual benefit of the shader adjustment would be in limiting the shaders to a set frequency and pushing my core even further
 
While true, there is a balancing act to perform. Especially since the G80 has 3 main sections that have to be regulated. A maxed out core vs default memory may not yield as large of a gain if you upped them both (duh). I know what you're saying though ;) :p.

I haven't run any true benchmarks (canned or game-play), but from what I can see myself, I notice that my games run smoother. Perhaps it's the drivers? I can definitely tell that the animation/motion/speed is MUCH more fluid and the hiccups I would get in-game are completely gone now. They greatly decreased in the 163.11/44s, but now they're completely gone. Overall it looks like the shader clock adjustment doesn't bring about too great of a change in fps performance. I was only able to set mine up to 635/1566 (2.5) and still be ATiTool stable (just a few Mhz short on the core of 1620's reach). I believe Unwinder documented that the stable ratio values are from 2.3 to 2.5. I could go all the way up to 688 on the core, but I'd have artifacts (quite a bit in ATiTool) lol.

[in version 2.05...] I've already added a slider for adjusting shader clock and I need a few G8x owners for alpha testing it and brainstorming final shader clock UI implementation. Keeping currently existing driver/hardware ratio limitations in mind (http://forums.guru3d.com/showthread.php?t=238083) I'm still unsure it is is a good idea to provide completely independent shader clock control, probably we should keep it ratio based.

-Unwinder
Click to expand...



I've got one of the original GTS' so I'm stuck with the A2 core, it's brought me great OCs nonetheless.
 
Running at 612/1620 right now with a 2.65 ratio. Stable in ATi Tool, 3dmark06 and ETQW so far. Not too sure about the benefit of bumping the shader clocks higher though. My GTS seems to be pretty much bandwidth limited with AA on so I'm not seeing any significant improvements.

Still very cool to be running at 1.6Ghz+ though :D
 
i don't know if i was clear with what i was trying to say or not lol...

my idea is that if i can lock the shaders at say 1620 like i have them, i think i'll be able to oc the core alot higher than if the shader clocks were increasing via ratio with the core clocks. that's a test for tomorrow tho and i'll have to break out the calculatrice :D
 
snaggletooth said:
i don't know if i was clear with what i was trying to say or not lol...

my idea is that if i can lock the shaders at say 1620 like i have them, i think i'll be able to oc the core alot higher than if the shader clocks were increasing via ratio with the core clocks. that's a test for tomorrow tho and i'll have to break out the calculatrice :D
Click to expand...

Easy. Flash the shader clocks to the card and use RivaTuner (with the ratio blank) and 163.67 or 163.69 to raise core alone. That's just making the ROPs run faster, though, and I doubt it'll do so much.

I've flashed 675Mhz core/1620Mhz shader/1026Mhz mem (2052Mhz DDR) to mine and it's running along happily. (using DOS version of nvflash booted from a memstick cause I'm paranoid like that...)
 
silent-circuit said:
Easy. Flash the shader clocks to the card and use RivaTuner (with the ratio blank) and 163.67 or 163.69 to raise core alone. That's just making the ROPs run faster, though, and I doubt it'll do so much.

I've flashed 675Mhz core/1620Mhz shader/1026Mhz mem (2052Mhz DDR) to mine and it's running along happily. (using DOS version of nvflash booted from a memstick cause I'm paranoid like that...)
Click to expand...


but with the shader clocks being tied to the core via a ratio, would flashing the shader clocks really matter? wouldn't they still rise and fall based on core speed?

well, i guess they wouldn't if you're not raising or lowering your core speed :p
 
How do you underclock to 100/100, won't let me go that low? To extend to life of my notebooks battery a bit more.

*EDIT* Still not working with 8XXXM cards :(
 
My 8800GTS is at 648/1640/1008 and is rock solid. Just raising the shader core from 15XX to 1640 gave me 212 more marks in 3dmark06 and 1 fps in all tests. Ambient temps went up 7C over stock clocks which isn't too shabby as its still below 60C.
 
You must log in or register to reply here.
Back
Top