Sign in with
Sign up | Sign in
Your question

ForceWare 174.74, the 9800GTX secret helper???

Last response: in Graphics & Displays
Share
a b U Graphics card
April 3, 2008 1:53:34 AM

Ok, I tried to make the title as eye-catcher as possible to get a few people to spend a few seconds to read this.

Since yesterday's release of the 9800GTX, many threads opened about whether or not go for a 9800GTX (either for a new system or EVGA's Step-Up program) over a 8800GTS; I myself wondered if I should pay the 45$ it would cost me to step-up my 512-P3-N841-A3 (8800GTS factory clocked @ 670MHz).

Looking to various reviews, I noticed that most tests were made using the cards "official" drivers, that is 169.25 for GF8 and 174.74 for GF9. Yeah, you see where I'm going with that ... Being a numbers freak, I did a little comparison, on my own system of how the 174.74 drivers would impact my performance (got the here).

I did my tests with Crysis (Crysis benchmark Tool) on Windows Vista Ultimate x64, 5 passes for each test. These tests were made on a recent installation of Vista (2 weeks old, not much installed yet), but not in a strictly controlled environment (still had AVG, Pidgin, ... running in background). Here is a summary of the results:



I encourage anyone to try it out (even on a few resolutions) and post their findings.

For those who want to know my rig's spec, just look in my profile, it's all there.
April 3, 2008 4:13:27 AM

WARNING DO NOT CLICK ON THE ABOVE IMAGES - THEY WILL TRY TO LOAD JAVA SCRIPT
a b U Graphics card
April 3, 2008 5:36:18 AM

As far as I know the 8800GTS 512 (G92) has exactly the same GPU as the 9800GTX, just with slower clocks. It has been a long time (Dec 2007) since nVidia has released a new driver for the series 8 cards. I am sure once they roll up the driver improvements found in the 174 driivers into the unified driver, all the G92 core GPU's (possibly G80 as well, depending on whether the improvements are general application improvements or architectural improvements) should see the same sort of improvements.
April 3, 2008 5:49:48 AM

That takes away any reason to step up save the name.

I was just going to change to these, so I'll post improvements as well, when I do.
April 3, 2008 6:13:19 AM

I ran those beta drivers with my 8800gts and saw a 3dmark improvement of about 200. I dont know if that adds anything to this topic because I cant say I understand it completely.
April 3, 2008 7:01:30 AM

I can't seem to install the 64-bit Vista drivers. I get errors stating it can't detect the hardware the drivers support, and that I am trying to run a 32-bit installed on a 64-bit OS.

I downloaded the 64-bit version.
April 3, 2008 7:58:22 AM

Anybody remembers the Geforce2 "Ti" series? The new cards were rebranded and boasted a 28-50% performance improvement over the regular model's performance upon release, but anybody with an older model could update their card's driver and get those same performance improvements for the price of a download (as a side note, even older chips - like my TNT, NV4 - got a definite boost from the driver update).

More recently, we had the Geforce6 and Geforce7 families; asking people who seem to know quite a lot about these (the Xorg Nouveau staff, currently writing a reverse engineered driver, with working 3D for these cards), Geforce6 (especially later chip revisions like 6600) and Geforce7 have a lot in common, and are in fact addressed by the official driver as being the same family -just different chip revisions.

Having tested both NV1x chip revisions (as the "ti" chip was, indeed, slightly tweaked), I'd say that the G9x could (if the parallel persists), indeed, contain some fixes over erratas and may indeed be much better performing in as of yet unseen corner cases, but that's it - if you already have a top-notch G8x chip, don't upgrade (except if it's a 8600, as the 9600 does indeed leave it in the dust). Remember, if you were the proud owner of a 6600GT, getting a 7600 card wasn't really worth the money...

For those wondering: the "ti" chip revision for Geforce2 (NV1x family) seems to perform much better on some VRAM-intensive operations (like GL_EXT_texture_from_pixmap) than the original, with the very same driver revision. Said extension was added to the NVIDIA driver 2 years ago, so...
April 3, 2008 8:35:38 AM

I've been running 174.20's for about 6 weeks or so now.
After installing my 2nd GTX I was getting big lock-ups in CoD4 the moment I had turned SLi on, so as recommended by pauldh (thanks! :) ) I tried the drivers from Guru3D. Performance was much, much better and I didn't have all the damn locking up either! I should add I was running the 169.21's prior to that, at the the time the mose up-to-date, and I was getting lock-ups, thanks Nvidia! :kaola: 
I'll try the 74's later I think :) 
a b U Graphics card
April 3, 2008 10:20:30 AM

dragonsprayer said:
WARNING DO NOT CLICK ON THE ABOVE IMAGES - THEY WILL TRY TO LOAD JAVA SCRIPT
I always hosted my images on imageshack, any alternatives?
a b U Graphics card
April 3, 2008 11:00:58 AM

nkarasch said:
I ran those beta drivers with my 8800gts and saw a 3dmark improvement of about 200. I dont know if that adds anything to this topic because I cant say I understand it completely.
My point was that you could probably get 9800GTX performance from a 8800GTS (G92) simply by upgrading drivers. The 3DMark improvement might not seem like much because default 3DMark tests are rather "low-res", don't seem to use AA or AF and the driver seems to have better improvement more in the upper range.
a b U Graphics card
April 3, 2008 11:15:04 AM

Zenthar said:
I always hosted my images on imageshack, any alternatives?

Flicker.com

side note: The 174.74 drivers don't work right with Rivatuner/AtiTool with XP/Vista x64. The OCing tools are unable to stick the OCing changes. They revert back to stock.
April 3, 2008 11:20:43 AM

Im still waiting for somebody to test the 9800GTX and 8800GTS with the same drivers at the same clocks. I wonder what the difference would be.
April 3, 2008 11:22:39 AM

The card I have now if better :p  lol. I will keep her for a long time. I will be getting one of these Q9450s however though.
April 3, 2008 11:41:20 AM

mitch074 said:
Anybody remembers the Geforce2 "Ti" series? The new cards were rebranded and boasted a 28-50% performance improvement over the regular model's performance upon release, but anybody with an older model could update their card's driver and get those same performance improvements for the price of a download (as a side note, even older chips - like my TNT, NV4 - got a definite boost from the driver update).

More recently, we had the Geforce6 and Geforce7 families; asking people who seem to know quite a lot about these (the Xorg Nouveau staff, currently writing a reverse engineered driver, with working 3D for these cards), Geforce6 (especially later chip revisions like 6600) and Geforce7 have a lot in common, and are in fact addressed by the official driver as being the same family -just different chip revisions.

Having tested both NV1x chip revisions (as the "ti" chip was, indeed, slightly tweaked), I'd say that the G9x could (if the parallel persists), indeed, contain some fixes over erratas and may indeed be much better performing in as of yet unseen corner cases, but that's it - if you already have a top-notch G8x chip, don't upgrade (except if it's a 8600, as the 9600 does indeed leave it in the dust). Remember, if you were the proud owner of a 6600GT, getting a 7600 card wasn't really worth the money...

For those wondering: the "ti" chip revision for Geforce2 (NV1x family) seems to perform much better on some VRAM-intensive operations (like GL_EXT_texture_from_pixmap) than the original, with the very same driver revision. Said extension was added to the NVIDIA driver 2 years ago, so...


"Remember, if you were the proud owner of a 6600GT, getting a 7600 card wasn't really worth the money..."

Mate I had both a 6600gt and a 7600gt and the difference was ENORMOUS. If you meant 6800gt then yeah fair enough.
April 3, 2008 11:48:38 AM

techgeek said:
As far as I know the 8800GTS 512 (G92) has exactly the same GPU as the 9800GTX, just with slower clocks. It has been a long time (Dec 2007) since nVidia has released a new driver for the series 8 cards. I am sure once they roll up the driver improvements found in the 174 driivers into the unified driver, all the G92 core GPU's (possibly G80 as well, depending on whether the improvements are general application improvements or architectural improvements) should see the same sort of improvements.


If thats true that nvidia hasn't made a driver for G80 cards since december, then thats pretty shocking indeed. The 174 drivers inflate the 9800gtx & gx2 scores a bit against the old gtx and newer 512mb gts. The 9800gtx sooo isn't worth the premium when you could be a good brand factory oc'd gts and up the clocks a bit more.
April 3, 2008 12:25:34 PM

nkarasch said:
http://en.expreview.com/2008/04/03/geforce-9800gtx-review/12/


Wow!

The 9800GTX will definately have more headroom and beat the GTS when overclocked but the GTX is a waste of time. My GTS overclocked to 820 core, I wonder how high the 9800 will go



Probably about the same... 820/2008/2204 is what my 8800GTS 512 will do.
a b U Graphics card
April 3, 2008 12:28:53 PM

Shadow703793 said:
side note: The 174.74 drivers don't work right with Rivatuner/AtiTool with XP/Vista x64. The OCing tools are unable to stick the OCing changes. They revert back to stock.
Might be so, I don't overclock, my EVGA came factory clocked @ 670MHz. Any specific version of RivaTuner causing problem? The 2.08 version is said to support 174.XX drivers and x64 (anyone can confirm?).
a b U Graphics card
April 3, 2008 12:55:42 PM

nkarasch said:
http://en.expreview.com/2008/04/03/geforce-9800gtx-review/12/


Wow!

The 9800GTX will definately have more headroom and beat the GTS when overclocked but the GTX is a waste of time. My GTS overclocked to 820 core, I wonder how high the 9800 will go
Very good article, many numbers to support the fact that 9800GTX isn't really worth the extra cash. Too bad I paid a bit more for an EVGA card (300$) for the "Step-up option" to 9800GTX ...
April 3, 2008 1:33:36 PM

I mentioned this back in the GTX review thread. It is good to see someone else bring this up. It is comparing apples to oranges when doing a vs benchmark test if you give the GTX fresh drivers. It produces a false sense of performance gains.

I am sticking with the name 8800 GTS v3.0, I refuse to recognize it as a GTX.
April 3, 2008 1:55:37 PM

Big trouble in little china for NVIDIA wow,, I like and have a card of theirs to so, no fan boy stuff just wow,
April 3, 2008 2:19:00 PM

That's quite pathetic; I just downloaded the new drivers for my 8800GTX but have yet to run anything. It's sad how Nvidia has simply re-branded old hardware with a new naming scheme and a new set of drivers to try and pretend it's better hardware.
a b U Graphics card
April 3, 2008 2:37:31 PM

To their defense, they might have optimized the PCB and improved some component's quality so I consider this a bit more than simple rebranding. What I have trouble with is the price difference; to me, the value of a 9800GTX over a 8800GTS is about 25$ top.

Other than that, I would like to see how far the 9800GTX overclocks. If it can be pushed farther than 8800GTS, then it might be "worth it" (like DDR2-800+ might be for some overclockers).
April 3, 2008 3:19:34 PM

If I read right, even tough the 174.74 mention only supporting the new 9 series of GeF VPU, I can install it for my 8800GT? I'm on Vista 64-bit, if that change anything. I'm getting fed up of using the same old Dec2007 169.25 driver, I'll be more then glad to upgrade to this one if possible.
a b U Graphics card
April 3, 2008 3:24:57 PM

NightlySputnik said:
If I read right, even tough the 174.74 mention only supporting the new 9 series of GeF VPU, I can install it for my 8800GT? I'm on Vista 64-bit, if that change anything. I'm getting fed up of using the same old Dec2007 169.25 driver, I'll be more then glad to upgrade to this one if possible.
Running Vista Ultimate x64 here and I got my drivers from Guru3D, I think you could find them in the beta section of NVidia's website as well.
April 3, 2008 3:35:56 PM

It's called better process over time. As time pass better yields which equate to better stock clocks and overclocking ability.

8800gts have better stock clocks so it overclocks a little better than 8800gt and 9800gtx does the same with 8800gts.

9800gtx is still based on G92 chip. Maybe a little tweak or not even.
a b U Graphics card
April 3, 2008 8:19:44 PM

^ Probably volt mod and RAM heatsinks.
April 3, 2008 8:51:29 PM

Did he now see my post? I got 820/2008/2248~ which equals 1124. I have heat sinks on them and no volt mod.
a b U Graphics card
April 3, 2008 10:30:23 PM

truehighroller said:
Did he now see my post? I got 820/2008/2248~ which equals 1124. I have heat sinks on them and no volt mod.
Cool, got an EVGA card too so I will cross my fingers.
April 4, 2008 5:05:00 AM

Zenthar said:
Anyone knows how he was able to get his VRAM to 1100MHz, watercooling or something?

9800GTX uses better ram than the G92 8800's, Samsung I think.
It's also rated @ 2200mhz, as opposed to 2000mhz.
April 4, 2008 8:30:30 AM

So then my 1024MB GTS @ 800/2000/2200 out does an ultra and an oc'd g92 GTX? Very happy with my purchase in that case...
a b U Graphics card
April 4, 2008 10:17:41 AM

LAN_deRf_HA said:
So then my 1024MB GTS @ 800/2000/2200 out does an ultra and an oc'd g92 GTX? Very happy with my purchase in that case...
Might be so yes, do a little benchmark with latest beta drivers and you'll know.
a b U Graphics card
April 4, 2008 7:44:31 PM

truehighroller said:
Did he now see my post? I got 820/2008/2248~ which equals 1124. I have heat sinks on them and no volt mod.

I stand corrected ;) .
April 4, 2008 8:06:04 PM

Is there a way to find out which company made your videocard memory without taking the heatsink off?
!