View Full Version : So, I got the Det. 52.14, there's a problem.... :(
I've just downloaded the Inno3D 52.14 Detonators for WinXP and made a 3D MARK2001SE run.... everything is OK xcept for the Dragotic Low and the Nature scores..... they're BIG GARBAGE!!!! :(
I'm just aming for 22k and with a 2.3G A64/252FSB and a FX5900 500MHz core/920MHz mem It seems possible.... just that weird fps drop in those two...
any solution?
btw. with the 44.03 everything is OK, the Nature and Dragotic Low are OK.
Soulburner
10-18-2003, 05:51 PM
Any driver past 44.03 got slower and slower. They aren't good for 3DM01. Your Nature should be about 150.
Soulburner
10-18-2003, 05:52 PM
Don't you think ATI has optimizations too....ever wonder how they score 100fps more on drag low than Nvidia, yet Nature is almost the same and that test is all card....:rolleyes:
Originally posted by MickeyMouse
lol why on the 44.03's was all good....
becasue of optimizations nvidia added and now they are using "real" drivers you get just what you got a lower score.. now Im sure image detial has gone up, but all drivers after 45.23's had optimaizations for 2k1 removed
Hahahaha, Idiots.... why the hell am I trying with that NV anyway..... oh, well..... will try again with the 44.03 and then back to the R9800Pro.... it's better in any way....
Soulburner
10-18-2003, 06:20 PM
Like I said, all drivers after 44.03 are the same in 3DMark01. 45.23 on is slower.
And about the DX9 + PS deal, Nvidia got screwed because their code path was a helluva lot more complicated than ATIs. After close work with Nvidia, Microsoft is releasing DirectX9.1, and along with 52.16 WHQL, and the 5950, Nvidia should be doing fine in DX9. The next month or so will be interesting to see how it plays out.
As for 3DMark, ATI will always be the benchmark whore. Funny thing is that's why i'm getting one.....I need help :D.
I'm back to the 44.03 and it's all good now :)
Will stay there for a day and will jump rapidly to a 9800Pro, where I belong :D
shrae
10-20-2003, 12:24 AM
Originally posted by Soulburner
And about the DX9 + PS deal, Nvidia got screwed because their code path was a helluva lot more complicated than ATIs. After close work with Nvidia, Microsoft is releasing DirectX9.1, and along with 52.16 WHQL, and the 5950, Nvidia should be doing fine in DX9. The next month or so will be interesting to see how it plays out.
When will DX9.1 be out by? And can MS possibly change DX9 so much that it can overcome nVidia's hardware deficiencies (no 24-bit shader) that we've heard so much about? Obviously they can optimize DX9.1 codepath for 32-bit shaders (instead of 24-bit as it is currently), but that won't change 32-bit being slower than 24-bit.
Soulburner
10-20-2003, 02:25 AM
Well if they can get it working fast in 32 bit Nvidia will have an advantage because the most ATI can currently do is 24 bit since its all the hardware supports. Nvidia does 16 or 32.
st0nedpenguin
10-20-2003, 03:00 AM
Way to go nVidia, everyone else creates DirectX 9 compliant hardware, they just insist on changing DirectX to suit their hardware.
shrae
10-20-2003, 07:59 AM
Ah, I just did some reading on DX9 - I didn't realize that DX9 doesn't support anything but 24-bit shaders. If DX9.1 has optimizations for either 16-bit or 32-bit, then yeah, I can understand the performance increase.
Originally posted by st0nedpenguin
Way to go nVidia, everyone else creates DirectX 9 compliant hardware, they just insist on changing DirectX to suit their hardware.
In nVidia's defense, the 5800/5900 tapped out a long time before DX9 was released. I believe actually before Beta2 even. So did the R300, and 24-bit shaders on ATI's part was a gamble just like nVidia doing 16/32-bit. Let's say DX9 wound up supporting 32-bit instead (and it well could have, according to xbitlabs' interviews with various game developers who were surprised that DX9 only supported 24-bit). Then it would be ATI's customers who would have gotten screwed, and perhaps even ATI who would have "cheated" on drivers by reducing all shading to 24-bit to be compatible with their hardware (just as nVidia reduced 24-bit to 16-bit). The better solution, of course, would have been for DX9 to be fully compatible with 16, 24, and 32-bit shading in the first place.
And in that sense, we can actually blame MS. Remember that optimization codepaths that originally were relegated to be handled by drivers in the pre-DX8 era are now being built into DX. DX8.1 had at least 5 (maybe more) pixel shader optimizations which allows flexibility of hardware. So with regards to your "everyone else" statement, one of the main reasons to move from driver to standard optimization is so that "everyone else" will have compatability without having to develop specific codepaths in drivers to do so. But clearly it failed this time with MS' choice to only use 24-bit PS.
Honestly, the more I read about DX9 and the whole PS controversy, the more it starts sounding like a conspiracy. It's funny how people will take a story, jump to conclusions, and just run with it.
st0nedpenguin
10-20-2003, 08:22 AM
But surely MS released the specs for DirectX 9 to the hardware manufacturers before release, in order for them to have some idea of what to aim for hardware wise?
shrae
10-20-2003, 08:40 AM
That's something that is hard to find the answers for. The big question I see asked a lot on many of the tech websites is not whether nVidia had a pretty good idea that DX9 would focus on 24-bit, but rather if they knew that DX9 would ONLY support 24-bit. It is hard to believe (yet possible certainly) that nVidia would continue hardware production with a totally incompatible shader... there are obviously a lot of associated costs with revising hardware after a tapping out, after all. Then again it is just as hard to believe that MS would release DX9 with only 24-bit PS all the while knowing that nVidia's hardware wouldn't support it. My biggest problem with this issue is that it just doesn't add up from any of the perspectives you look at it from. I am definitely not a conspiracy theorist, but I can see where some people would start here.
I guess it could also be naivety on nVidia's part. Perhaps they believed that DX9 would eventually come out with 16, 24, and 32-bit PS compatability no matter what. Remember that game developers want this flexibility as well, so it's not just nVidia asking for the changes that (hopefully) we'll see in DX9.1. The benefit of standard as opposed to driver optimization applies to software devs just as it does for hardware companies.
No doubt it will be interesting to see how all this drama plays out if DX9.1 actually supports nVidia cards. I just bought a new 9800NP tho, so I will be waiting for the NV40 vs. R420 next year for my next purchase.
mluckey
10-20-2003, 09:15 AM
Your statements jibe with everything that I've read about the 24 bit shader dilema. NVidia just assumed that the 32 bit support would follow, when in fact, it wasn't part of the plan by DirectX programmers.
As far as Detonator 52.14 performance. I ran close to 30 benches, an two hours game time in UT2003. Acroos the board they looked nicer, but were consistantly slower than the fastest that I have used, the Omega Detonator 45.23.
Soulburner
10-20-2003, 12:31 PM
The fastest drivers were the 44.03, but image quality improvements and bug fixes have been made since then so there is a reason for a little loss in older games but in DX9 games performance has increased three fold at least.
Nebulous
10-20-2003, 12:37 PM
The 52.14 dets are still in beta. Not whql
Soulburner
10-20-2003, 12:47 PM
Originally posted by Nebulous
The 52.14 dets are still in beta. Not whql
Nope they are out and are WHQL. The final build is 52.16.