Poor 3DMark03 result :(

[NewbieOneKenobi]

Before: ~5200

Celeron 2.4 GHz
Abit IS7-E
512 MB RAM / 400 MHz (133 used due to processor limitations)
FX5900XT 390@435/700@795
HDD WD1200JB

With another 512 MB RAM and Dual Channel: 5400+

With Pentium 4 2.6c and GF7600GT: 7148...

And mind you that this time, the whole processor-memory bandwidth capability is enabled. It's several times higher - the processor's FSB is an effective 800, the memory is 400 going in Dual Channel for 800. This alone should have given the increase as specified. The card itself is credited with 13000 by the press.

I'm sad... I spent some money on the upgrade and volt-modding my 5900XT (perma-tweaking the volts, adding more cooling, flashing BIOS to 5950 Ultra etc) plus OC-ing the Celeron (they go over 3 GHz on stock cooling if you know what you're doing and have some luck) should probably have given the same.

One thing is that 3D Mark shows my card with 0 clocks for both the processor and memory. Can this be the case? Or is my result actually normal?

3DMark05, however, detects my clocks correctly and gives me 5000 with change, which is correct and looks not so rewarding but mostly fair for my config. Still, what's wrong with 03?

 

[autolex84]

honestly that 2.6c isnt that much faster (limiting at that) compared to the overclocked celeron 2.4. And you did get over a 20% boost in points, so thats not too bad!

You quote the vid card as hitting 13000 by "the press." Well, its possible that they are using something faster on the processor and memory side (even core2) which may account for a lot of that score.

 

[JLK03F150]

IMO you do have a problem. I ran a test for kicks with my system & scored 12k+. I have a 6800Ultra running stock speed. It is supposed to compare fairly even with a 7600GT. My P4 3.4 can use a 14x multi, which dropped the clock speed to 2.8. I then underclocked the fsb to 186mhz (14x186=2607). I compared it against your submitted score. Our cpu scores were only 10 points apart. chevalieratspamless.o2.pl <- that's you, correct?

Here's a few ideas of things to check. I'm sure there's more but I'm drawing a blank.
In the Advance Display Setting are AA, AF, and vsync set to Off or Application Controlled?
How many processes are you running in task manager?
Got any yellow question marks in device manager?
Did you run a video driver cleaner after you updated from your old card?
Consider uninstalling the video drivers, run a driver cleaner, update the chipset drivers, then installing the video driver again.

 

[NewbieOneKenobi]

They got 13400 on an Athlon64 FX-60... And 6000 in 3DMark05. I got 5000 in the latter, but in the former I didn't get 5/6 or 83% of their score (11167) but some measly 7148, which is 53%. It looks like my computer is better with new games than it is with old ones and I really doubt it's all about HT. If I had a bottleneck, it should show in 05 sooner than 03, I reckon. Also, people with similar rigs (Pentium IV class processors up to 3 GHz, 7600GT graphics cards) get around 10000 according to Futuremark's stats database. It's strange...

And yeah, that's my e-mail. I think the settings were set on quality before speed, but I'm not sure about the specifics. In fact, I ran the 05 version on the same settings and still got 83% of what they got with probably a much better CPU and mobo and RAM and a SATA(2) HDD etc. Vsync always off everywhere for performance reasons (I rarely see tearing, whereas I often notice fps reductions quite painfully - even though I'm well used to less than 30). AA and AF probably application controlled. I generally set it to 16 for Warcraft 3 TFT and would probably for some aged games, but not in the general setting.

My hard drive could be the bottleneck but the SE from Western Digital runs more or less comparably to ATA133, even though it's ATA100, so I guess it doesn't suck that much compared to, let's say SATA1. My RAM seems to be a big lover of stock settings, but it's still DDR400/Dual Channel at 800 data rate from CPU, which isn't that bad (~6 GB bandwidth on either side). This IS7-e is not a monster, but it's a good mobo. What else... My 7600GT has a 6-pin additional power connector with a molex 4 adapter that spreads into two molex 4 plugs. I connect both of them. Should I connect only one?

 

[JLK03F150]

I broke that address Just making sure that was you in the search & compare at Orb.
So the 3dM05 score seems to be in line but it's just 3dM03 that scores low? That is strange. I wonder if you just have a bad install of 3dM03?

Both power connectors should be hooked up on the AGP cards.

 

[NewbieOneKenobi]

Yeah, downloading it again and hoping this time it'll be better. I've just reinstalled Windows. We'll see what happens. Can you think of anything specific I might have on/off that reviewers are likely to have off/on respectively? I suppose they should run it on default settings to give a general idea, but who knows. I guess I'll end up volt-modding the card since I've already tested OC and voided the warranty. But if I had wanted that in the beginning, I'd have volt-modded the old MSI 5900XT (and I'm sure it would have been more receptive than this Palit, having been a high-end card and all) and flashed it with a 5950 Ultra BIOS and further overclocked, which would probably have been safer than modding this one. Plus the Celeron 2.4 Northwood would have OC'ed quite well on stock cooling even, as it didn't exceed 54 degrees no matter how I tormented it, and Celerons aren't affected by SNDS... But oh well, I can see what I get out of this. If I don't burn anything, I can always get a C2D/AGP/DDR1 transitional mobo plus the cheapest proc (hello, Allendale, let me introduce you to SLK-497U... ).

Edit: Yeah, noticed. Thanks for breaking it. I've moved on to a gmail account with my domain and "work" aliases because the o2.pl one was getting spammed beyond limits.

Edit2: 3DMark03 still detects my clocks as core 0, mem 0. Additionally, the only options I have in my nVidia control panel seem to be monitor ones.

Edit3: Got 11700. Not bad. It's double my pre-upgrade score, so I'm quite happy with the result. I suppose it's quite good for the hardware generation I have.

 

[JLK03F150]

If you are using the "quaility" setting in nvidia's new control panel the default is set to 4xAA & 8xAF. I just scored a 7857 with the driver set that way. I set everything back to application controlled using the control panel's classic view.

 

[NewbieOneKenobi]

Between quality and efficiency (just one notch, not the maximum), I get about 60-70 points of difference, which is about what I got later on from using different drivers. Between vsync on and off, I got the exact same score and I mean exact, not a single point of difference, while I normally get up to 50 point differences randomly. I ran it on NGO "optimised" drivers before going to sleep and I got a 50 point cut, so now I'm looking for the right version of ForceWare between the latest ones and 84.21 that so many people still respect. Will take me some benchmarking, I guess. BTW, thanks a lot for taking all those benchmarks for me and even adjusting your CPU clock just for that.