Futuremark 06 Benchmark results on AGP 1950Pro

[HeadRusch]

Interestingly, their ORB browser says I am #2 in line (out of 4 entries).
My score of 3796 was close to 3rd place, but 1st place was up with like 4200 points.

The interesting thing is that the only real difference I see between his system and mine, besides some basic parts, is that he's running the 512meg board. Every review I saw said that the 512meg parts were no faster, in some cases slower, and required *twice* the power of the 256meg version of this card. :P

But in looking at the specific scores, there isn't much difference. Where I averaged 15fps he/she averaged like 19 or 20......

Anyhow, I guess 3796 is a respectable score in 3dmark 06 for an AGP system on an old Intel P4 platform....no A64 or dual core, and those would all be PCIe anyhow.

<shrug>

Playing single player BF2 with 4xaa and 4xaf I never had a problem with the framerate......I also decided that FRAPS is evil....I'd be playing fine, having a good time, not detecting any choppiness or slowdown...then I'd fire up fraps and see "my god I'm only at 42fps!". Ugh....evil. What you dont know wont hurt you :)

[SteelWill]

I'm not sure where those power claims come from, but there's no way a 512mb card requires twice the power of a 256mb version of the same card. It will require more power, IF their are actually twice as many chips, but the core is the big power drain on these things. And if they are using the same number of RAM chips just of a higher density then it shouldn't require any more power at all.

As for the performance difference, in most games up to this point there isn't one. Oblivion I think is one game where the difference is significant, and there might be a couple of others as well. It's entirely dependent on the app, so just because in one game it doesn't make a difference, you can't make the blanket statement to all games (and I do see A LOT of people make that statement). Also, in future games, that extra RAM is more likely to make a difference than not.

I'm not trying to crap on your purchase, you got a good card and it's going to give your rig much longer legs, but I think you got some bad info that might have negatively affected your decision and given you some expectations that were a bit higher than they should have been. But really, IMO the 512mb AGP x1950 Pros are overpriced anyway (heck, I got a PCIE version and a new mobo to use it in for less), so just think of it as money saved to go towards your next build. When things like Crysis and UT2K7 come out I highly suspect either card is going to start to feel very dated.

[HeadRusch]

I'm not sure where those power claims come from, but there's no way a 512mb card requires twice the power of a 256mb version of the same card. It will require more power, IF their are actually twice as many chips, but the core is the big power drain on these things. And if they are using the same number of RAM chips just of a higher density then it shouldn't require any more power at all.


Well, you can take that up with Sapphire who require 2 PCI-E molex-ish connectors, and they have to be seperate connectors, can't use the same 12v rail, on their 512 meg card. If you try to use the same rail, the card wont power up.
Don't tell me, go tell Sapphire (currently the only company out with a 512 meg card last time I checked).

As it is ATI must be smoking crack....their recommendations are that you feed the GPU with a power supply that has at least 30 AMPS on the 12v rail. Now the 1950 is a modern GPU yes, but its a little bit gimped...a mainstream card really.....whys it sucking so much juice. 30amps? They kill people with 30 amps!

Now, not a whole hell of alot of people have a $180+ power supply that would deliver 30amps on a single 12v rail, particularly in a 2 year old AGP system.

I am hoping thats a misprint and they mean 15 amps on the 12v rail, X2 if you bought the 512 meg card.


As for the performance difference, in most games up to this point there isn't one. Oblivion I think is one game where the difference is significant, and there might be a couple of others as well.


Oblivion I can see.....its one gigantic massive detailed environment with tons of sprite-based foliage...I'll look into those benchmarks, see what the difference is.


It's entirely dependent on the app, so just because in one game it doesn't make a difference, you can't make the blanket statement to all games (and I do see A LOT of people make that statement). Also, in future games, that extra RAM is more likely to make a difference than not.


Dont accuse me of making a blanket-statement when you just made one yourself: "In future games that extra RAM is more likely to make a difference than not".

How long have we been hearing that? For years we've been seeing a whole 3-5fps increase in performance with the extra ram added, all it came down to was texture swapping. The payoff for doubling your video memory is currently mostly nil.


I'm not trying to crap on your purchase, you got a good card and it's going to give your rig much longer legs, but I think you got some bad info that might have negatively affected your decision and given you some expectations that were a bit higher than they should have been.


I knew going into this that the card wasn't an 8800GTX....my goal was 1080p at 60fps. I'm close to or above that with this card on all the games I currently play, so I'm satisfied.


But really, IMO the 512mb AGP x1950 Pros are overpriced anyway (heck, I got a PCIE version and a new mobo to use it in for less), so just think of it as money saved to go towards your next build. When things like Crysis and UT2K7 come out I highly suspect either card is going to start to feel very dated.

Agreed...mainstream cards are the first to fall to new technology. The interesting thing will be to see how UT3 scales with older hardware. Since its likely to be shader heavy, this card may still run it acceptably....tho add HDR into the mix and I'll likely be dropping the resolution down and scaling to my display...we'll see.

[SteelWill]

Sorry if parts of my my post came off as confrontational, I didn't intend for it to be.

Sapphire's power reqs are bogus. The only hardware diffence between AGP and PCIE parts are the PCB and the bridge chip on the AGP card. There's no reason in the world why power consumption wouldn't be virtually the same between them. The only difference is how that power is delivered. PCIE gets a good chunk through the mobo, AGP needs the extra connectors.

As for the extra RAM, I was really directing my comment about that to all those reviews that just bench HL2 and maybe a couple of other games and from that declare that there's no difference from the extra RAM. You said yourself it seemed to be making a difference of about 25% (15 vs. 19 or 20 fps) in certain instances between your system and the one you were comparing it to.

[HeadRusch]

np, didn't think you were throwing gas on a fire or anything......I'm just quoting what they're printing, which seems absurd if those requirements are true.

30amps on the 12v rail...they're crazy!

[StreetPreacher]

Anyhow, I guess 3796 is a respectable score in 3dmark 06 for an AGP system on an old Intel P4 platform....no A64 or dual core, and those would all be PCIe anyhow.


Thanks for posting your score, I was worried this Radeon was going to crush the 7800GS I grabbed in the summer (didn't really want it but the old card died), but it's not quite as bad as I had worried :).

Would you mind breaking down the individual sm/hdr/cpu scores? I just ran 3DMark06 and scored 3302 with my setup (p4 3.2@3.2/2gig), so I'd be interested to see exactly how things stack up.

[HeadRusch]

Heres how it lays out:
Abit..or asus..motherboard...meh, whatever! :)
Intel 3.0Ghz NOrthwood o/c'ed to 3.45Ghz
2 gigs of Ballistix ram running at...I dunno..whatever speed to reach 3.45Ghz..
PowerColor Radeon X1950 Pro AGP in 8x mode.
Drivers set to "let application decide" so I think these tests all run with 4xAF turned on for 3DMARK06, but I could be wrong....

3DMark Score 3743 3DMarks
SM 2.0 Score 1733 Marks
SM 3.0 Score 1937 Marks
CPU Score 733 Marks

Detailed Test Results

Graphics Tests
1 - Return to Proxycon 13.362 FPS
2 - Firefly Forest 15.515 FPS

CPU Tests
CPU1 - Red Valley 0.225 FPS
CPU2 - Red Valley 0.382 FPS

HDR Tests
1 - Canyon Flight (SM 3.0) 18.759 FPS
2 - Deep Freeze (SM 3.0) 19.984 FPS

[HeadRusch]

UPDATE: I overclocked the 1950pro, bumped the core up to almost 608 and the mem went up to like 722.

3DMARK06 score jumped massively to....3792. :P

I have read online that supposedly when you increase the RAM speed on the 1950's, the card automatically adjusts the memory timings, thereby slowing down the card...all you get is more heat :(

The bump in the core speed means this card is either locked somehow, or its design is simply maxxed out in its current iteration.

The curious thing is how some companies are selling the factory-overclocked cards on NEWEGG that run just a bit faster than stock...I wonder if the benchmarks on those are identical like mine are?? Or perhaps 3DMARK06 isn't a good litmus?

When I bumped up the speed in 3DMARK03 I got a higher score...could be that at 1024x768(native rez of 3dmark03) my CPU isn't bottlenecking the card, but at 1280x1024 my CPU is necking the card, along with the types of tests being run....(the rez 06 runs at)

(shrug)

[ChrisFB]

When I bumped up the speed in 3DMARK03 I got a higher score...could be that at 1024x768(native rez of 3dmark03) my CPU isn't bottlenecking the card, but at 1280x1024 my CPU is necking the card, along with the types of tests being run....(the rez 06 runs at)

(shrug)

That was my first thought. I don't see any other explanation - but because I don't see it doesn't mean it is necessarily so.