bucket wrote:

I believe that the Radeon X1950 is the best DX9 card

Why bother? Just go for a HD3850 or HD4670 and never have to worry about upgrading again. They will run DX9 games even faster than a X1950 anyway.

sliderider wrote:

bucket wrote:

I believe that the Radeon X1950 is the best DX9 card

Why bother? Just go for a HD3850 or HD4670 and never have to worry about upgrading again. They will run DX9 games even faster than a X1950 anyway.

The newer AGP cards seem to add some additional CPU overhead. I've tested the X800XT, X1950Pro, and HD3850 on my P3 @ 1596MHz. All three of them were slower than a Radeon 9800 Pro in ALL dx7/8 games and most DX9 games. Same story with a P4-2400/400. However with a S939 X2-4600 at the helm, the newer cards were able to pull far ahead of the 9800 Pro, even at lower resolutions.

Interesting considering R6x0 and later have a command processor that's supposed to make the driver less CPU intensive. But you know marketing.... 😉 Of course who really knows what's going on. P3 does not support SSE2 and that is possibly a performance issue with video drivers and games. There's also the question of that quirky AGP/PCIe bridge chip used by some X800s and all later GPUs.

http://techreport.com/review/12458/amd-radeon … ics-processor/2

Our tour of the R600 began, appropriately, with the GPU's command processor. Demers said previous Radeons have also had logic to process the command stream from the graphics driver, but on the R600, this is actually a processor; it has memory, can handle math, and downloads microcode every time it boots up. The reason this command processor is so robust is so it can offload work from the graphics driver. In keeping with a DirectX 10 theme, it's intended to reduce state management overhead. DirectX 9 tends to group work in lots of small batches, creating substantial overhead just to manage all of the objects in a scene. That work typically falls to the graphics driver, burdening the CPU. Demers described the R600 command processor as "somewhat self-aware," snooping to determine and manage state itself. The result? A claimed reduction in CPU overhead of up to 30% in DirectX 9 applications, with even less overhead in DX10.

swaaye wrote:

Interesting considering R6x0 and later have a command processor that's supposed to make the driver less CPU intensive. But you know marketing.... 😉 Of course who really knows what's going on. P3 does not support SSE2 and that is possibly a performance issue with video drivers and games. There's also the question of that quirky AGP/PCIe bridge chip used by some X800s and all later GPUs.

http://techreport.com/review/12458/amd-radeon … ics-processor/2

Our tour of the R600 began, appropriately, with the GPU's command processor. Demers said previous Radeons have also had logic to process the command stream from the graphics driver, but on the R600, this is actually a processor; it has memory, can handle math, and downloads microcode every time it boots up. The reason this command processor is so robust is so it can offload work from the graphics driver. In keeping with a DirectX 10 theme, it's intended to reduce state management overhead. DirectX 9 tends to group work in lots of small batches, creating substantial overhead just to manage all of the objects in a scene. That work typically falls to the graphics driver, burdening the CPU. Demers described the R600 command processor as "somewhat self-aware," snooping to determine and manage state itself. The result? A claimed reduction in CPU overhead of up to 30% in DirectX 9 applications, with even less overhead in DX10.

If SSE adds so much performance, then that might be reason enough to build a Willamette P4 system instead of a Tualatin.

sliderider wrote:

swaaye wrote:

Interesting considering R6x0 and later have a command processor that's supposed to make the driver less CPU intensive. But you know marketing.... 😉 Of course who really knows what's going on. P3 does not support SSE2 and that is possibly a performance issue with video drivers and games. There's also the question of that quirky AGP/PCIe bridge chip used by some X800s and all later GPUs.

http://techreport.com/review/12458/amd-radeon … ics-processor/2

Our tour of the R600 began, appropriately, with the GPU's command processor. Demers said previous Radeons have also had logic to process the command stream from the graphics driver, but on the R600, this is actually a processor; it has memory, can handle math, and downloads microcode every time it boots up. The reason this command processor is so robust is so it can offload work from the graphics driver. In keeping with a DirectX 10 theme, it's intended to reduce state management overhead. DirectX 9 tends to group work in lots of small batches, creating substantial overhead just to manage all of the objects in a scene. That work typically falls to the graphics driver, burdening the CPU. Demers described the R600 command processor as "somewhat self-aware," snooping to determine and manage state itself. The result? A claimed reduction in CPU overhead of up to 30% in DirectX 9 applications, with even less overhead in DX10.

If SSE adds so much performance, then that might be reason enough to build a Willamette P4 system instead of a Tualatin.

I don't believe the absence of SSE2 is what's causing the additional CPU overhead. After all, I had the same "problem" with a Northwood 2400/400.

I think swaaye may be correct about the bridge chip. Those PCIe-AGP chips can do some very strange things. I had a hell of a time flashing the BIOS on my x1950 because of the Rialto bridge. And the HSI bridge on my 7800GS AGP would trick older Forceware drivers into believing that I had a multi-GPU, SLI-ready system. 😒

I do also own an old IBM server system with dual 3.06ghz pentium 4-based xeons. Not sure of what core. Maybe Northwood? If anyone is interested I could drag that loud, noisy, hot beast out and run some benches. Didn't know if it would get any love on here, seemed to be a little overpowered compared to what is usually posted.

I have access to an x800gto, in a computer I built for an elderly family member. I'm sure they wouldn't notice if I swapped it out for a geforce4 mx. All they do is basic web surfing.

Still planning on doing some more benchmarks on the p3 system. Will get it back out this weekend and run it through some various benchmarks with one and two CPU configs and different sound cards. Threw out several of the older AGP cards I accumulated over the years a while back... Some TNT2s, various 3dlabs workstation cards, handful of 3DFX cards, various X800s in differing varieties... so I don't have many graphics cards I can bench with.

i over clocked the shit tits out of a 2.8 p4 to 3.2x. paired with an hd2600 pro/xt and it plays YT just fine it can even run stalker clear sky.

feipoa wrote:

Blue ZIF's and a never soldered-in ISA slot. What happened here?

If one would solder the missing ISA slot, would it actually work?

Win7 would probably run fine on that machine, but only after extensive tweaking/maxing out the RAM. 3GB would be plenty. Assuming there's driver support, 32-bit Win7 would give you the best of both worlds with modern computing and software compatibility. The thread management is vastly improved, and discrete CPUs beat cores or hyperthreading any day.

Standard Def Steve wrote:

sliderider wrote:

bucket wrote:

I believe that the Radeon X1950 is the best DX9 card

Why bother? Just go for a HD3850 or HD4670 and never have to worry about upgrading again. They will run DX9 games even faster than a X1950 anyway.

The newer AGP cards seem to add some additional CPU overhead. I've tested the X800XT, X1950Pro, and HD3850 on my P3 @ 1596MHz. All three of them were slower than a Radeon 9800 Pro in ALL dx7/8 games and most DX9 games. Same story with a P4-2400/400. However with a S939 X2-4600 at the helm, the newer cards were able to pull far ahead of the 9800 Pro, even at lower resolutions.

what about on socket A? and what OS? i used the HD2600 AGP with a Pentium 4 on windows 7 and it could even handle some HD content, as to the overhead i wouldn't know as i never compared it to any thing.

Schizofrik wrote:

Here is a link to a few photos: http://imgur.com/a/ZJ6U2

For some reason, I really like dual or even quad cpu's in favour of dual/quad-core systems.
It just look's so much better, having duble cpu-fans. Really like it.

brostenen wrote:

Schizofrik wrote:

Here is a link to a few photos: http://imgur.com/a/ZJ6U2

For some reason, I really like dual or even quad cpu's in favour of dual/quad-core systems.
It just look's so much better, having duble cpu-fans. Really like it.

I agree - in some way... These configs are fascinating.