Definitive 2d Performance Thread - need your help!

[Skaven]

Ok folks - here is the story. I built what I thought would be the ultimate performance machine for the money to whip out 2D and 3D images for the Visualization lab here at the university. The specs are as follows:

AMD Athlon 64 4000+
Asus A8N-SLI Deluxe
EVGA 6800 Ultra x 2 (SLI and non SLI, explained later)
2gb PC3200 Dual Channel
250gb SATA

We do a lot of image manipulation. Some in 2D and some in 3D. The main purpose for this machine is to power the Geowall stereoscopic projectors. The idea is that the machine outputs to two projectors; one is the left eye image and one is the right eye image. Using polarized filters and glasses (just like IMAX 3D) you see a 3D image. The problem is that for all the money that we put into this machine the 2D performance isn't as stellar as I had expected. We are manipulating images that can be up to 10,000 x 10,000 pixels in resolution (Mars surface images, as well as panoramic images from the mars rovers). When moving the image around and zooming etc... we still get a good amount of tearing in the image. Granted that these images are 150mb+ uncompressed, so its no small task even for this beast of a machine. We have tried running dual display in SLI as well as running each display off of an individual card (SLI disabled) and the results are the same.

I have decided that more investigation is needed. I have run benchmarks in (get ready to laugh!) Winbench 99 - the last free 2D benchmark software that I have found. If anybody else has other free benchmark software that can be used to test 2D - I'm all ears!

For now - download Winbench 99 (the 100mb version) and see what your card has got. I'm interested in both stock and overclocked speeds. Lets see who really wins out in the end!

Your help is greatly appreciated!

I'm going to run it again, stock and overclocked speeds and post my results to get this started.

(edit) Ok.. I only downloaded the 10mb version which only has the DirectDraw tests. The 100mb version has the High End Graphics Benchmark. It only gives you ONE number instead of 30! DUH. Ok... going to redo the results.

 

[Skaven]

Athlon 64 4000+ 4x512mb PC3200 Dual Channel

EVGA 6800Ultra 2D Clock/memory 400mhz/1.1ghz
High End Graphics Winmark 99: 2230

Overclocked to 500/1.5ghz
High End Graphics Winmark 99: 2300

 

[Skaven]

Athlon XP 2400+ 2x512 PC2700 Dual Channel

GeForce TI4400 Default Clock/Mem 275/552
High End Graphics Winmark: 1560

Overclocked 300mhz/600mhz
High End Graphics Winmark: 1560

 

[Skaven]

Ok - so I fixed the results so that they actually show the high-end graphics winmark scores. Looks like overclocking helps minimally. I'm going to try on my Radeon 9000 when I get home. I'm really interested in high-end ATI or Matrox scores if anybody is up for it!

 

[Skaven]

Athlon XP 1800+ 2x256 PC2700 Dual Channel

ATI Radeon 9000 250mhz/200mhz
High End Graphics Winmark: 1180

ATI Radeon 9000 300mhz/250mhz
High End Graphics Winmark: 1220

 

[mwmorph]

ill try it on a oced rage pro 8mb. i hope i get more than 100!

 

[Goi]

How relevant is Graphics Winmark 99 today though? It's a 6 year old benchmark afterall, and may not be that accurate in evaluating 2D performance. Besides 2D performance, do you also care about the signal quality?

 

[NokiaDude]

Are you getting real image tearing or is it stuttering? I know in games, to fix image tearing, I enable vertical sync and it goes away, although my FPS, it doesn't really matter as I get 60FPS with max IQ.

 

[Skaven]

Yea.. it may not be very relevant today... but at least you get some kind of comparison between the cards. Image quality is also important... but I want to get an idea of raw 2D performance. I guess this would be akin to looking at the original 3DMark benchmark scores. Sure it may not be testing Directx 9 shaders etc... but will give you an idea of the raw 3D performance.

Could that much have changed in the 2D acceleration world since 1999 anyway??

Anybody have a top of the line ATI card they would be willing to test?


As for the tearing, yes, there are actually tears in the image while moving it around. I believe that v-sync is only for 3D apps...

Here is a link to a sample image that tends to show tearing while zooming in and panning with the wallview stereoscopic viewing program.

 

[TGS]

(Honest questions)

Originally posted by: Skaven
Granted that these images are 150mb+ uncompressed, so its no small task even for this beast of a machine.

Someone, correct me if I'm wrong here, but isn't 2-D imagery loaded into whatever program is being used to view it. Along with the 2-d quality being a by product of the signal that is being output. I may be confused here, but I though the memory requirements on the 3-d portion of a card had to do with buffers, shaders, textures, and geometery data, moreso than the 2-d aspect.

As long as the signal quality is within your quality specifications, any card should be able to drive the 2-d image.

We have tried running dual display in SLI as well as running each display off of an individual card (SLI disabled) and the results are the same.

Again I could be off here, but I thought SLI only allowed one monitor to be on. While disabling it would allow for multi-monitor setup though no SLI capabilities. So technically your results should be the same, as SLI can't drive two or more monitors.

If you are getting tearing in 2-d apps perhaps whatever viewing program is becoming CPU bound. the only time I've ever experienced tearing so to speak is with a fully loaded
cpu long with manipulation in 2-d apps.

 

[biostud]

Matrox?

 

[Skaven]

Athlon 64 4000+ 4x512mb PC3200 Dual Channel

EVGA 7800 GTX 275mhz/1.2ghz
High End Graphics Winmark 99: 2470

 

[BTA]

Matrox for 2D hands down

 

[HDTVMan]

Wouldnt run on my windows XP kept complaining too many things were running. When disabling them it eventually hung the system.

Sorry cant help.

 

[Peter]

My minimal notebook here (Sempron 2600, SiS chipset w/ integrated graphics on 32 MBytes of borrowed PC2700 RAM out of 256 total) just scored

1070

on High-End Graphics Winmark, with everything running, PowerNow! enabled and 1024x768/32-bit.

1160

after a bit of cleanup in the system tray and PowerNow! set to "Desktop" (full speed all the time)

1460

with display depth reduced to 16-bit.

Absurdly fast, compared to dedicated graphics that cost more than the entire notebook?

Not really. Come to think of it, 2D performance is mostly about the bandwidth available to the CPU pushing image data into graphics memory - and shared system memory actually is the fastest path the system has.
I'd expect a properly implemented PCIE system to be considerably faster than an AGP system in this - simply because without fragile "Fast Writes" disabled, the CPU has a 250 MB/s ceiling in pushing data onto the graphics card, while I've already seen a rather miserable GF6200 consume 420 MB/s on an 8x PCIE link.

 

[JBT]

2230 with the settings in my sig.

 

[Ronin]

Downloading and about to run it now. I'll run it on a 4600+ with an x800 XL 512MB tomorrow, as well as a 2400+ with a Matrox Parhelia.

 

[Ronin]

FX55/2x6800U/1GB RAM - 2610
4600+ X2/x800 XL 512MB/2GB RAM - 1800
P4HT 3.6GHz/6800 Go/2GB RAM - 1510
AMD 2400+/Matrox Parhelia 256MB/1GB RAM - 1640

 

[Fox5]

I remember back in the day, 3dfx and Matrox used to be at the very top of these tests. Hmm, 3dfx may have been one of the last to implement a 2d/3d combo card, but at least they did it right.(it probably ate into their profits too)

BTW, what scores am I supposed to be looking at?

With a 2.3ghz athlon xp barton and an overclocked geforce 6600gt(575mhz core and 500mhz ram) I got 2010. Looks like nvidia rules 2d, probably because their cards have faster framebuffer access than ATI cards.

 

[CCityInstaller]

System Specs:
A64 3200@ 2.5ghz
1GB PC3200 @ 210(2-3-3-8)
ATI AIW PRO 9800 (411/364)
High End Graphics score was 2250