2D, Acceleration, And Windows: Aren't All Graphics Cards Equal?

[knutjb]

Interesting story. I have an XP machine with an X800 , a Vista machine with a X1900, and a Win7 machine with a 5770. I had all the machines hooked up to the same set of monitors for some work after moving. The Win7/5770 looked stunning compared to the other two which weren't that far apart in general appearance. This was with aero enabled and disabled. Either way Win7 produced the best looking 2D view.

 

[ktasley]

Nice job Tom's/

2D matters and it looks like some people at some companies forgot that...

 

[kvik86]

Hi!

Great article! I'd like to ask where is WPF in this case (different OS-es, performance compared to GDI)? Is it also affected? Is it faster or slower? I'm really interested!

Thanks!
KV

 

[knowom]

I think if anything this shows both Nvidia and Ati have lost focus in 2D in recent years at least for consumer cards not too sure about their workstation cards.

 

[JohnnyLucky]

Great article! I actually understood most of it. I think I'm going to tinker with my new system and see what happens.

 

[michaelahess]

Windows ME was not polarizing, it was hated by all!

 

[Guest]

The most telling sentece in the article, and this directly addresses a current debate in the Graphics Forum.
GtX 285 vs 5870........
The GeForce GTX 285 runs as much as 11 times faster than ATI's Radeon HD 5870. Worse yet, an on-board graphics processor from a two year-old $50 motherboard bests this $400 graphics card by an order of magnitude.

 

[aneasytarget]

Great article, Igor Wallossek. This is why I read THG, solid investigative analysis of a problem, dissect the problem, work with the manufacture to address the issues.

Looking forward to the 2nd part.

More articles like this please.

 

[TheGreatGrapeApe]

If you're going to review with S3 and older hardware, then MATROX is a MUST !!

They have current hardware as well as older hardware that has always been the 2D standard.

Also including a GMA series card would be a good idea as they have always outpaced the other two (ATi/nV) in 2D speed, and usually by a wide margin (similar to that of the GF7050). Speaking of which it's interesting that the focus was limited to the ATi solution, which was definitely poor, however in many of the other 2D tests (like the Polygon and Arc tests) the GTX285 was also left in the dirt by the integrated solution. It beat the ATi solutions, there is no doub,t but it's still performing at 1/3 of the speed of the GF 7050.

Seems to me we need more options like the intel GMA series, Matrox and S3 to see if 'better than the other guy' performance really is 'good enough' for any of them.

 

[wh3resmycar]

shouldnt an intel igp be included in those benchmarks??

 

[Guest]

gj tom's... if you could, plz make 2d a regular part of your benchmarks... thnx!!!

 

[TheGreatGrapeApe]

Why was the benchmark download not included in the English version of the article, but included in the German one? It's interesting seeing the variety of tests from other out there.

Just ran it myself straight on the laptop (not the connected PCI Matrox G550) thus using just the onboard GMA X3100 and got the following;

BENCHMARK: DIRECT DRAWING TO VISIBLE DEVICE

Text: 5949 chars/sec
Line: 31901 lines/sec
Polygon: 5656 polygons/sec
Rectangle: 1232 rects/sec
Arc/Ellipse: 8889 ellipses/sec
Blitting: 2617 operations/sec
Stretching: 331 operations/sec
Splines/Bézier: 18182 splines/sec
Score: 767

The benchmark can be found here;
http://www.tomshardware.com/de/download/Tom2D,0301-26150.html

Also for those interested, part 2 is already done, and unfortunately no S3 or Matrox cards;

http://www.tomshardware.com/de/wddm-2d-performance,testberichte-240487.html

 

[scotth108]

Thanks very much for this article. I don't game, and neither do my two main clients; but they and I all push on 2D pretty heavily at times. Nice to have the info-gap filled!!

 

[Guest]

check out the related article "Preview of 3Dfx Voodoo Banshee, S3 Savage3D and NVIDIA RIVA TNT"

One of the first paragraphs emphasises the acceleration of gdi functions in hardware!

I loved my banshee card, i'm sure its in my component "memories" box full of stuff. From what i remember, it was massive.

I have to say, when i started looking at the results, my first thought was catalyst was the culprit.

Looks like lazyness on ati's part, can't blame them though, before this artical came out, we've been buying 3d cards blindly for a good 10 years without thinking about 2d performance.

I would like to see benchmarks for the quadro and firegl lines as a comparison. I imagine the results will be night and day between the cad based cards and consumer cards.


P.s. good work guys, this is really something!

 

[ravewulf]

@Belardo

GDI is not part of DirectX

 

[cangelini]

[citation][nom]TheGreatGrapeApe[/nom]Why was the benchmark download not included in the English version of the article, but included in the German one? It's interesting seeing the variety of tests from other out there.Just ran it myself straight on the laptop (not the connected PCI Matrox G550) thus using just the onboard GMA X3100 and got the following;BENCHMARK: DIRECT DRAWING TO VISIBLE DEVICEText: 5949 chars/secLine: 31901 lines/secPolygon: 5656 polygons/secRectangle: 1232 rects/secArc/Ellipse: 8889 ellipses/secBlitting: 2617 operations/secStretching: 331 operations/secSplines/Bézier: 18182 splines/secScore: 767 The benchmark can be found here;http://www.tomshardware.com/de/dow [...] 26150.htmlAlso for those interested, part 2 is already done, and unfortunately no S3 or Matrox cards;http://www.tomshardware.com/de/wdd [...] 40487.html[/citation]

Benchmark was included in Part 2, which is still in translation in the US. I'll pass along the feedback re: those legacy cards to see if the German team has them available!

 

[gwolfman]

What about Intel graphics? Does the Quadro and FireGL series suffer from these issues as well!?! I'd hope not!

I agree with former posts about possible skewing of acceleration to keep idle temps and clocks down.

 

[earthwormsvx]

Now where did I put my copy of Winbench? It must have 2 inches of dust on it by now. Winbench was a 2D benchmark tool from the early 90's.

 

[uwave101]

My Matrox G550 on a 3520x1200 display (dual) with a 2.8GHz P4 in XP gave me a total score of 441. An 2GHz Core 2 duo (essentially a laptop) with an Intel GM/GME chipset gives a total score of 1169.

 

[climber]

I think this might be a result of Microsoft trying to include in Windows 7 the ability for people to split the tasks of 3D work and 2D work between onboard and discrete graphics chips. If one was to make this assumption, there is an error in this approach, that being, people who want serious gamming or workstation class motherboards, don't want the onboard graphics taking up the room on the board or any of the PCI-express or remnant PCI-X or PCI lanes with that low end stuff. It might be great for notebook battery life, but not for serious power workstation that are doing serious 2D drawing or CAD work. I do lots of Corel DRAW and Canvis work, where I have hundreds of thousands of polygons and don't need slow downs in performance due to this new "feature" in windows. I also do GIS which to a large extent is 2D points, lines and polygons.

 

[jes_pt]

Excellent article.

 

[climber]

[citation][nom]pcwlai[/nom]Without this article, we may need to accept slow 2D performance in the new product generation...Well, I would like to address why programmers not trying to minimize the redraw of windows which is actually encouraged by Microsoft as well.Imagine you need to update a rectangle in the center of a window while some other window is overlapping portion of it. If is already not easy and time consuming to compute the lines remains to form the visible portion.If there is some other complex drawings... it is just too time consumign and sometimes impossible to get that uncovered part calculated.So, we programmers rely on the hardware acceleration. Most of the time, it is actually to redraw everything especially when 2D hardware acceleration is present.[/citation]

I think it's lazy programming to simply say, redraw everything, especially when there's a good possibility that only a tiny portion of the screen has changed. I've worked in GIS for years, MapInfo only redrew what was in the minimum bounding rectangle of the object, text, whatever was modified, ArcGIS redraws everything in every layer, one by one. I've had situations which too minutes to redraw the screen because of this, lazy, legacy programming model, don't fix or optimize, just patch and build around your old outdated code.

 

[Guest]

This is a shockingly great article. It also applies to more cards than what they tested here. For example in 2D/2.5D, an ORIGINAL Radeon is faster than a Geforce 6600GT. This must be an ongoing issue that no graphics maker has ever tried to repair until recently.

 

[tsnor]

Outstanding article. Bet the author was freaking at the first measurements and grinning at the AMD/ATI "We now have our driver team engaged to optimize this path and will release a new driver to address this workload as soon as possible."

 

[Guest]

I don't understand the real point of this. AutoCAD users (let alone most every other user) don't go out and buy a gaming graphics card to accomplish what they want... They buy a workstation graphics card. This article seems like a joke.

The people who are using the apps mentioned here are probably most likely using either onboard graphics because they don't want to afford the expensive hardware or they work for a company that goes all out and buys workstation graphics cards.