Tom's Hardware guide: Matrox G450

[Kruzin]

So you dont think it's a valid point that they are in fact using an SDR, and passing it off as a DDR in the benchmarks?
Most of here know what a huge difference this one simple point mskes in the benchmarks, when run at the resolutions they ran the tests at. The bandwidth required at these resolutions chokes the SDR version down, while the DDR version continues to perform at a significantly higher speed of the SDR.

Here is a thread that shows what I mean.
http://forums.murc.ws/ubb/Forum5/HTML/009935.html
Scroll down to Ashleys scores on both types of card:
"G450 16MB DDR SDRAM @ 166MHz = 2551
G450 16MB SDR SDRAM @ 166MHz = 1702"

[Brian R.]

Tom is responsible for everything published under his name. The editor gets the heat in a newspaper if he allows falsehoods to be printed. The newspaper owner is responsible for answering lawsuits placed against published material. Tom is both on that site. It is editorial policy (set by Tom) which dictates the tone and accuracy of every article published on his site.

Kruzin - To give them on Tom's site the benefit of the doubt, it might just be incompetence and not malice influencing their accuracy. Neither is applaudable, however.

[This message has been edited by Brian R. (edited 09 December 2000).]

[Joel]

Now here's a fairly decent review of the G450 looking at it based on it's intended market.

www.tweak3d.net/reviews/matrox/millg450/

Joel

[Greebe]

rubank,

Tom is the publisher not the author

And being such has a responcebility of ensuring what is written is accurate... ie he should proof read it first. If he had done either then we wouldn't have any problems with his publishing/dribble.

and of course you should do the same too! D'oh!

[Indiana]

Originally posted by Kruzin:
So you dont think it's a valid point that they are in fact using an SDR, and passing it off as a DDR in the benchmarks?

This is just plain wrong. From the performance numbers you can clearly see that they actually must have tested the DDR G450. There's no way a G450SDR will get 16-17 fps in QIII 1280x1240x32 HQ..., all the other results are very indicating to the DDR version as well since they're slightly below the scores of a G400.
But I forgot: a good review is only one where 1. Matrox is coming out on top and 2. you MUST NOT mention NVIDIA, ATI or 3DFx and 3. if you do speak of the aforementioned you have to say that they're crap and the only thing they're giving you are fps in a range that noone needs....

And: a review SHOULD come to a conclusion (that's, of course, ALWAYS opinion-based). Of course you should do the review first and then get the opinion and not vice versa...

One last thing: I would see the oc thing as an disadvantage as well. Maybe Tom's (the reviewers) english is not that good but as I read him he's complaining not sppcificly that MGATweak is not able to oc the G450 but that there's currently no way to oc this card AT ALL, including the MGATweak tool, which is known as the best tool for oc'ing cards of the G400 line (and you can count the G450 in there).
This is a disadvantage as it was claimed (and seems logical) e.g. here on this forum that the G450 should in fact be quite oc'able due to it running at the same core-speed in spite of it's manufacturing in a lower micron process.

[Rags]

Indiana,

you are so far removed from reality, it's not funny.

I have a G450 DDR and it scores double that at 1024.

He didn't say that the G450 has to come out on top, he was just being critical of ambiguous information, and it is.

Rags

[Greebe]

Indiana, all of us Matrox beta testers (not just a single BB) can confirm just what Rags said. His review is BS, full of errors and based on misleading half truths!

OK

[Rags]

I'm not saying it was intentional or whatever, but what is written is misleading.

Rags

[Joel]

With my G450 DDR I get about the same scores as Tom did running Quake III using the same drivers. The only real differences between his test setup and mine is that I intend to run alot of stuff in background, I leave access to the CD on, I leave sound on (to get a real world performance look, I don't know too many people that play without sound), and my system is an 800MHz AMD Thunderbird whereas his was an Intel Pentium III 1GHz. So I think that kind of shows that he was in fact using the SDR version.

Joel

[Indiana]

Originally posted by Rags:
Indiana,


you are so far removed from reality, it's not funny.


I have a G450 DDR and it scores double that at 1024.

What exactly does it score and where? I never spoke of 1024 scores, BTW...
A G450DDR should score roughly below 30fps in QIII 1024x768x32 HQ (and this is with trilinear ON!) but it won't reach 20fps in 1280x1024x32 HQ.
These tests in this res/depth are severely bandwidth limited even on the G400/G450DDR, the SDR version WILL get much lower scores (probably not much more than the half) than those on the review.

But maybe someone on this list with a G450DDR (Maggi?) can run Q-III and MBTR in the tested resolutions and post his numbers for comparision.

[Indiana]

Originally posted by Joel:
With my G450 DDR I get about the same scores as Tom did running Quake III using the same drivers. The only real differences between his test setup and mine is that I intend to run alot of stuff in background, I leave access to the CD on, I leave sound on (to get a real world performance look, I don't know too many people that play without sound), and my system is an 800MHz AMD Thunderbird whereas his was an Intel Pentium III 1GHz. So I think that kind of shows that he was in fact using the SDR version.

Joel

No, this again shows that he was using the DDR version. All the things you mention (running other software, having sound enabled,...) only cost CPU-time, not gfx-mem bandwidth - and the latter is the one and only limiting factor for Q-III performance in the resolutions Tom tested, CPU power is irrelevant as long as you have a 500MHz+ CPU.
This, btw, would be a much more valid complaint about Tom's test: why didn't he test 800x600x16 or turn off trilinear (a real performance-killer on the G4x0 boards that only gives a very slight quality improvement)?


[This message has been edited by Indiana (edited 09 December 2000).]

[Joel]

CPU power is irrelevant as long as you have a 500MHz+ CPU.

I guess that is why I get much better scores on my 800MHz machine vs. my 500MHz machine.

Joel

[This message has been edited by Joel (edited 09 December 2000).]

[Rags]

And that's why my gameplay is actually smoother when I clock from 550 to 850, right?

Rags

(same cpu, btw)

[Indiana]

Originally posted by Joel:
I guess that is why I get much better scores on my 800MHz machine vs. my 500MHz machine.

Much better scores... what game, what resolution?
If you're saying that you get much higher scores for Q-III in 1280x1280x32 HQ with a P-III (or Duron/TBird) 800MHz than with a 500MHz one, I don't believe you, sorry.

Originally posted by Rags:
And that's why my gameplay is actually smoother when I clock from 550 to 850, right?


Rags


(same cpu, btw)

We were not talking about general gameplay (which is of course smoother with a faster CPU in many cases), but specificly about the Q-III demo001 in 1280x124x32 with the HQ settings otherwise. And in this case the CPU-speed does NOT significantly influence the results when a P-III/Athlon/TBird/Duron with 500+MHz is used (again, noone talks about anything being smoother, just about the resulting average benchmark-result).


[This message has been edited by Indiana (edited 09 December 2000).]

[rubank]

Kruzin,
come on, you know better than this!
To validate your point you refer to Ashleys measly scores (for Gods sake, his G450 SDR score is in the range of my old G200) in a benchmark THG didn´t use. What does that prove? Nada.
When I had my G400, clocked at MAX speed, I got numbers in MBTR demo just in excess of THG:s G450 results. That should validate the G450 as being of the DDR persuasion, at least in that test.

Other than that I think Indiana has made some good points.

Brian R, if you think you can bring Tom down on a falsehood charge, by all means sue him.

rubank

[This message has been edited by rubank (edited 10 December 2000).]