tomshardware review online!

[thop]

i just checked tecchannel.de they announced a review for tommorow. their reviews are usually very good and unbiased

[CaineTanathos]

Originally posted by Indiana
@Edward: exactly what I'm thinking.

And thop: Lars always "disguises" his Bias so that at first view you could think he wrote an unbiased article.
E.g. take the R8500: first he uses an OEM clocked version in this review and of course tests it against the highest clocked GF4. Then he says that the Radeon does tricks with anisotropic filtering and thus he didn't measure it.
That's partially correct: the Radeon does have a special way to do anisotropic filtering that can give non-optimal results at a few specific angles, but you normally only see this in artificial setups and not in real games. In overall the R8500 allows for MUCH higher anisotropic settings than the GF4 giving MUCH better/sharper textures in nearly every real-world case. And this with nearly no performance-loss (<10%). The Radeon8500 even at OEM clockings simply would've crushed the GF4ti4600 at the anisotropic filtering benchies and this is something that can't be (a NVidia card not winning) so he left these benches out.
This is with the Radeon8500 where I know what's going on and how fake those benchies are. Most likely he did similar things with the test-structure to make the Parhelia look worse.

P.S.: I'm sure if NVidia had integrated the Radeons anisotropic filtering method they would praise it over anything other, giving FREE anisotropic filtering.
And if he's going to say "it's doing tricks, leave it out", OK. But then he must be consequent about this. Why does he include the Quincunx which really looks shit (esp. without anisotropic)? Yes, because it makes the FSAA on the GF appear fast even though it's doing major tricks that really deteriorate image quality

Yep.
Even on the first page (I printed , just before the th.de removed the paged) there is an error :
in the section Matrox Parhelia - the card in detail
he wrote that the GPU clock is 220 but hist storing ats is 250 (instead of 275 or the gpu clock is wrong)

but like indiana sais that th.de has an bulk version and not a retail

What kind of hardware reviewing website makes such errors

[maharajah]

THG is a 1 product company; I think every geek knows that.
Hard is a one benchmark company. FPS only.
Anand seems to realize that product features and IQ are important. I'll be reading their review tomorrow.
But, more importantly, I'll be waiting for the BBs reviews tomorrow. Their word holds a little more credibility with me.
Sweet dreams tonight

[CaineTanathos]

eheh, you to

and I think the drivers of parhelia need some more tweaking

one thing I am wondering about the 3d mark vertex shader
does it use all 4 vertex shaders of the parhelia

[Reckless]

All this talk about Tom's being biased... Well maybe the site does have a bias but wouldn't the benchmark graphs/values be the same anyway? They're not going to artificially fudge those .... OK, they may choose certain benchmarks that show problem areas but hey that's not biased either - consider that I MAY WANT to run whatever they're showing? In a review I like to see what doesn't work as well as what does. If I wanted a fully rosey picture I'd ask Matrox...

The text of an article can be 100% biased but who the hell is realling reading that level of stuff now. We've seen the chip specs, seen the product specs and now all most really want is confirmation that in the real world (games that we all play) the card performs. Whatever any of the beta testers who frequent this board have to say WILL be biased towards the card. Why should that change anyones mind either way?

Reading the posts it would seem that the majority are dissapointed and if it's down to bad drivers then I'd have to say that's a terrible situation for Matrox to be in. My (the) G200 and G400 had exactly the same issues on release (god help them if this continues!). If it isn't drivers then maybe it's a games 'reliance' on expected hardware (ATI/nVidia?) and if so Matrox need to step up the developer relations effort to ensure this is dealt with. Either way if the right numbers are not quoted then lots of people will vote with their wallets FAA 'for free' or not.

And ANT - did I read this correctly - you DO NOT have a Parhelia? That would be a sin of immense proportions. Your review judgement has (to me) always been fairly sensible since I've been coming here (many years!)...

[Helevitia]

Originally posted by Rags
I don't get it.

You see very playable framerates in the lower detail benchs, but the part that matters with today's cards is the quad texturing performance and how well it will handle the details maxed up.

Parhelia CRUSHES Ti4600 in the benchmarks that measure this. Look at the multitexture fillrate, and you get the idea of how much faster than an R8500 this card is in games that use quad texturing. Turn FAA and Aniso on, and you are still left with VERY playable framerates, whereas you don't with the 8500 or the Ti4600.

Beware these drivers are fairly crude IMHO and the performance is only going to get higher, that you can count on. 8500 and GF4 are about as mature as they are going to get.

Rags

If you are referring to the 3Dmark2k1 benchmark with multitexture, I can't really put any weight on that since most of us would agree that 3dmark scores do not translate into real world numbers.
As far as "VERY playable framerates" are concerned, I would be most interested to see what the low framerates are for all 3 cards with aniso & AA(FAA vs FSAA). If the P drops to below 30 frames, I would not call that "VERY playable framerates". And what do you mean exactly by saying "whereas you don't with the 8500 or the Ti4600."? I see slightly less numbers with the 4600 in those scenarios with Quake3. Rags, I'm not trying to single you out, but your opinion, so far on this thread, is the most valuable to me since you have all 3 cards and I see you as trying to be unbiased.
Also, do you agree the aniso is not as good as the 4600? And the most important questions to me are, how does this card play on UT2k3 performance test? It will be interesting to see how P performs with next gen games. Do you think it will surpass the 4600 on next games? What about iamge quality, I know it's a stupid question, but how does everything look in windows? Text? How much better(or worse) does it look than a G400? G450? G550? Have you tried 1600x1200x32 with aniso and FAA on 3 monitors? Is the game still playable? It seems that with the current benchmarks, games will nto be playable wiht 3 monitors? Please educate us BBs if you can.

Dave

[MrGaribaldi]

Reckless> Well, Tom has managed to score lower when he tests a card than other sites... Check is V5 review vs Anands or another site...

[Liquid Snake]

I hope at least one site tomorrow will post some benches with low, average and max FPS, not just the average. An average by itself can be pretty meaningless.

[Indiana]

As already said THG seem to be doing LOTS of test and then afterwards throwing those resolutions/settings away where the NVidia-card does not win and only use the others for the review.
Ever seen a GF4 vs. Radeon8500 anisotropic filtering benchmark on THG? I guess not!

And if a non-NVidia card is as fast or faster (like the Parhelia on the FAA&aniso settings) then this fact will be disguised by overloading the graph with other NVidia results with lower quality but higher fps.

The same pattern can be seen with CPUs: Tom managed to get better performance out of an P4 than an AthlonXP with DVDx/DivX5. Hmm strange, here my AthlonXP1800+ beats my P4@2GHz?
Maybe this has something to do with the fact that I use high quality settings and THG use settings that give bad quality but favor the P4?

[Reckless]

Originally posted by MrGaribaldi
Reckless> Well, Tom has managed to score lower when he tests a card than other sites... Check is V5 review vs Anands or another site...

Well, I just had a quick look and the graphs I seen show various levels of FPS. The problem with comparing the two is that Anandtech used an Athlon 750 whilst Tom's site used a P3-1Ghz. (In fact on those graphs the Tom's values were higher).

When doing cross-site comparisons a whole bunch of factors need to be considered. Mobo, CPU, type of RAM, speed of FSB, driver versions. If the two systems are different in any way the values are not comparable.

[Joel]

Where are they by the way? I think they can say something now after the first review and the date being 25th.

We're here. I personally can't comment on a review that I have not read yet. I know nothing about the setup he used or the apps he benched and there settings. And besides it's still the 24th where I am.

Joel

[Cobos]

Just to answer the question to me...I'm currently running a Dualhead solution with 2 17'' nothing spectacular but decent, and I've got another 17''. But I need a bigger apartment first, and unfortunately the rent prices in Oslo are bad enough that I can't use triplehead. The point I was trynig to make was that Aquanox has special OpenGL or Direct3D (can't remember) call's that only a Geforce can use, thus it's a really really unfair benchmark. About as fair as running a triplehead test, though the triplehead test is a bit more obvious.... Which just shows that Lars's benchmarking conditions are "interesting". And as a scientist I find that annoying....

Cobos
BTW: Hello Mr Garibaldi

[YngwieM]

I think it's too early to us comment on those numbers...

Soon there'll be lot's of reviews, and hopefully some of them will add some good answers... (like how the P will maintain good playabillity with FAA+aniso thru the games to come)

P.S.- Guru, your t-shirt rules...

[Wienerschnitzl]

At www.chip.de there is "another" Review, they gave the Parhelia a 1.0 !!!

Well, again it is in german, let me try to translate :
"faster and more beautiful than a GF4"


more direct link:
Comback of the Year : Matrox Parhelia 512

[thop]

thank you wienerschnitzl! nice nick btw.