Typedef Enum
Regular
So...what are you trying to say?
Ich Kann nicht Deutsche verstehen
- Thread starter ben6
- Start date
So...what are you trying to say?
Jerry Cornelius
Newcomer
This reminds me a lot of TBR vs IMR discussions. They've obviously worked to minimize the worst case scenarios. Benchmarks will never show it, since the card that get's 200+ fps in that flat spot in the middle of the bench looking at a blank wall takes all the marbles.
I'm likely to bet my $500+ CAN it's going to have excellent visual quality, useable AA and less jarring slowdowns than other cards on the market.
I really couldn't care less about top end performance. I'm interested in how low it bottoms out and how often. Judging from the specs this card looks promising.
They may end up snagging some IMG customers that are sick of waiting for Series 4.
I'm likely to bet my $500+ CAN it's going to have excellent visual quality, useable AA and less jarring slowdowns than other cards on the market.
I really couldn't care less about top end performance. I'm interested in how low it bottoms out and how often. Judging from the specs this card looks promising.
They may end up snagging some IMG customers that are sick of waiting for Series 4.
Hellbinder
Banned
Nappe1,
I am humled and honored.
Seriously, You all know that even though they did not show it the 8500 would whip its ass in aniso. Yes they are different methods. However look at the facts. The 8500 beats it in 90% of the tests at least. With faster Aniso that is at least 80% of the quality. On top of that the Gf4 straight FSAA kicks its ass to. With the release of the next dets for GF4 Aniso is expected to increase by as much as 30%. Parhelia simply cant compete with that.
Here is the problem. Who flipping cares if you are the fastest with 4x FSAA and Aniso if its only 40 FPS on an average non taxing demo? It will be very hard to use this card in any serious Gaming action.
What if you want to take the brakes of and get into a serious FPS battle? You are screwed with this card. It starts out to low.
There is just NO justification for a 399$ price tag. The Radeon 8500 is faster in 90% of all benchmarks. Its Aniso is faster also. for 250$ this would have been an interesting card. but not at 399$. Especially when you consider that the R300 and Rv250 are mere weeks away for at least 100$ less and offer huge ammounts of performance.
I am humled and honored.
Seriously, You all know that even though they did not show it the 8500 would whip its ass in aniso. Yes they are different methods. However look at the facts. The 8500 beats it in 90% of the tests at least. With faster Aniso that is at least 80% of the quality. On top of that the Gf4 straight FSAA kicks its ass to. With the release of the next dets for GF4 Aniso is expected to increase by as much as 30%. Parhelia simply cant compete with that.
Here is the problem. Who flipping cares if you are the fastest with 4x FSAA and Aniso if its only 40 FPS on an average non taxing demo? It will be very hard to use this card in any serious Gaming action.
What if you want to take the brakes of and get into a serious FPS battle? You are screwed with this card. It starts out to low.
There is just NO justification for a 399$ price tag. The Radeon 8500 is faster in 90% of all benchmarks. Its Aniso is faster also. for 250$ this would have been an interesting card. but not at 399$. Especially when you consider that the R300 and Rv250 are mere weeks away for at least 100$ less and offer huge ammounts of performance.
Typedef Enum
Regular
I would have liked to have said something like "Yeah, but the Parhelia's Anisotropic filtering is second to none..." or similar...
But the one thing that really was dissapointing, perhaps more than the lackluster performance, was the fairly poopy filtering that was displayed in the review.
If you were to compare it to nVidia's implementation, you're looking at something like "2x." Apparantly, this is what Tom actually said...I thought it looked like "2x," but I have read (it has obviously been yanked) that Tom actually stated this.
But, let's just assume that we're talking about a driver issue, and the card is capable of much higher levels...Then there will be the issue of performance. You _cannot_ make the argument that Parhelia is immune to performance degredation in the area of Anisotropic filtering...I did see that one page that showed the performance dropoff, and it was much more steep than expected...If Parhelia were capable of delivering something like "8x," then it seems likely the performance drop _might_ be even larger.
Of course, you could possibly chalk that up to driver immaturity...But like some of the guys @ nVidia have stated many a time, Anisotropic filtering eats up processor cycles...and the 220 MHz. clock doesn't help.
The one thing you can say right now is that the FAA implementation looks very good.
But the one thing that really was dissapointing, perhaps more than the lackluster performance, was the fairly poopy filtering that was displayed in the review.
If you were to compare it to nVidia's implementation, you're looking at something like "2x." Apparantly, this is what Tom actually said...I thought it looked like "2x," but I have read (it has obviously been yanked) that Tom actually stated this.
But, let's just assume that we're talking about a driver issue, and the card is capable of much higher levels...Then there will be the issue of performance. You _cannot_ make the argument that Parhelia is immune to performance degredation in the area of Anisotropic filtering...I did see that one page that showed the performance dropoff, and it was much more steep than expected...If Parhelia were capable of delivering something like "8x," then it seems likely the performance drop _might_ be even larger.
Of course, you could possibly chalk that up to driver immaturity...But like some of the guys @ nVidia have stated many a time, Anisotropic filtering eats up processor cycles...and the 220 MHz. clock doesn't help.
The one thing you can say right now is that the FAA implementation looks very good.
Heres another review of it http://www.chip.de/produkte_tests/produkte_tests_8737618.html
Hellbinder
Banned
no way jazz, It is way more than 50-60%.... please...
it was less that 60 FPS in about 15... Think about it....... They also did not show the Radeeons Aniso scores in comparrison.
So first, I totally disagree with your 50-60% adn secondly with aniso enabled the 8500 roasts it in *every* benchamrk.
No big deal except this was the ENTIRE case for the cards superiority.
it was less that 60 FPS in about 15... Think about it....... They also did not show the Radeeons Aniso scores in comparrison.
So first, I totally disagree with your 50-60% adn secondly with aniso enabled the 8500 roasts it in *every* benchamrk.
No big deal except this was the ENTIRE case for the cards superiority.
LittlePenny
Regular
Jazz said:
Interesting on the 3dMark Multitexturing and FSAA test they had.
Ok i had the benchmarks saved on my hard disk and i counted up how many time the parhelia beats the r8500, out of the 30benchmarks i saw the parhelia was better in 13, that means the ati card was better in 57% of the benchies, and i can't believe i actually counted them up
Also the chip.de portrays this card alot better than tomshardware have, its worth a read
Also the chip.de portrays this card alot better than tomshardware have, its worth a read
Typedef Enum
Regular
Matrox is going to have to play nVidia's game. Obviously, I'm talking about the GF4 strengths over the GF3: better performance w/ Antialiasing @ higher resolutions.
If you disregard non AA/AF performance numbers altogether, and strictly focus on AA/AF, then Parhelia does flex some muscles.
Of course, we're comparing this card to the Ti4600, generally regarded as the highest performance 3D board available...That will, of course, change in the very near future.
But the Chip.de article does show the Parhelia potential...It will be interesting to see how things change with a good amount of driver tuning over the next few months.
If you disregard non AA/AF performance numbers altogether, and strictly focus on AA/AF, then Parhelia does flex some muscles.
Of course, we're comparing this card to the Ti4600, generally regarded as the highest performance 3D board available...That will, of course, change in the very near future.
But the Chip.de article does show the Parhelia potential...It will be interesting to see how things change with a good amount of driver tuning over the next few months.
Matt Burris
Newcomer
The small FAA/Aniso performance hit is impressive. UNFORTUNATELY the FPS it puts out with FAA and Aniso are just borderline playable, if at all.
Most people view FAA/Aniso as extra quality, to use when you have performance to spare. With the Parhelia, not only is the performance below what I'd consider playable, but even disabling the features doesn't result in much better performance.
I really think they made a mistake here... I mean, I'd be impressed by their FAA/Aniso numbers, except I have no intention of ever running with FSAA/FAA. So I could buy this card and get aniso, but ATi/Nvidia already have that and with only aniso it seems like the Parhelia is slower.
That's not even mentioning that this is against current graphics accelerators. Within a couple of months the R300 will hit the market and then what? It will stomp the Parhelia straight into the ground. Not really a pretty picture (no pun intended).
Most people view FAA/Aniso as extra quality, to use when you have performance to spare. With the Parhelia, not only is the performance below what I'd consider playable, but even disabling the features doesn't result in much better performance.
I really think they made a mistake here... I mean, I'd be impressed by their FAA/Aniso numbers, except I have no intention of ever running with FSAA/FAA. So I could buy this card and get aniso, but ATi/Nvidia already have that and with only aniso it seems like the Parhelia is slower.
That's not even mentioning that this is against current graphics accelerators. Within a couple of months the R300 will hit the market and then what? It will stomp the Parhelia straight into the ground. Not really a pretty picture (no pun intended).
Hellbinder
Banned
potential???
That 3dchip article showe no aniso, no normal FSAA modes etc etc etc... please.. this is exactly the kind of artice that matrox wanted. IT does not show the real picture at all.
That 3dchip article showe no aniso, no normal FSAA modes etc etc etc... please.. this is exactly the kind of artice that matrox wanted. IT does not show the real picture at all.
Just as the 8500 has realized huge performance improvements with later drivers, so I expect Parhelia to quickly ramp up speed as Matrox comes to grips with it and developers target it. Something must be wrong when it's being outperformed in vertex shader tests--it has four to nVidia's two, and Parhelia's four are not half the clock speed of the 4600's pair.
Similarly, I expect aniso image quality to improve--I seriously doubt Parhelia has a bandwidth issue.
Obviously, disappointing debut benchmarks, but I'm sure Parhelia will turn out to be a worthy purchase ... just not to a hardcore gamer, who would do better to turn to an R300 or NV30.
Heck, how's it gonna come close to 27,000 3DMarks?
Similarly, I expect aniso image quality to improve--I seriously doubt Parhelia has a bandwidth issue.
Obviously, disappointing debut benchmarks, but I'm sure Parhelia will turn out to be a worthy purchase ... just not to a hardcore gamer, who would do better to turn to an R300 or NV30.
Heck, how's it gonna come close to 27,000 3DMarks?
Hellbinder
Banned
First i dont remember seeing any aniso enabled in those numbers...
Second if it did PArhelia is only doing 2x to GF4's 8x. if you look at toms numbers when both cards are doing 2x the GF4 more than beats the parhelia. No screenshots etc.. and those FPS ARE NOT ACCEPTABLE to most of teh gaming public...
I just took a look at the matrox forums. Totally rediculous. They are also trying to say that that 3dchip article is a "more fair view" how totall and comletely absurd. toms article shows 2x the information and screen shots...
This is just silly.
Second if it did PArhelia is only doing 2x to GF4's 8x. if you look at toms numbers when both cards are doing 2x the GF4 more than beats the parhelia. No screenshots etc.. and those FPS ARE NOT ACCEPTABLE to most of teh gaming public...
I just took a look at the matrox forums. Totally rediculous. They are also trying to say that that 3dchip article is a "more fair view" how totall and comletely absurd. toms article shows 2x the information and screen shots...
This is just silly.
Quite good, yes. But still far from acceptable. I guess this is exactly what they meant (matrox) when they supposedly told HardOCP that they wont get a review sample in the first place.....not the "right" kind of benchies...
THG for example (and Lars stated that he left out R8500 with AF switched on because of their "different" (actual word was afair questionable) approach to AF) showed not only raw scores, in which the Parhelia sucked badly (and still does so in the Chip.de test) but also with FSAA AND AF turned on.
Chip only tests with FSAA turned on and many people already bashed nVidia for delivering an inferior IQ with their multisampling technique compared to ATis Smoothvision and it's raised texture quality throughout the whole picture. Now, Matrox does the same thing, maybe even less, 'cause the driver has to separate whole objects to remove the jaggies from and that, as THG and Chip state, does not work out in 100% of the scenes....
Well, if you combine the two surfaced tests, you get an idea of the whole picture, imho. Matrox is fpr FSAA-Gaming only. If you want real IQ (with both FSAA and AF turned on that is) you can't use any of the exisiting cards in decent resolutions today, they're just too slow, except for older Games.
THG for example (and Lars stated that he left out R8500 with AF switched on because of their "different" (actual word was afair questionable) approach to AF) showed not only raw scores, in which the Parhelia sucked badly (and still does so in the Chip.de test) but also with FSAA AND AF turned on.
Chip only tests with FSAA turned on and many people already bashed nVidia for delivering an inferior IQ with their multisampling technique compared to ATis Smoothvision and it's raised texture quality throughout the whole picture. Now, Matrox does the same thing, maybe even less, 'cause the driver has to separate whole objects to remove the jaggies from and that, as THG and Chip state, does not work out in 100% of the scenes....
Well, if you combine the two surfaced tests, you get an idea of the whole picture, imho. Matrox is fpr FSAA-Gaming only. If you want real IQ (with both FSAA and AF turned on that is) you can't use any of the exisiting cards in decent resolutions today, they're just too slow, except for older Games.
Not a total loss
Now the question is, will it still do that with next year's games? For a $400 card, it should.
It's just too close to the hairy edge of being too slow. And by hairy edge I mean if it were to lose 30-40% of its speed in Game X next year, users will have to really start dropping detail levels and features. Most people who drop $400 on a card to play games will expect it to deliver 60-80fps with everything cranked for at least a year.
At least with that 4600, you can lower the resolution and turn AF/AA off. Doing so on the Parhelia regains very very little ground back. That is, the scaling across resolutions and detail levels seems to be poor. There are two sides to the 'performance loss' coin. On one side, the Parhelia loses very little performance by turning on the goodies. On the other side, it doesn't gain much ground turning them back off.
Perhaps the drivers will mature like Nvidia's have over the last few years, with huge improvements in low resolutions and smaller but notable improvements even with all the goodies turned on. Perhaps that will happen in enough time so our 2004 release games can still be run at 1024x768x32 at medium to low quality levels. The cold hard fact is, eventually the Parhelia won't be able to run at that super detail level for which it has been optimized with newer games, just as that shiny GF2 Ultra won't run SOF2 at 1600x1200x32 even thought it eeked by on Quake 3. But at least the GF2U is still doing well at 800x600 and 1024x768.
Not bad for Matrox's 'return-to-the-market' product. And it cannot be said that the current ATI or Nvidia products beat it across the board. At this point I think I'm fairly confident in saying their anti-aliasing method beats out both ATI and Nvidia in terms of quality and speed. ATI's is just too slow by the time it starts looking good (2x or 3x quality). Nvidia's is damn fast, but isn't tops on quality. It is just going to be very important for Matrox to keep artifacts to an absolute minimum.
Now the sadest part is the R300 and NV30 should pretty much smear the Parhelia in another few months (can you say '256bit bus?'). Makes dishing out $400 right now a bit difficult. These last two months may have really hurt Matrox in terms of timing.
A bit off topic, but really, what is Matrox's marketing angle with this? It's not really a workstation card, and it's not really a very well proportioned card for the enthusiast either without being able to bust out impressive sounding (however worthless they may be at times...) numbers. And even the enthusiast market isn't a very profitable one with such a small market and high competition. Can Matrox make money with this thing? Here's hoping for a Parhelia 2--I think they could use the chance to catchup.
muted said:
Now the question is, will it still do that with next year's games? For a $400 card, it should.
It's just too close to the hairy edge of being too slow. And by hairy edge I mean if it were to lose 30-40% of its speed in Game X next year, users will have to really start dropping detail levels and features. Most people who drop $400 on a card to play games will expect it to deliver 60-80fps with everything cranked for at least a year.
At least with that 4600, you can lower the resolution and turn AF/AA off. Doing so on the Parhelia regains very very little ground back. That is, the scaling across resolutions and detail levels seems to be poor. There are two sides to the 'performance loss' coin. On one side, the Parhelia loses very little performance by turning on the goodies. On the other side, it doesn't gain much ground turning them back off.
Perhaps the drivers will mature like Nvidia's have over the last few years, with huge improvements in low resolutions and smaller but notable improvements even with all the goodies turned on. Perhaps that will happen in enough time so our 2004 release games can still be run at 1024x768x32 at medium to low quality levels. The cold hard fact is, eventually the Parhelia won't be able to run at that super detail level for which it has been optimized with newer games, just as that shiny GF2 Ultra won't run SOF2 at 1600x1200x32 even thought it eeked by on Quake 3. But at least the GF2U is still doing well at 800x600 and 1024x768.
Not bad for Matrox's 'return-to-the-market' product. And it cannot be said that the current ATI or Nvidia products beat it across the board. At this point I think I'm fairly confident in saying their anti-aliasing method beats out both ATI and Nvidia in terms of quality and speed. ATI's is just too slow by the time it starts looking good (2x or 3x quality). Nvidia's is damn fast, but isn't tops on quality. It is just going to be very important for Matrox to keep artifacts to an absolute minimum.
Now the sadest part is the R300 and NV30 should pretty much smear the Parhelia in another few months (can you say '256bit bus?'). Makes dishing out $400 right now a bit difficult. These last two months may have really hurt Matrox in terms of timing.
A bit off topic, but really, what is Matrox's marketing angle with this? It's not really a workstation card, and it's not really a very well proportioned card for the enthusiast either without being able to bust out impressive sounding (however worthless they may be at times...) numbers. And even the enthusiast market isn't a very profitable one with such a small market and high competition. Can Matrox make money with this thing? Here's hoping for a Parhelia 2--I think they could use the chance to catchup.
