PDA

View Full Version : Something's fishy...


Pages : [1] 2

Cota
05-04-04, 03:10 PM
After reading several reviews I've found a few things that catch my attention.


But the most interesting one is that on some reviews like hard ocp the X800pro is almost as fast as the X800XT and even faster than the 6800ultra but on others like extremetech it looses in almost every bench with the 6800GT.

Could it be that some sites received lower clocked X800xt instead of real X800 pro? It would make sense since the performance difference in the hardocp review seem more because of clockspeed rather than pipelines.

Of course Nvidia did their homework and released the 6800ultra Extreme Edition (450/1100) which may be replaced by the 6800 super duper mega hyper ultra XTGT+4 Extreme extended remixed Director Cut Final Edition (600/1800??) and new optimized (read "full of bugs") drivers to steal ATIs thunder.

Now more than ever this is as close a race as it can be. Nvidia has more features that have yet to be seen like SM3.0 and hardware video codecs. ATI is faster or slower depending what review you read but has better image quality and temporal AA.

I think I'll just flip a coin to help me decide.

particleman
05-04-04, 03:39 PM
Extremetech's numbers are definitely extremely messed up, that or everyone else's numbers are messed up. Extremetech has the 6800 Ultra performing at the same speed as a X800XT in FarCry, this does not match up with any other review out there.

As for HardOCP you need to be a bit careful because HardOCP uses a system where they bench at the best image quality at a playable framerate for each card. The X800XT is faster than the Pro, it's just that HardOCP benches the Pro at a lower resolution than the XT.

jimmyjames123
05-04-04, 03:42 PM
I don't think that is the case of different reviewers sneakily getting cards with different clocks. Many websites have their own custom demos, they have unique PC equipment, they have different ways of benchmarking cards.

You just have to understand what each reviewer is doing. Some reviewers test mainly at 4xAA/8xAF, at various resolutions. Some test up to max AA, whether that be 6xAA or 8xAA (which obviously makes the 8xAA card look very bad in comparison). Some test up to playable settings (which again tends to make the 6xAA card look more favorable because it incurs much less of a performance penalty than 8xAA). Some use only one or two resolution settings (which doesn't show the impact of any higher resolutions).

No review is perfect, but the better reviews tend to show the impact of various resolutions and various AA/AF settings on the cards performance. Personally, I like TechReport's graphical layout, as it is easy to see how performance changes with resolution or with AA/AF settings.

eesa
05-04-04, 03:45 PM
it's very easy to make on card look a lot better than another depending on which tests you choose. No one can say one card is definitively better than the other since it all depends on what resolutions/settings/games you'll be running. Performance aside, the power consumption and dual slots alone turns me off the gf6. But that's just me.

Casper
05-04-04, 03:47 PM
From what I have noticed the faster CPU the more it seem to benefit the Radeons for whatever reason.

eesa
05-04-04, 03:51 PM
Oh yes, and I as well am puzzled over extremetech's results.

Anandtech's was extremely neutral.
Tom's tested really useless settings.
Hardocp, though they didn't lie, really catered the tests to make ati look good (tested max playable settings)

I read more, but don't have time to go into each one. Overall though, I'm just surprised ppl didn't kick up a big fuss about how much more elegant a solution the x800 is if you completely ignore performance(cooler, less power). Maybe it just doesn't matter that much to other people? I know the OEMs would sure care a lot...
And if you got to hold the card, you'll realize how small it is. Almost cute if you ask me :) 6800 is a beast. But again, not everyone cares about that.

jimmyjames123
05-04-04, 03:51 PM
A faster CPU probably benefits the X800 XT PE edition most, because it has massive fillrate. In general, all of these new gpu's benefit very much from faster cpu's.

zakelwe
05-04-04, 03:54 PM
Anand did the worst review I have ever seen them do, it looks like they were very short of time .. just threw it at the page

Hexus review as well was rubbish, though that is also probably due to time limit.

Xbitlabs is also split into two sections .. why ? Time limits ?

These cards are geting more and more complicated and the reviewers are getting les and less time to review them in.

Need to wait for the dust to settle me thinks

Regards

Andy

jimmyjames123
05-04-04, 03:55 PM
Overall though, I'm just surprised ppl didn't kick up a big fuss about how much more elegant a solution the x800 is if you completely ignore performance(cooler, less power). Maybe it just doesn't matter that much to other people? I know the OEMs would sure care a lot...

People did take note of that, and that is definitely a point in it's favor. At the same time, the 6800 GT seems to be a 16 pipeline single slot/single molex design that doesn't eat up nearly as much power as, say, a 6800UE. In other words, very OEM friendly. Also, the NV4x cards were really not noted as being excessively loud, excessively warm to the touch, or excessively power hungry. It is really only the 6800 Ultra Extreme Edition that requires the very beefy power supply. But again, very nice and elegant design from ATI, and most reviewers have noted this.

OWA
05-04-04, 03:58 PM
I read more, but don't have time to go into each one. Overall though, I'm just surprised ppl didn't kick up a big fuss about how much more elegant a solution the x800 is if you completely ignore performance(cooler, less power). Maybe it just doesn't matter that much to other people? I know the OEMs would sure care a lot...
And if you got to hold the card, you'll realize how small it is. Almost cute if you ask me :) 6800 is a beast. But again, not everyone cares about that.
I think it is something you just get use to. :) I was use to really big cards having owned just about every nvidia card. When I got the AIW 9800 Pro I was pretty shocked at how small the card was. So, yes, I do like that about ATI but I'm also so use to having such large cards, etc. I don't really pay that any mind these days.

fallguy
05-04-04, 04:06 PM
Oh yes, and I as well am puzzled over extremetech's results.

Anandtech's was extremely neutral.
Tom's tested really useless settings.
Hardocp, though they didn't lie, really catered the tests to make ati look good (tested max playable settings)



Wait, so you're going to buy a $500 top of the line card, and not play with max settings? Uh, ok...

Also, they used medium, and maximum details in games.

stncttr908
05-04-04, 04:10 PM
Hexus review as well was rubbish, though that is also probably due to time limit.Yeah, something was really askew with those benchies. :screwy:

eesa
05-04-04, 04:12 PM
Wait, so you're going to buy a $500 top of the line card, and not play with max settings? Uh, ok...

Also, they used medium, and maximum details in games.

I didn't say I didn't like hardocp's review. I actually found it quite telling... I was just trying to be fair and make the point that you can make the results go both ways.

Besides, who ever said about paying 500 for a card? ;) I'm not exactly getting a 420 purely out of loyalty ya know :)

eesa
05-04-04, 04:13 PM
Yeah, something was really askew with those benchies. :screwy:

yeah, like using the crippled 8x fsaa mode. Sure I like ati to look good, but that just made no sense.

Cota
05-04-04, 04:38 PM
I am aware of the methodology that hardocp used for their review, the thing that puzzles me is that is almost every review the performance difference between the X800pro and X800-XT varies a lot, in some are almost neck to neck and some others the x800-xt takes the lead by a wide margin.

Its too early to draw conclusions, since there are so many factors to take into consideration.

Drivers, processors, memory, mobos, resolution, image quality settings, all can make a difference in benches.

Also both cards have options that will increase speed and image quality, we'll have to wait for developers to use them.

The battle is getting really interesting

Edge
05-04-04, 05:33 PM
I think this whole comparision thing is just plain crap at this point. We have half the reviews test totally different settings between cards (8xSAA vs. 6xA), we have half that post VERY questionable results (6800 being slower than a 9800xt in one test?), and the few that post somewhat comparable results do so at such low settings that it's hard to even tell if it's GPU limited (for these monster cards, even at 1600x1200 you're CPU limited in most tests with "average" CPUs). So basically, we can't even draw any logical conclusions so far.

Then there's the "other" issues that affect your choice, like the size/heat of the card (though didn't many benchmarks say the 6800 was unusually cool at default clocks?), SM3.0 compatability, final clocks vs. reviewed clocks, different driver versions (the 61.xx drivers actually REDUCE performance in many games),

Wait, so you're going to buy a $500 top of the line card, and not play with max settings? Uh, ok...

Also, they used medium, and maximum details in games.

Actually, I think he's refering to the way that HardOCP's original article didn't do ANY direct comparision between the 6800 and other cards, just "the 6800 is running at 4xAA while the 9800 is running at lower resolution without AA" type situations, which is almost totally useless for a BENCHMARK. And also they tested 8xS on Nvidia cards, which should NOT be compared to ATI's 6xAA because 8xS does MUCH more work and affects things differently, it would be like comparing 4x Super-sample AA on a 5900 card to 4x MSAA on an ATI card.

eesa
05-04-04, 07:46 PM
there's different testing philosophies.

- Some base it on absolute fps at a specific resolution/setting.
- Some base it on how high of a resolution you can play comfortably at while having the eye candy cranked up.
- Some base it on using image quality as a baseline (subjective) and compare the cards on that merit (hence the 6x vs. 8xS comparison).

I think each methodology has its own merit and offers a unique perspective. They're focusing on different areas. For example, I would consider the first method as more "synthetic" since it tries to ensure that the "amount of work" being done is the same. Has no data points for image quality. The second method is less quantitative and is more gameplay oriented. The third method focuses on image quality as the equalizer and only compares performance once that is the same. How much work that's actually being done is irrelevant. Wouldn't it be cool if there was one comprehensive review had all three? :)

I think if you just take one benchmark or even one method of testing and make a decision based on that, you are missing a lot of the whole picture.

walkndude
05-04-04, 08:24 PM
Thanks for the level headed posts eesa, you've been making alot of good points in other threads as well... nice change of pace from the whats been crawling out of the woodwork lately. :)

Have you started at ATI yet ?

eesa
05-04-04, 08:27 PM
yup, except today I didn't errr.... get as much work done as other days. heh.

hovz
05-04-04, 09:44 PM
its because hardocp doesnt use standard demos for the most part, they use their own or just do a level run through with fraps. therefor ur getting the performance of the hardware, and not the driver optimisations. its much harder to optimize every scene in an entire game then a benchmark being used all over the web. obviously this isnt set in stone, but its what i gather from reading many of their reviews

fallguy
05-04-04, 10:32 PM
Actually, I think he's refering to the way that HardOCP's original article didn't do ANY direct comparision between the 6800 and other cards, just "the 6800 is running at 4xAA while the 9800 is running at lower resolution without AA" type situations, which is almost totally useless for a BENCHMARK.

I doubt it. This review does an "apples to apples" comparison, for people who cry foul with the other system, or just dont like it.

Edge
05-04-04, 10:35 PM
They didn't do that comparision with their original 6800 review, though, which is what I was refering to when I mentioned their "original" article. Also some of those "apples to apples" comparisions are very questionable...a 6800 coming in slower than a 9800 in Farcry? That totally goes against every other bit of evidence I've seen.

walkndude
05-04-04, 10:54 PM
Honestly I like hardocp's idea behind their testing methods - it attempts to cut right to the meat of the question, unfortunately in practice it seems rather difficult to quantify what constitutes "playable" settings...

Edge
05-04-04, 11:15 PM
Yeah, that's why I didn't like the way they were doing it: they had an odd way of choosing the "best" solution. For example in one game they felt 1600x1200 with 2xAA was best for the x800, while 1280x1024 with 4xAA was best for the 6800. Umm, how close would the two cards have been if they used the same settings on them? Seems rather odd to turn down the resolution by a notch on one card and then raise the AA up a notch.

eesa
05-04-04, 11:16 PM
personally, I'd start with max AF, then see if I can reach 1280 comfortably. In that order.

Then turn up AA and see how far that goes.