Hmm, dailytech's X1950xtx scores seem VERY low for q4 compared to AT's...

[dug777]

http://www.anandtech.com/video/showdoc.aspx?i=2833&p=9

http://www.dailytech.com/article.aspx?newsid=4812

I know who i trust in this, AT articles are usually spell checked and grammatically correct before they post them up for starters 😉

EDIT: i'm aware they don't say which bench they use for it, but to get them that far apart you'd have to set out to intentionally make the score as low as possible 😛

EDIT: to make the thread OP slightly less imflammatory, i do understand what you are saying BFG :thumbsup:

 

[GTaudiophile]

Anandtech is very bias towards nVidia. Didn't you know?

 

[schneiderguy]

Dailytech probably did a benchmark run with more action/more complex scenes/more complicated effects (or all of those) than anandtech did.

 

[coldpower27]

Originally posted by: dug777
http://www.anandtech.com/video/showdoc.aspx?i=2833&p=9

http://www.dailytech.com/article.aspx?newsid=4812

Seriously, how could they possibly be that incompetent? :roll:

I know who i trust in this, AT articles are usually spell checked and grammatically correct before they post them up for starters 😉

EDIT: i'm aware they don't say which bench they use for it, but to get them that far apart you'd have to set out to intentionally make the score as low as possible 😛

You can't compare benchmarks across different websites.

Dailytech =/= Anandtech.

They probably utilized one of the most heavy situations in Quake 4 to help showcase the power of the Geforce 8800 GTX, no sense CPU limiting the card. 😛

 

[Stumps]

meh...if it isn't trident I'm not interested

TVGA FTW

 

[josh6079]

I brought this up before here.

Originally posted by: schneiderguy
Dailytech probably did a benchmark run with more action/more complex scenes/more complicated effects (or all of those) than anandtech did.

Then it looks like Kentsfield and the X1950XTX is a waste, since I regularly get that kind of a score with my X2 4400+ and X1900XT at that same resolution and AA setting.

IMO, DailyTech really should have checked and re-checked those scores before throwing them out on the net. Kentsfield, an X1950XTX @1600x1200 w/4xAA brings more to the table than that in Q4. Especially when AT has the same card with a Conroe @ 2048x1536 w/ 4xAA and is getting ~20 frames more than that. If I do benches and see that my results are that far off from other various benches around the web, I test, test, and retest to be sure.

 

[dug777]

Originally posted by: schneiderguy
Dailytech probably did a benchmark run with more action/more complex scenes/more complicated effects (or all of those) than anandtech did.

I'm sure someone might have noticed that the AT time demo results give you framerates that are over twice as fast as playable conditions in the game, even on average in the hardest level, on a faster computer 😛

This kind of difference might be explained if dailtytech set out to find the hardest possible scene then just stared at it, but even then, i doubt the AT time demo consists of staring at a wall to get maxty11tybillionfps 😛

 

[dug777]

Originally posted by: coldpower27

Originally posted by: dug777
http://www.anandtech.com/video/showdoc.aspx?i=2833&p=9

http://www.dailytech.com/article.aspx?newsid=4812

Seriously, how could they possibly be that incompetent? :roll:

I know who i trust in this, AT articles are usually spell checked and grammatically correct before they post them up for starters 😉

EDIT: i'm aware they don't say which bench they use for it, but to get them that far apart you'd have to set out to intentionally make the score as low as possible 😛

You can't compare benchmarks across different websites.

Dailytech =/= Anandtech.

They probably utilized one of the most heavy situations in Quake 4 to help showcase the power of the Geforce 8800 GTX, no sense CPU limiting the card. 😛

yeah, because quad core kentsfield and cpu limitation go hand in hand...wait...no they don't :Q

 

[BFG10K]

Seriously, how could they possibly be that incompetent?

It's nothing to do with incompetence, you just don't appear to understand simple benchmarking concepts.

Do you understand the fact that different sections of the game have different performance dynamics and unless the demo is exactly the same it's invalid to compare two scores?

I'm sure someone might have noticed that the AT time demo results give you framerates that are over twice as fast as playable conditions in the game, even on average in the hardest level, on a faster computer

Again I'll ask whether you understand the very simple concept I have just presented to you?

To compare scores from two different benchmarks is beyond ludicrous.

 

[dug777]

Originally posted by: BFG10K

Seriously, how could they possibly be that incompetent?

It's nothing to do with incompetence, you just don't appear to understand simple benchmarking concepts.

Do you understand the fact that different sections of the game have different performance dynamics and unless the demo is exactly the same it's invalid to compare two scores?

I'm sure someone might have noticed that the AT time demo results give you framerates that are over twice as fast as playable conditions in the game, even on average in the hardest level, on a faster computer

Again I'll ask whether you understand the very simple concept I have just presented to you?

To compare scores from two different benchmarks is beyond ludicrous.

You're still as abrasive as ever BFG 😛

I think it's perfectly valid to expect a reasonable degree of similarity between two different benchmarks of the same game, assuming that the benchmarks are constructed to show an average cross section of play (indoors/outdoors etc), and if they aren't constructed to do that, they're utterly useless to us as a general guide to how the card performs.

This is a HUGE disparity, i'm sure even you with that infamous donkey-like stubbornness might be prepared to concede that?

 

[BFG10K]

yeah, because quad core kentsfield and cpu limitation go hand in hand...wait...no they don't

Err, what? Quad core brings nothing to the table for Quake 4 performance over dual core.

Not that the CPU has anything to do with the issue given it's clearly caused by differing benchmarks.

It's trivially easy to construct two Quake 4 demos and have one perform at twice the speed (or more) of the other.

 

[BFG10K]

You're still as abrasive as ever BFG

Abrasive? You're the one calling people incompetent because your understanding is sorely lacking.

I think it's perfectly valid to expect a reasonable degree of similarity between two different benchmarks of the same game,

Then you still don't understand benchmarking or gaming.

Like I said it's trvial to construct cross-sections of actual gameplay and get wildly different scores.

This is a HUGE disparity,

That doesn't mean either result is wrong.

i'm sure even you with that infamous donkey-like stubbornness might be prepared to concede that?

Pot-kettle-black. I expect the same tactics from you here as we got when you were trying to convince us that 512 MB cards weren't needed because you couldn't understand that a single benchmark isn't representative of the whole game.

 

[dug777]

Originally posted by: BFG10K

You're still as abrasive as ever BFG

Abrasive? You're the one calling people incompetent because your understanding is sorely lacking.

I think it's perfectly valid to expect a reasonable degree of similarity between two different benchmarks of the same game,

Then you still don't understand benchmarking or gaming.

Like I said it's trvial to construct cross-sections of actual gameplay and get wildly different scores.

This is a HUGE disparity,

That doesn't mean either result is wrong.

i'm sure even you with that infamous donkey-like stubbornness might be prepared to concede that?

Pot-kettle-black. I expect the same tactics from you here as we got when you were trying to convince us that 512 MB cards weren't needed because you couldn't understand that a single benchmark isn't representative of the whole game.

lol, if you want to go back to that, i saw you trying to recommend a 7950 pro 512mb over the 256mb x1900xt the other day, despite the xt being faster in every useful situation 😉 That's great advice :thumbsup:

I'm sticking to my guns here, by dailytech's scores the minimum card you need to play q4 at 1600x1200 4xAA is an x1950xtx, and we know that plainly isn't true 😉 When they got scores that are so totally out of line with every other site on the net, they should have at least checked them and explained why this was so...

 

[ronnn]

Originally posted by: BFG10K

Seriously, how could they possibly be that incompetent?

It's nothing to do with incompetence, you just don't appear to understand simple benchmarking concepts.

Do you understand the fact that different sections of the game have different performance dynamics and unless the demo is exactly the same it's invalid to compare two scores?

I'm sure someone might have noticed that the AT time demo results give you framerates that are over twice as fast as playable conditions in the game, even on average in the hardest level, on a faster computer

Again I'll ask whether you understand the very simple concept I have just presented to you?

To compare scores from two different benchmarks is beyond ludicrous.

Oh well in a few days we will know. The 8800gtx will be almost twice as fast as the x1950xtx or not. If is only 50% faster, than we will know that dailytech likely made this little piece up.

 

[BFG10K]

lol, if you want to go back to that, i saw you trying to recommend a 7950 pro 512mb over the 256mb x1900xt the other day, despite the xt being faster in every useful situation

So what if it's faster? There are other factors involved in a video card decision, factors that I mentioned in the recommendation.

I'm sticking to my guns here, by dailytech's scores the minimum card you need to play q4 at 1600x1200 4xAA is an x1950xtx,

For their particular benchmark it may well be true. Plenty of games have sections that slow down more than the usual sections but that doesn't necessarily mean DailyTech are incompetent.

 

[BFG10K]

Oh well in a few days we will know. The 8800gtx will be almost twice as fast as the x1950xtx or not. If is only 50% faster, than we will know that dailytech likely made this little piece up.

Sure, provided the other reviews use the same benchmark DailyTech did.

If they use more CPU limited demos - like Anand's one which is actually quite a light load compared to the rest of the game - then we'll be back to where we started.

 

[schneiderguy]

Originally posted by: dug777

This kind of difference might be explained if dailtytech set out to find the hardest possible scene then just stared at it, but even then, i doubt the AT time demo consists of staring at a wall to get maxty11tybillionfps 😛

I never got the full version of Quake4, but the opening level of the demo (the one AT tested on) wasnt that stressfull on my system... at most there were two enemies on my screen at once.

Im assuming there were much larger battles in the game or huge outdoor scenes in later parts of the game, which is where Dailytech probably got their benchmark from. 🙂

 

[ronnn]

I played through the game and it never really seemed that demanding -only had to think twice :thumbsup: But I was busy during the busy parts and didn't have fraps running. Still if there are only a few select parts of current games that can use a 8800gtx - well it would be nice to know before upgrading. No doubting for 3dmark and benchmarking fans - the 8800gtx is the boss.

 

[dug777]

Originally posted by: schneiderguy

Originally posted by: dug777

This kind of difference might be explained if dailtytech set out to find the hardest possible scene then just stared at it, but even then, i doubt the AT time demo consists of staring at a wall to get maxty11tybillionfps 😛

I never got the full version of Quake4, but the opening level of the demo (the one AT tested on) wasnt that stressfull on my system... at most there were two enemies on my screen at once.

Im assuming there were much larger battles in the game or huge outdoor scenes in later parts of the game, which is where Dailytech probably got their benchmark from. 🙂

yeah i'd like to see some methodology behind their results that's for sure 🙂

 

[josh6079]

Whoa, this escalated quickly...

As far as the time-demos go, I believe that you can't compare AT's scores directly to DT's and say one or the other is wrong. I do believe though that one should expect similar results over a variety of benches while all using upstanding hardware.

IMO it's not that the the DT's score was a flawed representation of the time-demo they ran, just that whatever time-demo they ran was a misrepresentation of the game as a whole.

If they picked the most intense section of the game to bench, state that. Instead of presenting a vague, all-encompassing "Quake 4" bench, tell us that it was really just "Quake 4: [insert chapter or level name] section." Otherwise they're claiming that the average frame rate Quake 4 lets a Kentsfield/X1950XTX get using the least useful AA and resolution of 1600x1200 is a mere 34 fps.

For instance, in Oblivion benches we don't care what the dungeon frames are because 99% of the time they're around 60 or higher. That's why when you see relevant Oblivion benches, they're segregated into "Outdoor", "Foliage", or "Oblivion Gate" sections.

While I certainly wouldn't have made a thread about it, I don't see the crime in questioning two sites different benches when the results are in such opposition to eachother. Afterall, this was just a preview and I do think it's safe to say that whatever time-demo made the X1950XTX get it's 34fps average, the 8800GTX underwent the same torture and did quite well.

I'm not going to get my underwear in a twist about whose benches were better and what not. I know how Quake 4 plays and what it takes to play it well. The point of the preview was to show just what realm of frames to expect from G80, and it accomplished that. It was never designed to show what realm of frames to expect from the X1950XTX since there are a great many different benches that do that while being more detailed.

G80 is 3 days away, it's already been demonstrated that it has the frames, I just hope that nVidia didn't throw IQ out of the window for its still very dominant DX9 userbase.

 

[Modular]

I think we're looking at this all wrong. The only way that DT could have gotten such low scores for the 1950 is if they used a rediculously heavy timedemo right? Right.

So we can agree that the timedemo was incredibly taxing on even such an amazing card as the 1950.

Look at how well the 8800gtx did on the same timedemo! I can't wait to see what it's actual numbers are going to be. If the 1950 is about 50% lower than it "actually" is for most people, then we can assume the same for the 8800gtx! That means that it should put up like 50% higher scores than we see on that report! I can't wait.

 

[Nelsieus]

Originally posted by: Modular
I think we're looking at this all wrong. The only way that DT could have gotten such low scores for the 1950 is if they used a rediculously heavy timedemo right? Right.

So we can agree that the timedemo was incredibly taxing on even such an amazing card as the 1950.

Look at how well the 8800gtx did on the same timedemo! I can't wait to see what it's actual numbers are going to be. If the 1950 is about 50% lower than it "actually" is for most people, then we can assume the same for the 8800gtx! That means that it should put up like 50% higher scores than we see on that report! I can't wait.

Exactly.

What Dug and Josh seem to be forgetting is that the 8800GTX was tested on the exact same time-demo.

So even though the x1950XTX score seems low, I would also argue that the Geforce 8800GTX's score seems a bit low (compared to the others), as well. It still leads by a significant margin, but not as significant as the other games tested. That tells you the time demo probably was the most intense, and that the 8800GTX managed to hold up well.

Now, if it were a case of the x1950XTX being tested on the most intense, and the 8800GTX on the least intense, then you would obviously have a point to make. But both scores are consistent with eachother and whatever time-demo was used for the testing.

Nelsieus

 

[ronnn]

Originally posted by: Modular
I think we're looking at this all wrong. The only way that DT could have gotten such low scores for the 1950 is if they used a rediculously heavy timedemo right? Right.

So we can agree that the timedemo was incredibly taxing on even such an amazing card as the 1950.

Look at how well the 8800gtx did on the same timedemo! I can't wait to see what it's actual numbers are going to be. If the 1950 is about 50% lower than it "actually" is for most people, then we can assume the same for the 8800gtx! That means that it should put up like 50% higher scores than we see on that report! I can't wait.

No the issue is does the 1950 really perform that poorly playing these games? If dailytech had to cherry pick rare scenes to get these scores - it may mean in most situations the 8800gtx is cpu limited or deliver frame rates that are surplus to a good gaming experience. If this is the case - people may want to wait for games and/or cpu's to catch up before upgrading or spend their money more wisely on other parts. Always nice to see what you get in real terms for your money. I am looking forward to dx10 and the games it brings. Hope the adoption is not as slow as sm2 and 3 or I will be not buying until the midrange refreshes are out for compliant hardware. :beer:

 

[Nelsieus]

Originally posted by: ronnn

Originally posted by: Modular
I think we're looking at this all wrong. The only way that DT could have gotten such low scores for the 1950 is if they used a rediculously heavy timedemo right? Right.

So we can agree that the timedemo was incredibly taxing on even such an amazing card as the 1950.

Look at how well the 8800gtx did on the same timedemo! I can't wait to see what it's actual numbers are going to be. If the 1950 is about 50% lower than it "actually" is for most people, then we can assume the same for the 8800gtx! That means that it should put up like 50% higher scores than we see on that report! I can't wait.

No the issue is does the 1950 really perform that poorly playing these games? If dailytech had to cherry pick rare scenes to get these scores - it may mean in most situations the 8800gtx is cpu limited or deliver frame rates that are surplus to a good gaming experience. If this is the case - people may want to wait for games and/or cpu's to catch up before upgrading or spend their money more wisely on other parts. Always nice to see what you get in real terms for your money. I am looking forward to dx10 and the games it brings. Hope the adoption is not as slow as sm2 and 3 or I will be not buying until the midrange refreshes are out for compliant hardware. :beer:

I don't know how you can draw that conclusion? 😕

I would take 65FPS vs. 35FPS any day.
And if it were a matter of being CPU limited, all of the scores would be relatively close to eachother. We saw the Geforce 8800GTX lead in most of those apps by almost 92%, with it leading by 65% in the Q4 benchmark. That's hardly evident of CPU bottleneck.

Secondly, when will people realize that being CPU bottlenecked is a good thing and means free AA and IQ? You make it sound as if the extra GPU power is taken and thrown away somewhere - that's not the case.

Nelsieus

 

[Drayvn]

Originally posted by: Nelsieus

Originally posted by: ronnn

Originally posted by: Modular
I think we're looking at this all wrong. The only way that DT could have gotten such low scores for the 1950 is if they used a rediculously heavy timedemo right? Right.

So we can agree that the timedemo was incredibly taxing on even such an amazing card as the 1950.

Look at how well the 8800gtx did on the same timedemo! I can't wait to see what it's actual numbers are going to be. If the 1950 is about 50% lower than it "actually" is for most people, then we can assume the same for the 8800gtx! That means that it should put up like 50% higher scores than we see on that report! I can't wait.

No the issue is does the 1950 really perform that poorly playing these games? If dailytech had to cherry pick rare scenes to get these scores - it may mean in most situations the 8800gtx is cpu limited or deliver frame rates that are surplus to a good gaming experience. If this is the case - people may want to wait for games and/or cpu's to catch up before upgrading or spend their money more wisely on other parts. Always nice to see what you get in real terms for your money. I am looking forward to dx10 and the games it brings. Hope the adoption is not as slow as sm2 and 3 or I will be not buying until the midrange refreshes are out for compliant hardware. :beer:

I don't know how you can draw that conclusion? 😕

I would take 65FPS vs. 35FPS any day.
And if it were a matter of being CPU limited, all of the scores would be relatively close to eachother. We saw the Geforce 8800GTX lead in most of those apps by almost 92%, with it leading by 65% in the Q4 benchmark. That's hardly evident of CPU bottleneck.

Secondly, when will people realize that being CPU bottlenecked is a good thing and means free AA and IQ? You make it sound as if the extra GPU power is taken and thrown away somewhere - that's not the case.

Nelsieus

I think what hes trying to say is that are people going to get into such a heavy environment for your GPU or is it made so that it can stress it, much like 3DMock.