In House HD2900XT vs. 8800GTS 640

[apoppin]

Originally posted by: Noema
Looking forward to more benchies guys.

:thumbsup:

thanks ... i updated Post Number Four of this thread with my HD2900xt benchmarks ... i'm taking a brief break [a couple of hours between GPUs] before i get busy again ... D/L'ing the nvidia oc'ing tool ... and am going to play STALKER ... and then bench it on XP, right now

i also bought Call of Juarez ... does it have StarForce? if so, back it goes [unopened]

==================

strangely, FarCry installed fine on Vista ... i think i only used one in-between patch ... but how do you get FRAP to run during the demo? ... i need a console command or else i only get averages

otoh, anything over PainKiller patch 1.5 refused to install on Vista

and PREY looks like it has Starforce with patch v1.3 [some crap installed] ... it needs that damn activation... i could bypass it on XP, but not on Vista

and i have tomorrow and Wednesday "off" ... maybe also thru the weekend 🙂

 

[LittleNemoNES]

Originally posted by: apoppin
<div class="FTQUOTE"><begin quote>Originally posted by: Noema
Looking forward to more benchies guys.

:thumbsup:
</end quote></div>

thanks ... i updated Post Number Four of this thread with my HD2900xt benchmarks ... i'm taking a brief break [a couple of hours between GPUs] before i get busy again ... D/L'ing the nvidia oc'ing tool ... and am going to play STALKER ... and then bench it on XP, right now

i also bought Call of Juarez ... does it have StarForce? if so, back it goes [unopened]

==================

strangely, FarCry installed fine on Vista ... i think i only used one in-between patch ... but how do you get FRAP to run during the demo? ... i need a console command or else i only get averages

otoh, anything over PainKiller patch 1.5 refused to install on Vista

and PREY looks like it has Starforce with patch v1.3 [some crap installed] ... it needs that damn activation... i could bypass it on XP, but not on Vista

and i have tomorrow and Wednesday "off" ... maybe also thru the weekend 🙂

you can always use a crack for the .exe

 

[apoppin]

you can always use a crack for the .exe

\works for XP ... no prob, no acvtivation ... but walk me thru Vista

and what about 'Call of Juarez' [pronounced in Spanish as "Warez"] ... any good? ... and is it infested with StarForce?

... and STALKER looks just as good with the XT as the GTS
[i can't hear my XT over the gunfire, comrade!]!.. now for some benches ... 😛

 

[apoppin]

Here's my first GTS vs XT bench ... STALKER ... to be fair, HD2900xt was benched on Vista and GTS640 OC on XP

STALKER/Vista [FDL]HD2900xt

Building time demo [ran twice with FRAPS] 42/41 Min. // 365/364 Max. // 86.719/86.906 Av.

Short time demo [also 2x with FRAPS] 35/34 Min. // 345/349 Max. // 82.721/83.59 Av.
+++++++++++++++++++++
8800GTS-640M OC results - STALKER/XP [FDL]
Building time demo [ran twice with FRAPS] 10.5/11.5 Min. // 173/171 Max. // 74.867/74.153 Av.

Short time demo [also 2x with FRAPS] 24/21 Min. // 161/165 Max. // 64.476/65.441 Av.

WtH, i am getting some terrible minimums with the GTS
😕

 

[trake1]

Hi, Just started looking at this thread and I am having trouble reading any results.

Is there some readable table or graph that shows the results cleanly, next to each other?

I can't tell which FPS are for the 8800 and which are for the 2900 in most of the posts.

It's a great effort you are unertaking though ... Thanks!

 

[apoppin]

is Keys around ... ?
[i haven't heard anything from him since Sat]

He is updating the website ... check the first 4 posts ... mine is No 4
... i am just working on raw data now ... and keeping a notebook ... i will eventually organize it into something readable - and easily comparative ... GOOD NEWS ... i have today and tomorrow OFF ... MAYBE Thursday and Friday too --for sure Sat and Sun ..
--ONE of my cards goes back next Monday ... expect 'something' *finished* by then

and i DON'T know why my GTS is scoring so LOW in STALKER ... about to run 3DMARK05/6 to see if my rig is in line

AND i got Call of Juarez
-is it worth playing
-is it worth benchmarking?

EDIT OK, i need help ...

http://service.futuremark.com/...=14&XLID=0&UID=9937058

9515 3DMark06 ... almost 1000 points less than my HD2900xt [with CPU at 3.0Ghz] ... so that seems "close" ... but WtH is up with the GTS' STALKER scores ... it also stutters occasionally when playing with everything maxed in-game] so UN-like my HD2900xt

of course, i am "comparing Vista Stalker/3DMark 2900XT vs. XP Stalker/3DMark 8800GTS
- frankly i was expecting the results reversed 😛


huh?
😕

i will run a virus scan on this partition - just in case

 

[nullpointerus]

EDIT: I believe the HD2900XT is known for scoring very high in 3DMark06 relative to its real-world performance.

Did you test the 8800GTS under Vista? <--This could be an XP configuration problem of some kind.
Did you install chipset drivers for the 8800GTS on XP?
Did you defragment the disk? <--Diablo II actually stuttered wildly unless I did this after install 😕

I will PM you links about the "marforce" workarounds since they could be misconstrued as piracy.

 

[apoppin]

Originally posted by: nullpointerus
(1)Did you test the 8800GTS under Vista? <--This could be an XP configuration problem of some kind.
(2)Did you install chipset drivers for the 8800GTS on XP?
(3)Did you defragment the disk? <--Diablo II actually stuttered wildly unless I did this after install 😕

I will PM you links about the "marforce" workarounds since they could be misconstrued as piracy.

1. No ... that is my last setup ... GTS on Vista ... i want to get the bugs out of this XP install first

2. not "specifically" ... i used my Gigabyte drivers for the chipset and the drivers off the BFG install disk to set up GTS, then i ran the latest ForceWare uninstaller and the latest Forceware off nvidia's site

3. Not yet ... i didn't derag the disks for the XT benches either ... when i run all my benches again - everything all over - [is it fast and easy once you set up and know exactly what you are doing] i will do that and virus scans too [from the only partition with AntiVir installed]

thanks for your help
--i want to get this right!

what is strange is that 3DmMark06 is so "close" .. chipset drivers should tank it ... i guess i'll defrag all my drives .. now .. anyway, i'd expect it 1000 points higher ... thanks for the PM ... i'll see what i can 'do' for PREY and CoJ 🙂

 

[Keysplayr]

Originally posted by: apoppin
is Keys around ... ?
[i haven't heard anything from him since Sat]

He is updating the website ... check the first 4 posts ... mine is No 4
... i am just working on raw data now ... and keeping a notebook ... i will eventually organize it into something readable - and easily comparative ... GOOD NEWS ... i have today and tomorrow OFF ... MAYBE Thursday and Friday too --for sure Sat and Sun ..
--ONE of my cards goes back next Monday ... expect 'something' *finished* by then

and i DON'T know why my GTS is scoring so LOW in STALKER ... about to run 3DMARK05/6 to see if my rig is in line

AND i got Call of Juarez
-is it worth playing
-is it worth benchmarking?

EDIT OK, i need help ...

http://service.futuremark.com/...=14&XLID=0&UID=9937058

9515 3DMark06 ... almost 1000 points less than my HD2900xt [with CPU at 3.0Ghz] ... so that seems "close" ... but WtH is up with the GTS' STALKER scores ... it also stutters occasionally when playing with everything maxed in-game] so UN-like my HD2900xt

of course, i am "comparing Vista Stalker/3DMark 2900XT vs. XP Stalker/3DMark 8800GTS
- frankly i was expecting the results reversed 😛


huh?
😕

i will run a virus scan on this partition - just in case

--------------------------

Sorry I've been so quiet. I've been working on all of this in silence for a bit. Rethinking things. On good authority, I was told that the buildings_timedemo results are not really reliable results. The 2900XT gets insane maxfps because some graphics cards render open sky much better than others. in that demo, you can see that the person who recorded it pans to the sky on a few occasions. The best way to bench stalker is to record our own demos by doing a straight walkthrough. position yourself in the middle of a gunfight or something with a lot of activity. open the console, type "demo_record [demoname]". then use that demo for your benchmarks. It may take you several times to aquire a good bench that your satisfied with.

Back to work for me. 🙂

 

[bfdd]

Apoppin, make sure the drivers are fully uninstalled and try reinstalling I guess. If that doesn't fix it then *shrug* your GTS just hates stalker then lol. Those min frames seem weak as all hell. Hope you find a solution.

 

[manowar821]

You're awesome, keep up the good work!

 

[apoppin]

Originally posted by: keysplayr2003

Originally posted by: apoppin
is Keys around ... ?
[i haven't heard anything from him since Sat]

He is updating the website ... check the first 4 posts ... mine is No 4
... i am just working on raw data now ... and keeping a notebook ... i will eventually organize it into something readable - and easily comparative ... GOOD NEWS ... i have today and tomorrow OFF ... MAYBE Thursday and Friday too --for sure Sat and Sun ..
--ONE of my cards goes back next Monday ... expect 'something' *finished* by then

and i DON'T know why my GTS is scoring so LOW in STALKER ... about to run 3DMARK05/6 to see if my rig is in line

AND i got Call of Juarez
-is it worth playing
-is it worth benchmarking?

EDIT OK, i need help ...

... ing[/i] the results reversed 😛


huh?
😕

i will run a virus scan on this partition - just in case

--------------------------

Sorry I've been so quiet. I've been working on all of this in silence for a bit. Rethinking things. On good authority, I was told that the buildings_timedemo results are not really reliable results. The 2900XT gets insane maxfps because some graphics cards render open sky much better than others. in that demo, you can see that the person who recorded it pans to the sky on a few occasions. The best way to bench stalker is to record our own demos by doing a straight walkthrough. position yourself in the middle of a gunfight or something with a lot of activity. open the console, type "demo_record [demoname]". then use that demo for your benchmarks. It may take you several times to aquire a good bench that your satisfied with.

Back to work for me. 🙂

glad to hear from you ... i am also finding that some benches appear to favor nvidia and some favor AMD ... i do NOT like the STALKER benches at all ... NO game does what they do .. [no clip and flying into and out of the cells]

AND i think HardOCP's testing is "pure evil" ... i can "make" HD2900xt beat the GTS 640 OC every time and in most games .. max everything with 0/AA .. want to make GeForce win? ... Max AA
--nope ... not the way to test, imo and i will disregard every bench form that site as long as they have their silly method ... no wonder NO ONE uses it besides Kyle ... my criticism stands for that site [so far]

i definitely *needed* to defrag the only HD that GTS drivers were installed on ... that COULD account for the extra low scores ... but i defragged them ALL anyway

scanning all 5 Partitons showed no Viruses

so ... more testing
i am hearing CoJ is not so good ... i don't know if it is worth $40 or not and am hesitant to open it.

and expect a call tomorrow, Keys ... if you have the 4th 'off' and the free minutes ...

 

[apoppin]

quick follow-up

i am SO hesitant to create my own time demos ... that could also be a problem and we could have to exchange them to have something to really compare - your rig to mine

and most of the HW sites appear to use many the same time demos

'Max FPS' never meant much to me .. the minimum and average are what i always based it on

EDIT: NOPE, after defragging and scanning for Viruses ... same low 3DMark06 score

9548

i will run STALKER bench again and then R&R nvidia drivers

 

[apoppin]

UPDATE on STALKER

getting the same 3DMARK06 score ... but after defragging, the "short" demo in Stalker brought the minimum way up:
64.23 average // 25.7 minimum // 163.07

Buildings gave the same crap minimum ... ~11.62 each run .. there is a single "chug" that might account for the one blip in the time demo that keeps it 30FPS lower than Key's rig

i am *guessing* that 9500 is a "normal-low" for a GTS640OC ... and my AMD card is a "normal-high" ... could it be the CrossFire MB or is AMD better optimized for 3DMark06 better?

should i do the ForceWare driver reinstall?
--just to be sure

and i am getting my nap ... now [while D/L'ing Patch 1.07 to 1.08 for FEAR ... we'll update those] ... it's my day off ... and tomorrow ... and Thurs, Fri, Sat ... and Sun
... why do i feel like i am still working?

 

[pcslookout]

Originally posted by: apoppin

... why do i feel like i am still working?

Well because its what us computer geeks do for fun! Work on computer tests and benchmarks , which is still work.

 

[bfdd]

Originally posted by: apoppin
<div class="FTQUOTE"><begin quote>Originally posted by: keysplayr2003
<div class="FTQUOTE"><begin quote>Originally posted by: apoppin
is Keys around ... ?
[i haven't heard anything from him since Sat]

He is updating the website ... check the first 4 posts ... mine is No 4
... i am just working on raw data now ... and keeping a notebook ... i will eventually organize it into something readable - and easily comparative ... GOOD NEWS ... i have today and tomorrow OFF ... MAYBE Thursday and Friday too --for sure Sat and Sun ..
--ONE of my cards goes back next Monday ... expect 'something' *finished* by then

and i DON'T know why my GTS is scoring so LOW in STALKER ... about to run 3DMARK05/6 to see if my rig is in line

AND i got Call of Juarez
-is it worth playing
-is it worth benchmarking?

EDIT OK, i need help ...

... ing[/i] the results reversed 😛


huh?
😕

i will run a virus scan on this partition - just in case</end quote></div>

--------------------------

Sorry I've been so quiet. I've been working on all of this in silence for a bit. Rethinking things. On good authority, I was told that the buildings_timedemo results are not really reliable results. The 2900XT gets insane maxfps because some graphics cards render open sky much better than others. in that demo, you can see that the person who recorded it pans to the sky on a few occasions. The best way to bench stalker is to record our own demos by doing a straight walkthrough. position yourself in the middle of a gunfight or something with a lot of activity. open the console, type "demo_record [demoname]". then use that demo for your benchmarks. It may take you several times to aquire a good bench that your satisfied with.

Back to work for me. 🙂</end quote></div>
glad to hear from you ... i am also finding that some benches appear to favor nvidia and some favor AMD ... i do NOT like the STALKER benches at all ... NO game does what they do .. [no clip and flying into and out of the cells]

AND i think HardOCP's testing is "pure evil" ... i can "make" HD2900xt beat the GTS 640 OC every time and in most games .. max everything with 0/AA .. want to make GeForce win? ... Max AA
--nope ... not the way to test, imo and i will disregard every bench form that site as long as they have their silly method ... no wonder NO ONE uses it besides Kyle ... my criticism stands for that site [so far]

i definitely *needed* to defrag the only HD that GTS drivers were installed on ... that COULD account for the extra low scores ... but i defragged them ALL anyway

scanning all 5 Partitons showed no Viruses

so ... more testing
i am hearing CoJ is not so good ... i don't know if it is worth $40 or not and am hesitant to open it.

and expect a call tomorrow, Keys ... if you have the 4th 'off' and the free minutes ...

Well in all fairness, when you buy a high end video card don't you want to have max settings AND AA/AF turned on? I don't really think they were being shisty at all in their review, seeing as they kept it to max settings except when it was deemed unplayable. They always tried to keep AA/AF on because that's what most people will expect out of a high end card.

 

[Keysplayr]

Originally posted by: apoppin
<div class="FTQUOTE"><begin quote>Originally posted by: keysplayr2003
<div class="FTQUOTE"><begin quote>Originally posted by: apoppin
is Keys around ... ?
[i haven't heard anything from him since Sat]

He is updating the website ... check the first 4 posts ... mine is No 4
... i am just working on raw data now ... and keeping a notebook ... i will eventually organize it into something readable - and easily comparative ... GOOD NEWS ... i have today and tomorrow OFF ... MAYBE Thursday and Friday too --for sure Sat and Sun ..
--ONE of my cards goes back next Monday ... expect 'something' *finished* by then

and i DON'T know why my GTS is scoring so LOW in STALKER ... about to run 3DMARK05/6 to see if my rig is in line

AND i got Call of Juarez
-is it worth playing
-is it worth benchmarking?

EDIT OK, i need help ...

... ing[/i] the results reversed 😛


huh?
😕

i will run a virus scan on this partition - just in case</end quote></div>

--------------------------

Sorry I've been so quiet. I've been working on all of this in silence for a bit. Rethinking things. On good authority, I was told that the buildings_timedemo results are not really reliable results. The 2900XT gets insane maxfps because some graphics cards render open sky much better than others. in that demo, you can see that the person who recorded it pans to the sky on a few occasions. The best way to bench stalker is to record our own demos by doing a straight walkthrough. position yourself in the middle of a gunfight or something with a lot of activity. open the console, type "demo_record [demoname]". then use that demo for your benchmarks. It may take you several times to aquire a good bench that your satisfied with.

Back to work for me. 🙂</end quote></div>
glad to hear from you ... i am also finding that some benches appear to favor nvidia and some favor AMD ... i do NOT like the STALKER benches at all ... NO game does what they do .. [no clip and flying into and out of the cells]

AND i think HardOCP's testing is "pure evil" ... i can "make" HD2900xt beat the GTS 640 OC every time and in most games .. max everything with 0/AA .. want to make GeForce win? ... Max AA
--nope ... not the way to test, imo and i will disregard every bench form that site as long as they have their silly method ... no wonder NO ONE uses it besides Kyle ... my criticism stands for that site [so far]

i definitely *needed* to defrag the only HD that GTS drivers were installed on ... that COULD account for the extra low scores ... but i defragged them ALL anyway

scanning all 5 Partitons showed no Viruses

so ... more testing
i am hearing CoJ is not so good ... i don't know if it is worth $40 or not and am hesitant to open it.

and expect a call tomorrow, Keys ... if you have the 4th 'off' and the free minutes ...

Sure thing. I'll be home until about 2:30pm EST. Then off to a party. I'm playing a block party with my old band. Should be a blast.

 

[MrStryker]

I do not mean to flame at all or anything but... 😱 Are you trying to kill the 2900XT? You're using one of the most avoided PSUs, the Aspire 600W. They are near generic quality and the 2900XT might kill it soon. No kidding...

 

[apoppin]

http://www.gamershell.com/news/39093.html

http://downloads.guru3d.com/download.php?det=1642

is this for real? Call of Juarez has DX10 features?
:Q

my first "dx10" game ?
$40 is "cool" ... considering 😉

a BIG D/L ... 694 MB .. letsee IF i can find a site that allows D/L managers, i'll have it in 3 overnights 😛

Call of Juarez DirectX 10 Benchmark (694.7MB)

http://www.extremetech.com/art.../0,1697,2147111,00.asp
. . .has a June 18 bench

Physics based particle water system - Physics done via geometry shader

apoppin rips CoJ box open 😛
. . . lets take a looksee

now do i need the 694 MB d/l for the full game? according to this ...

http://www.gamespot.com/pc/action/thelawman/index.html

the patch i already have - v 1.1.0.0 - 94.7 MB already has the benchmark included and the DX10 path

Call of Juarez (DX10) v1.1.1.0 Patch

This patch includes the DirectX 10 benchmark and includes several bug fixes. This patch can only be installed in the DirectX 10 version of the game on Windows Vista. Download »

Posted Jun 20, 2007| Size: 44.3MB

Oh yeah ... gotta install it in Vista ... sigh ... ANOTHER [final] clean install for the GTS and CoJ
. . . see you in a few hours with DX10 "impressions"

 

[VinDSL]

Originally posted by: apoppin
a BIG D/L ... 694 MB .. letsee IF i can find a site that allows D/L managers, i'll have it in 3 overnights 😛

Jeez, apoppin...

Dude, why don't you buy a 4GB USB thumb drive... go to your local library... and borrow their bandwidth?

Every library I know of has computers with Internet access.

That's what the 911 terrorists used, so you should feel right at home! 😀

 

[superbooga]

Originally posted by: apoppin
UPDATE on STALKER

getting the same 3DMARK06 score ... but after defragging, the "short" demo in Stalker brought the minimum way up:
64.23 average // 25.7 minimum // 163.07

Buildings gave the same crap minimum ... ~11.62 each run .. there is a single "chug" that might account for the one blip in the time demo that keeps it 30FPS lower than Key's rig

Perhaps you should just ignore the minimum score. If the minimum fps only occurs for a split second, and the other scores are consistent with Key's scores, then it really shouldn't matter. A low minimum isn't the end of the world. In actual gameplay it's consistently near the minimum or getting a low fps in a critical area that hurts.

Besides you may just realized that simply reporting min,avg,max frame rates is not a perfect way to evaluate video cards. A fraps log would actually be much more useful.

 

[apoppin]

Originally posted by: superbooga

Originally posted by: apoppin
UPDATE on STALKER

getting the same 3DMARK06 score ... but after defragging, the "short" demo in Stalker brought the minimum way up:
64.23 average // 25.7 minimum // 163.07

Buildings gave the same crap minimum ... ~11.62 each run .. there is a single "chug" that might account for the one blip in the time demo that keeps it 30FPS lower than Key's rig

Perhaps you should just ignore the minimum score. If the minimum fps only occurs for a split second, and the other scores are consistent with Key's scores, then it really shouldn't matter. A low minimum isn't the end of the world. In actual gameplay it's consistently near the minimum or getting a low fps in a critical area that hurts.

Besides you may just realized that simply reporting min,avg,max frame rates is not a perfect way to evaluate video cards. A fraps log would actually be much more useful.

yes and no.

i AM ignoring that score ... BUT .. when actually playing STALKER with GTS, there is the occasional sudden drop of FPS that i used to find with Cats 7.5 that is now gone with HD2900xt ... so for my rig, STALKER is smoother on the XT
--not at all surprising as it has no AA.

You Do know that FRAP actually does slow down the game a bit and sometimes it actually interferes with the benchmark tool ... you have to run them over-and-over to actually "see" what is happening with all the combinations you run... then make it IDENTICAL for the other card.
==========

Originally posted by: VinDSL

Originally posted by: apoppin
a BIG D/L ... 694 MB .. letsee IF i can find a site that allows D/L managers, i'll have it in 3 overnights 😛

Jeez, apoppin...

Dude, why don't you buy a 4GB USB thumb drive... go to your local library... and borrow their bandwidth?

Every library I know of has computers with Internet access.

That's what the 911 terrorists used, so you should feel right at home! 😀

well, you seem to know a LOT about it ... actually, i didn't ... but thanks

and i have a 1GB and a 2GB drive but no library is open in the USA tomorrow and i ALREADY have the DX10 patch and bench for CoJ ... so you weren't paying attention. i think i have everything i need to bench GTS and XT ... and i almost had a heart attack ... i thought for an instant that i did a fresh install over my drive with the CoJ patch! [5 partitions do get a little confusing]

ANYway, GTX doesn't seem to like 3DMark06 with Vista any better

9172 in 3DMark06 ... not way off from other GTS scores with C2D ... so if you "play" 3dMark, i recommend the XT 😛
... and i found a later driver for my chipset .... so that didn't make any difference

Well in all fairness, when you buy a high end video card don't you want to have max settings AND AA/AF turned on? I don't really think they were being shisty at all in their review, seeing as they kept it to max settings except when it was deemed unplayable. They always tried to keep AA/AF on because that's what most people will expect out of a high end card.

Well, now that you know about the 2900 and AA, you can look at HardOCP review in a new light ... the 320-GTS is not the equal of the 2900xt EXCEPT when AA is fully maxed on the XT giving it twice the performance hit of the GTS [my wild guess ... i "made up" the 2x ... just from my 'impression' ... let's just say the XT takes a harder hit when 4x AA is enabled the the GTS does. ... from my limited testing and observations, so far]

on to STALKER .. just to compare XP with Vista and then i will install CoJ and look at my first "dx10" game ... hopefully i can get it to run

 

[apoppin]

OK ... last update before bedtime at a reasonable time ... i DARE not play CoJ or i will be up all night .. i must have a redneck's weakness for Westerns

Anyway , it installed on Vista - it IS patched 1.1.0.0 from the DVD ... so it has the DX10 pathway ... i only played into the tutorial, but it runs OK with every in-game setting maxed on the GTS ... so far ... i haven't seen anything special but it has only been a few minutes

as to STALKER, i ran the Vista benches with everything maxed ... the GTS is not ideal for this as it drops below 30FPS in a normal game ... needs better drivers, imo

building time demo - 87.9/88.0 average // 10.3/11.5 min // 600.85/621.22 max
short time demo --- 79.09/79.37 average // 22.57/20.42 min // 510.9/610.12 max

seeing those crazy high numbers with the GTS in Vista

for my rig it is no contest, i would pick the XT for STALKER with the current drivers

and i need to start the 'Juarez' benchmark D/L ... big one, i think

 

[DerekWilson]

Originally posted by: apoppin
quick follow-up

i am SO hesitant to create my own time demos ... that could also be a problem and we could have to exchange them to have something to really compare - your rig to mine

and most of the HW sites appear to use many the same time demos

'Max FPS' never meant much to me .. the minimum and average are what i always based it on

EDIT: NOPE, after defragging and scanning for Viruses ... same low 3DMark06 score

9548

i will run STALKER bench again and then R&R nvidia drivers

That's the tricky thing about MAX framerate ...

If I have a bench that's 100 seconds long and a /single/ frame gets rendered at 5000 fps (say, cause you're looking at a sky that can render at 5000 fps for one frame), it's not going to matter and I don't care. Lets say that arbitrarily this demo ended up getting an average of 40 fps. (generating a total of 4000 frames in 100 seconds).

But ...

what if, instead of one instantaneous frame that no one cares about, you pause on the sky for even 1 second. That will insert 5000 more frames into your benchmark, giving you an average of 90 fps rather than 40. And all this boost comes from 1/100th of the time you spend doing the benchmark.

yes my numbers are exaggerated for demonstration (though the oblivion pause screen can render at 3k+ fps on some cards) ...

but while in the real world we don't see instances where the difference is >2x performance, we do see real skew in the averages because of things like this.

What we really really like to see is a tight set of numbers -- min, average, max all near each other. This means we get a very steady experience.

...

While HardOCP's technique and presentation of the data leaves quite a lot to be desired, it does accomplish the task of presenting this type of variance. That said, we disagree with the way they do things on other levels, and we think there is a better way.

We've got something in the works to try and present this type of information (in addition to what we already show) in our articles ... but you'll have to wait and see what it is 🙂

 

[apoppin]

What we really really like to see is a tight set of numbers -- min, average, max all near each other. This means we get a very steady experience.

no kidding .. so instead, i find i get to do many many different benches over and over to get a "picture" of each card's respective performance ... and the hard part is conveying that picture to the reader.

While HardOCP's technique and presentation of the data leaves quite a lot to be desired, it does accomplish the task of presenting this type of variance. That said, we disagree with the way they do things on other levels, and we think there is a better way.

We've got something in the works to try and present this type of information (in addition to what we already show) in our articles ... but you'll have to wait and see what it is 🙂

AWESOME ... and looking forward to it and i think you know what we all want to see ... and thank-you!

yes, both Keys and i noticed and discussed the inconsistencies in the benchmarks in our PMs ... and i believe i know what HardOCP is attempting to present ... but their testing samples are way too few and we only get to see the "variance" the reviewer will to show us