I think the generation of games reflected by the 3DMark predictions should be benchmarked side-by-side the various flavors of 3DMark to see how much validity their predictions held.
i.e. Benchmark 3DMark05 or 3DMark03 against current games, 3DMark2001SE against games from 2? years ago... that sort of thing.
What a horribly, boring benchmark. I paid for 2k5 but not this one. The last two CPU tests should be measured in seconds per frame instead of frames per second.
I nearly fell asleep when it was running on my X800PRO... I'm not even going to bother with my SM3.0 machines... boring + more tests = no thanks.
I got a nice 1418 score which means "everything was really, really slow!" on an X800XL.
Never mind though, I suppose at the rate the cards are coming along this might work decently in a couple of years time. They have to raise the bar somehow and I think this is probably quite a good benchmark but the hardware it's intended for barely exists yet.
It took hours to download via bittorrent, and when I finally ran it, my impression was "Same as last year's, but slower".
My (non-aggressively) overclocked AMD 3800+ X2 and 800 GTO2 got 2204 3DMarks, without performing the HDR/SM3.0 tests. I must say it was quite disturbing to have plunked down money I've been saving for years in Nov. to build a new machine to replace my antique PC, only to see a benchmark running like a slideshow once again by January. :-) And I'd just run the F.E.A.R. demo on it last night, too....
There are several outcome from PCDVD @ TWN
http://forum.pcdvd.com.tw/showthread.php?t=582913">http://forum.pcdvd.com.tw/showthread.php?t=582913 Dual core CPU score almost double in comparison to single core
Opteron 2.9G dual core score 2200, Opteron 2.9G single score 1110, a 1.99x increase
Furthermore, Intel score abnormally high to AMD, may due to hyperthreading
P4 3.0G HT (prescott) score 900 while as Athlon 64 2.5G also result in 900(san diego)
Additionally, old K7 also score desociated with real gaming,
K7 2.2G score 725 which is higher than K8 1.8G at 660, but we know it never happen in real world..
So, 3D mark 05 make everything out of reality, but 3D mark 06 still have "unreal" CPU score.
I find it handy for another purpose: to see if the performance of my computer has deteriorated over the time and to see if there are any problems I'm not aware of. So I run 3D Mark after freshly installing the system and I write down the score. Every now and then, or after software updates I run it again to see if everything is all right.
I wanna see some bences on a system usind an AGEIA PhysX card. Since this version of 3DMark supports that sdk it would be nice to see if there is any performance imporvement on systems using the card vs. same system not.
Very good observation A$$ for brains. What I was suggesting was that a hardware site, for instance Anandtech, could contact BFG or ASUS to get a card for testing. Both have said that the cards are ready and BFG even had a card at CES.
Fixed and tweaked. I will definitely be running 3DMark06 in the future, but more as a stress test than for performance comparisons, like in the overclocking articles I've done.
LOL! I love Elimidate, although Elimidate would not provide a deep, meaningful connection, but more shallow beauty (admitedly, whoring themselves in public and bitching at each other at an ever increasing volume). Good times.
"Right now, the hardware that is available is prompting advancements in game development, and we can't easily predict what types of games we might see in the near or semi-near future."
You mean we won't have more WW2 shooters, with the occassional relief Zombie Mutant Alien? But now everything will have Bright Lights and Dark Shadows! Because that seems extremely likely to me. Game developers rarely chase new game types, and it's not really the hardware that motivates them AFAICT.
X800 Pro = 12 pipelines at 475 MHz and 980 MHz GDDR3 RAM.
X800 GTO = 12 pipelines at 400 MHz and 980 MHz GDDR3 RAM.
However, many people have had success in unlocking and/or overclocking GTO cards. If you can get 16 pipelines at 475 MHz, for example, it would be 33% faster on the core than the Pro. If you just overclock to 475 MHz and don't unlock the pipelines, you've got an X850 Pro. (R480 core vs. R420 core, I think? It doesn't make much difference, though.)
3dMark is and always has been primarily a GPU-oriented benchmark, Intel vs. AMD wouldn't tell us much if the GPU is the bottleneck(and if it isn't, all it would tell us is that AMD outperformed Intel like they tend to do in these kinds of tests).
3DMark06 actually has a CPU test function that runs a test at 640x480 and 2fps. This should take out the effect of any card able to run 3DMark06. So you could test them. I don't know that you can make a definite conclusion as the test will run multithreaded - and I don't know how many multi-threaded games we will see for this current crop of cards.
The ATI cards take a big hit in the testing because they can't run 24 bit depth stencil textures. 24 bit DSTs are optional for DX9 and ATI only supports the required 16 bit DSTs. On the other hand, the reason there are no results with AA enabled is that the nVidia cards don't do muntipoint blending and multisampling AA at the same time, so 3DMark06 doesn't report a score.
3dMark has a number of "Feature Tests" that test specific features such as fill rate, VS, PS, CPU, and triangle performance. These tests are outside of the "Game Tests" run to find a 3dMark score, hence they're effectively extra uses for the program. Also don't discount 3dMark for being a really good diagnostic program, both to determine if a rig is stable, and if it's performing at levels it should be at(thanks to the large comparison DB).
I would like to add that this "The overall 3DMark scores don't really give us much more information than we already have. Other than simply letting us know what hardware runs 3DMark better. "
is very true.
3dmark is meaningless, besides for competition. Which could be done in a more meaningful sense (real gaming benchmark comparisons).
Its one of the biggest crocks going today. Go upgrade your video cards so you can hit the magic 10,000 again with this years 3dmark.. um.. woot?
I pity the fool who "plays" 3dmark.
Agreed, the number of 3dmarks is basically useless. However, it's a good benchmark for yourself so you can see if a certain tweak you made to your system helped or not.
"4 X1600 cores (R580) will give some 6855 total score, not bad"
Will R580 have 4x memory bandwidth? X1600XT has 22GB/s of memory bandwidth.
I wouldn't call X1600XT scores very impressive. X1600XT is running at 590MHz core/590MHz memory and with 12 pipelines is 34% faster than 500MHz core/500MHz memory 8 pipeline 6600 GT (which is 1.5 years old chip).
I came across this old 3DMark06 and thought: why not test it with my current setup and see how it does against the fastest cards of early 2006 ?
The result is actualy shocking:
Total: 14301 SM2.0: 6180 SM3.0: 7886 CPU: 2780
It's not that I have any kind of super game machine. It's just a WindowsXP machine with an Intel E-7600 and a Nvidia 560 Ti. Not overclocked or anything.
That is roughly tripple the score. And not even tested on an high-end system B-)
Thanks for letting my add a comment to this old post Anandtech, many sites do not allow that.
If any one else reads this, please take the time to test your setup as well. If not, I may be back in a few years and test again :-)
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
45 Comments
Back to Article
neogodless - Thursday, January 19, 2006 - link
The future is here...I think the generation of games reflected by the 3DMark predictions should be benchmarked side-by-side the various flavors of 3DMark to see how much validity their predictions held.
i.e. Benchmark 3DMark05 or 3DMark03 against current games, 3DMark2001SE against games from 2? years ago... that sort of thing.
Or has this been done by someone?
Wellsoul2 - Thursday, January 19, 2006 - link
The CPU part seems to be a total joke.I saw very little difference between my XP2800 and Opteron 148.
Pretty much negligible, where in games I saw 10FPS.
MrSmurf - Wednesday, January 18, 2006 - link
What a horribly, boring benchmark. I paid for 2k5 but not this one. The last two CPU tests should be measured in seconds per frame instead of frames per second.I nearly fell asleep when it was running on my X800PRO... I'm not even going to bother with my SM3.0 machines... boring + more tests = no thanks.
stephenbrooks - Thursday, January 19, 2006 - link
I got a nice 1418 score which means "everything was really, really slow!" on an X800XL.Never mind though, I suppose at the rate the cards are coming along this might work decently in a couple of years time. They have to raise the bar somehow and I think this is probably quite a good benchmark but the hardware it's intended for barely exists yet.
alcalde - Wednesday, January 18, 2006 - link
It took hours to download via bittorrent, and when I finally ran it, my impression was "Same as last year's, but slower".My (non-aggressively) overclocked AMD 3800+ X2 and 800 GTO2 got 2204 3DMarks, without performing the HDR/SM3.0 tests. I must say it was quite disturbing to have plunked down money I've been saving for years in Nov. to build a new machine to replace my antique PC, only to see a benchmark running like a slideshow once again by January. :-) And I'd just run the F.E.A.R. demo on it last night, too....
nv40 - Thursday, January 19, 2006 - link
There are several outcome from PCDVD @ TWNhttp://forum.pcdvd.com.tw/showthread.php?t=582913">http://forum.pcdvd.com.tw/showthread.php?t=582913
Dual core CPU score almost double in comparison to single core
Opteron 2.9G dual core score 2200, Opteron 2.9G single score 1110, a 1.99x increase
Furthermore, Intel score abnormally high to AMD, may due to hyperthreading
P4 3.0G HT (prescott) score 900 while as Athlon 64 2.5G also result in 900(san diego)
Additionally, old K7 also score desociated with real gaming,
K7 2.2G score 725 which is higher than K8 1.8G at 660, but we know it never happen in real world..
So, 3D mark 05 make everything out of reality, but 3D mark 06 still have "unreal" CPU score.
shabby - Wednesday, January 18, 2006 - link
Ya the cpu test was pretty pointless, i think i got like 5 frames per minute!Btw i scored 900 with a gf6800nu.
Souka - Thursday, January 19, 2006 - link
Took 12min to download via my T3Tossed onto my laptop..T42 Pentium M 1.7ghz and Radeon 9800 w/64mb 0-1fps...stopped it after 5 min
tossed onto my desktop.. Pentium 4HT @2.8 and GF4 AGP 6800 w/128....0-1fps...stopped it after 5 min
Wow...what a piece. Yeah, its the demo, but still.
e4te - Thursday, January 19, 2006 - link
The cup test is capped at 2 fps I'm pretty sure.e4te - Thursday, January 19, 2006 - link
hehe cpuZak - Wednesday, January 18, 2006 - link
I find it handy for another purpose: to see if the performance of my computer has deteriorated over the time and to see if there are any problems I'm not aware of. So I run 3D Mark after freshly installing the system and I write down the score. Every now and then, or after software updates I run it again to see if everything is all right.<Z>
bigpow - Wednesday, January 18, 2006 - link
Would be curious to know what's the score of Apple's new Quadcore G5Powermoloch - Wednesday, January 18, 2006 - link
What happen to the X850XT/PE or X800XT/PE ?JamesDax - Wednesday, January 18, 2006 - link
I wanna see some bences on a system usind an AGEIA PhysX card. Since this version of 3DMark supports that sdk it would be nice to see if there is any performance imporvement on systems using the card vs. same system not.ViRGE - Wednesday, January 18, 2006 - link
What good would a PPU do? In the standard 3dMark benchmark, there's no physics calculations going on.Griswold - Wednesday, January 18, 2006 - link
Let me run to that hardware shop at the corner and buy one to test it. Oh wait...JamesDax - Wednesday, January 18, 2006 - link
Very good observation A$$ for brains. What I was suggesting was that a hardware site, for instance Anandtech, could contact BFG or ASUS to get a card for testing. Both have said that the cards are ready and BFG even had a card at CES.Griswold - Thursday, January 19, 2006 - link
Fuck you too, buddy.FrozenCanadian - Wednesday, January 18, 2006 - link
In the last line of the article:"While we will certainly be not be reporting 3DMark scores "
JarredWalton - Wednesday, January 18, 2006 - link
Fixed and tweaked. I will definitely be running 3DMark06 in the future, but more as a stress test than for performance comparisons, like in the overclocking articles I've done.Mant - Wednesday, January 18, 2006 - link
Quote: "But at the end of our testing, we are mostly left with shallow beauty rather than a deep, meaningful connection."wtf?
stephenbrooks - Thursday, January 19, 2006 - link
Quote: "But at the end of our testing, we are mostly left with shallow beauty rather than a deep, meaningful connection."Ah, but on the other hand, "There are likely many other uses for this program which we can't mention here".
In any case, this is way more interesting than your average graphics benchmark review.
Mant - Wednesday, January 18, 2006 - link
In case you think I'm making that up, its at the end of Page 3. Methinks Josh needs Elimidate more than 3DMark06Orbs - Thursday, January 19, 2006 - link
LOL! I love Elimidate, although Elimidate would not provide a deep, meaningful connection, but more shallow beauty (admitedly, whoring themselves in public and bitching at each other at an ever increasing volume). Good times.peldor - Wednesday, January 18, 2006 - link
"Right now, the hardware that is available is prompting advancements in game development, and we can't easily predict what types of games we might see in the near or semi-near future."You mean we won't have more WW2 shooters, with the occassional relief Zombie Mutant Alien? But now everything will have Bright Lights and Dark Shadows! Because that seems extremely likely to me. Game developers rarely chase new game types, and it's not really the hardware that motivates them AFAICT.
KingofL337 - Wednesday, January 18, 2006 - link
Whats the relation ship of these cards. Which is ahigher performance part?
JarredWalton - Wednesday, January 18, 2006 - link
X800 Pro = 12 pipelines at 475 MHz and 980 MHz GDDR3 RAM.X800 GTO = 12 pipelines at 400 MHz and 980 MHz GDDR3 RAM.
However, many people have had success in unlocking and/or overclocking GTO cards. If you can get 16 pipelines at 475 MHz, for example, it would be 33% faster on the core than the Pro. If you just overclock to 475 MHz and don't unlock the pipelines, you've got an X850 Pro. (R480 core vs. R420 core, I think? It doesn't make much difference, though.)
DigitalFreak - Wednesday, January 18, 2006 - link
Let the driver "optimizations" begin!PeteRoy - Wednesday, January 18, 2006 - link
No Intel vs AMD?ViRGE - Wednesday, January 18, 2006 - link
3dMark is and always has been primarily a GPU-oriented benchmark, Intel vs. AMD wouldn't tell us much if the GPU is the bottleneck(and if it isn't, all it would tell us is that AMD outperformed Intel like they tend to do in these kinds of tests).Skiplives - Wednesday, January 18, 2006 - link
3DMark06 actually has a CPU test function that runs a test at 640x480 and 2fps. This should take out the effect of any card able to run 3DMark06. So you could test them. I don't know that you can make a definite conclusion as the test will run multithreaded - and I don't know how many multi-threaded games we will see for this current crop of cards.The ATI cards take a big hit in the testing because they can't run 24 bit depth stencil textures. 24 bit DSTs are optional for DX9 and ATI only supports the required 16 bit DSTs. On the other hand, the reason there are no results with AA enabled is that the nVidia cards don't do muntipoint blending and multisampling AA at the same time, so 3DMark06 doesn't report a score.
Extremetech did an article about the technical issues (no real testing like Josh did) http://www.extremetech.com/article2/0,1697,1912131...">here.
Regards,
Chris
superkdogg - Wednesday, January 18, 2006 - link
"There are likely many other uses for this program which we can't mention here"Ummm, what's he talking about?
ViRGE - Wednesday, January 18, 2006 - link
3dMark has a number of "Feature Tests" that test specific features such as fill rate, VS, PS, CPU, and triangle performance. These tests are outside of the "Game Tests" run to find a 3dMark score, hence they're effectively extra uses for the program. Also don't discount 3dMark for being a really good diagnostic program, both to determine if a rig is stable, and if it's performing at levels it should be at(thanks to the large comparison DB).Rampage - Wednesday, January 18, 2006 - link
In the 2nd performance graph, "Shader Modle 2.0" should be Shader Model 2.0.Rampage - Wednesday, January 18, 2006 - link
I would like to add that this "The overall 3DMark scores don't really give us much more information than we already have. Other than simply letting us know what hardware runs 3DMark better. "is very true.
3dmark is meaningless, besides for competition. Which could be done in a more meaningful sense (real gaming benchmark comparisons).
Its one of the biggest crocks going today. Go upgrade your video cards so you can hit the magic 10,000 again with this years 3dmark.. um.. woot?
I pity the fool who "plays" 3dmark.
theslug - Wednesday, January 18, 2006 - link
Agreed, the number of 3dmarks is basically useless. However, it's a good benchmark for yourself so you can see if a certain tweak you made to your system helped or not.Phiro - Wednesday, January 18, 2006 - link
Not true, all you are going to do is tell if the tweaks/changes you made to your system helped or hurt 3DMark06, not Game X Y or Z.There's just too many ways to develop at this point for this artificial benchmark to be meaningful.
DerekWilson - Wednesday, January 18, 2006 - link
fixedRampage - Wednesday, January 18, 2006 - link
Whoa that was fast!gordon151 - Wednesday, January 18, 2006 - link
Pretty benchmark and looks to favour the x1600xt pretty nicely.MrKaz - Wednesday, January 18, 2006 - link
That means that R580 will boom GTX512 like X1800XT gets boomed by it now.4 X1600 cores (R580) will give some 6855 total score, not bad…
defter - Saturday, January 21, 2006 - link
"That means that R580 will boom GTX512 like X1800XT gets boomed by it now."Here are some X1900XT benchmarks: http://www.overclockers.ru/images/news/2006/01/21/...">http://www.overclockers.ru/images/news/2006/01/21/...
They used a faster CPU, but if we ignore the CPU tests results should be comparable with Anand's results:
PS2.0 score:
X1900XT (625/1450) with FX-60: 2081
7800GTX 512 with FX-55: 2167
X1800XT with FX-55: 1611
PS3.0 score:
X1900XT (625/1450) with FX-60: 2279
7800GTX 512 with FX-55: 2204
X1800XT with FX-55: 1697
X1900XT seems to be a little slower than 7800GTX 512 while X1900XTX should be a little faster. In any case the difference is only 2-3%.
defter - Thursday, January 19, 2006 - link
"4 X1600 cores (R580) will give some 6855 total score, not bad"Will R580 have 4x memory bandwidth? X1600XT has 22GB/s of memory bandwidth.
I wouldn't call X1600XT scores very impressive. X1600XT is running at 590MHz core/590MHz memory and with 12 pipelines is 34% faster than 500MHz core/500MHz memory 8 pipeline 6600 GT (which is 1.5 years old chip).
Boushh - Tuesday, June 19, 2012 - link
I came across this old 3DMark06 and thought: why not test it with my current setup and see how it does against the fastest cards of early 2006 ?The result is actualy shocking:
Total: 14301
SM2.0: 6180
SM3.0: 7886
CPU: 2780
It's not that I have any kind of super game machine. It's just a WindowsXP machine with an Intel E-7600 and a Nvidia 560 Ti. Not overclocked or anything.
That is roughly tripple the score. And not even tested on an high-end system B-)
Thanks for letting my add a comment to this old post Anandtech, many sites do not allow that.
If any one else reads this, please take the time to test your setup as well. If not, I may be back in a few years and test again :-)
Boushh - Tuesday, August 11, 2015 - link
And I'm back again !!Now the test was done with the same system, but with a GeForce GTX760. Oddly enough, the scores are lower ?
Total: 12295
SM2.0: 5342
SM3.0: 6031
CPU: 2785
I had at least expected double the score or something like that B-(
l, let's see how it goes when I have a new card....