I don't think it is the fact that it hasn't got a future, they want to push more people in to buying the bleeding edge cards. Just about every reviewer has recommended buying two GeForce 6800 GT's, as they clock up to GeForce 6800 Ultra clocks. They're trying to push people towards the top-end.
Also, if SLI hasn't got a future, why are ATI releasing their own implementation of multi-GPU rendering?
And considering they've had to create a chipset to make it work, and people need new mobos with dual slots etc, to pronounce SLI as "dead" seems somewhat premature tbh
I really don't see that they would do all that (above) and spend millions on marketing resurrecting the SLI name only to dump it a year later
Very true. The advent of PCI Express means that multi-participant rendering is here to stay. It'll only mature as time goes by, too.
As for X2 availability, I'd definitely put a finger on widespread availability of at least the 4200+ and 4400+ in June, given that they're floating around as we speak. I'm not so confident about the 4600+ and 4800+, though.
Nice work. I love that best-playable table. However on the midrange cards, most people won't be running an FX55, may wanna consider doing an addon with a 3500+ or something since most people would be around there with a pair of 6600s (or a single)
SLI will definately be here a while. I'm pretty sure that the reason nVidia only put the golden fingers on the 7800GTX (ie G70 Ultra) is so people don't do what they're doing now and buy the lower-clocked-same-pipelined variant and OC the difference. I'm not happy about that, but I think they'd have been better off going x800 style and just using cards with damaged pipelines to have the GTX be 32 pipes and the GT be 24 (or something like that) but at the same clock speed, if possible.
I think ATI's going the right way about multi-GPU over PCIE to some extent... trying to make sure they don't shaft their current customers and making it so any card can use it.
I really wanna see games once they're multithreaded and take that CPU bottleneck out. Because even 1600x1200 maxxed in CS:S, it "feels" like it's a CPU limitation.
Originally Posted by Firehed Nice work. I love that best-playable table. However on the midrange cards, most people won't be running an FX55, may wanna consider doing an addon with a 3500+ or something since most people would be around there with a pair of 6600s (or a single)
I think maybe that's why he ran the FX-55 @ 10 x 200
Originally Posted by Firehed However on the midrange cards, most people won't be running an FX55, may wanna consider doing an addon with a 3500+ or something since most people would be around there with a pair of 6600s (or a single)
I think everyone wants to know just how well the Bleeding Edge equipment runs. I dont want to know about midrange equipment on an SLI setup, because eh mid-range....there appears to be no such thing at the moment (Talking about graphics only here!). You can pretty much hazard a guess about where you would be with another CPU. But we all want an FX55 or atleast more of them :D
This SLI guide (and best on the net :o ) is huge, imagine eventual size when including additional processors
Originally Posted by Firehed Nice work. I love that best-playable table. However on the midrange cards, most people won't be running an FX55, may wanna consider doing an addon with a 3500+ or something since most people would be around there with a pair of 6600s (or a single)
Thanks, I think the table will stay, as it helps to sum up the article/review a little bit easier too. :)
Check the highlighted specification - this is a direct paste from our article creation engine with no emphasis added:
An A64 with 1MB L2 cache at 2200MHz is a 3700+, and an A64 with 1MB L2 cache at 2400MHz is a 4000+, so I think that an FX-55 at 2000MHz is roughly 3400-3500+ or there abouts. It's a fairly middle of the road CPU anyhow. If I had an Athlon 64 with 512K L2 Cache I would have used it for the third part, as I feel that there is no point in using a fast CPU when trying to find the best-playable settings on a GeForce 6600 GT, or GeForce 6600 GT SLI when using an FX-55, as it would not be a realistic system for the target user.
I've tried to make this as real-world as possible with the components I have available to me. Tomorrow's article should show you some more interesting findings. ;)
Im actually surprised at how little performance difference there is to be seen using higher end processors in a majority of games; glad I have my trusty Athlon XP 3000+ and Radeon 9800 Pro :p
Must have missed that bit bigz, thanks for pointing it out. I think today's CPU guide helps even more. And I think I finally got the HDR going right in Far Cry, and it finally does seem improved. Maybe I just was looking for graphical improvement and saw things I didn't before though, as I haven't really noticed the framerate drop that HDR seems to bring to most people.
zr_ox, yes, we want to know the bleeding edge, but not everyone can afford that. For SLI 6800GTs and an FX55, you're looking at almost $2000 when you add in the mobo, which could get you two sets of mid-upper range mobo/cpu/graphics.
I really didn't realise that HL2 was so CPU hungry but in line with your findings Bigz i did notice this without paying much attention as to why, and simply put it down to more applications running on my work machine (which probably does add to the overall reduced quality)
What i'm trying hard to come to terms with is that on two very similar machines (one Sli the other not) i can not see any difference at all in quality when playing HL2, both on max settings - 1600x1200 4xAA 16xAF Reflect All
First machine...
FX-53
Asus SK8V
Gainward 6800U Golden Sample
2x 512 OCZ 3200 ECC Reg (2-3-2-6)
Second machine
FX-55
Asus A8N Sli
2x Gainward 6800U Golden Sample
2x 512 OCZ 3200 EL Platinum Rev2 (2-2-2-5)
However on my work machine which is in line with the First Machine apart from CPU and all the junk running, i have to drop settings to 1024x768 with AA at 2x.
Work machine
3200+ 754
Lan Party UT nF3 250GB
Gainward 6800U Golden Sample
2x 512 OCZ 3200 EL Platinum Rev2 (2-2-2-5)
It could be as simple as down to bad eye sight or old age :D but would be interested to hear and invite any comments on this.
Irrespective, at the end of the day my two fav games are HL2 and NFS2 so your overall findings i'm happy with but would be interested on your thought and maybe add a couple of screenies to point me in the right direction as to see the benefits as i'll be damned if i can spot them
Oh yeah one last thing, does the OCZ PowerStream 520W have dedicated Sli connectors and would this make any difference as appose to the Noisetaker with dedicated lines??
Sorry for all that
I guess Part 4 must have answered as many questions as it did ask in my case lol.
The final part has now been published. It's a day later than planned, as we were awaiting some feedback from some of NVIDIA's partners before we could go ahead and publish. ;)
page 2 of the final part doesnt start until after the ads on the right in IE. as in scroll a page or 2 down. i recall this happening on another article too.
I only skimmed the article on CPU but Im pretty sure theres a basic mistake in there..
In your Doom 3 tests it shows average of just under 60.. thats because the default game config limits the FPS to 60 UNLESS you change it (max_fps If I remember correctly).
For non TFT users this command is the norm (put it to the monitor refresh rate) & once you do that you will see a difference between the CPU speeds, a big difference. It will give you results similar to HL2.
in both games you will also see that a higher av fps gives you a HIGHER MIN FPS, only reason i can think for this is its a % drop you get (especially in HL2 engine games), for example when i play HL2 or CS source with a capped fps of 60 the frame rate can drop to 40 or so in firefights, without capping or capping at 85 which I normally do it never goes below 60 as my FPS runs higher normally.
I would suggest trying those tests again with uncapped max FPS in both games & see what results you get.
Well after reading the articles and having used SLI, I agree on the most part. I'm a bit hesitant to recommend the DFI boards as much as you guys do, from what I've read in quite a few places they're really hard for a beginner to get working properly (ie, more than just putting things in the right places and enabling SLI in the BIOS), but on the other hand they're second-to-none for overclocking. I also remember reading somewhere that they wouldn't be 24-hour prime stable (or something to that extent) and although that's hardly a real issue, it's worth noting. The A8N in my experience is very easy to set up, quite stable and although doesn't have as much overclocking potential, still has a lot (mine would boot at 300htt, stable around 280-285 although I think it's my CPU holding me back at this point).
The point about driver support is excellent, it really could use some work. I'm not worried about game compatibility so much as the ability to use multiple monitors. I'm severely dissapointed by the fact that you can't use multiple displays when SLI is enabled, while disabling it requires a reboot and not just unchecking a box. Overall I love it, don't get me wrong, but in the next release, I'd be thrilled to see a taskbar icon (or an option to have one) that will let you very easily enable/disable SLI on the fly, or better yet an option for "Smart SLI" where it only does multi-VPU-rendering when in 3d mode, and leaves it as two seperate entities when in 2d mode.
If rebooting doesn't bother you or you have no inclination to use two+ monitors (remember, you'll have 4 DVI ports on a 6800GT/U setup), it doesn't matter in the slightest, but having the extra real estate was a great feeling. I'd be even more impressed if they added a display-size-scaling thing, as it's not uncommon to have two dissimilar-sided displays, and you'll have a dead spot where the mouse won't go over to the other screen even though it "should" (ie, displays don't line up pixel-for-pixel, you should be able to just indicate on the main monitor where the second monitor's boundaries are and have it scale the pointer position accordingly)
Just wanted to say thanks to Bit-net for this series of articles, some great info contained with in it, easily the most definitive acrticle on the subject I have come across on the net to date, keep up the good work people!
Originally Posted by mcmad I only skimmed the article on CPU but Im pretty sure theres a basic mistake in there..
In your Doom 3 tests it shows average of just under 60.. thats because the default game config limits the FPS to 60 UNLESS you change it (max_fps If I remember correctly).
For non TFT users this command is the norm (put it to the monitor refresh rate) & once you do that you will see a difference between the CPU speeds, a big difference. It will give you results similar to HL2.
in both games you will also see that a higher av fps gives you a HIGHER MIN FPS, only reason i can think for this is its a % drop you get (especially in HL2 engine games), for example when i play HL2 or CS source with a capped fps of 60 the frame rate can drop to 40 or so in firefights, without capping or capping at 85 which I normally do it never goes below 60 as my FPS runs higher normally.
I would suggest trying those tests again with uncapped max FPS in both games & see what results you get.
The games are left at their default settings, and Doom 3 has a capped FPS except for when you use the time demo function. I don't believe in benchmarking graphics cards with a time demo, because it does not represent true game play. The max_fps command you speak of does not work, and there are no commands that I can see relating to maximum frame rate in the console.
I don't believe in modifying the game from its default settings either, Doom 3 is capped at 60 fps as that is how ID Software designed the game to be. Half-Life 2 is totally uncapped, as was the developer's intention. There is no error, that is exactly how the end user will see the game if they install the game and play it IMHO.
I don't agree with the minimum frame rate statements you make either, it's dependant on any number of features, usually the speed of the video memory. If there is enough intensity in the graphics, you will see a massive drop, I've spent a great deal of time playing through all of the games I use to benchmark to be sure that I've found one of, if not the most graphic intense portions of the title for benchmarking purposes. The frame rates you see should be playable throughout the ENTIRE title, not just through my chosen section.
You see higher results when people use time demo's in Doom 3, but they are not realistic, and we all remember the cheating that went on during the GeForceFX era in 3DMark03.. using real games highlights exactly how your video card will perform when you play games. :)
Edit: Also, why would you want me to recommend that you spend even more money than you need to, when I've quite clearly proven that you can play these games on a slower CPU with SLI and still see very good performance? I don't get commissioned for every CPU or GPU sale I create, so I may as well advise you to spend your money wisely. That's what integrity is for.
Quote:
The point about driver support is excellent, it really could use some work. I'm not worried about game compatibility so much as the ability to use multiple monitors. I'm severely dissapointed by the fact that you can't use multiple displays when SLI is enabled, while disabling it requires a reboot and not just unchecking a box. Overall I love it, don't get me wrong, but in the next release, I'd be thrilled to see a taskbar icon (or an option to have one) that will let you very easily enable/disable SLI on the fly, or better yet an option for "Smart SLI" where it only does multi-VPU-rendering when in 3d mode, and leaves it as two seperate entities when in 2d mode.
I feel very confident that NVIDIA will continue to improve support of SLI, and I plan to continue to monitor its progress. As I said, I think its great, but it does need an awful lot of background work to make sure that the ongoing development continues. I'm sure that I will discuss this with NVIDIA when I am out in Taiwan at the beginning of next month.
Quote:
Bigz, you might want to give the article another read through; you messed up on a few occasions with useage of words and got a few tenses wrong
I'll have a look later - that's incredibly tired fingers and an incredibly tired bigz0r writing on the train home late thursday night/friday morning. Forgive me, I've written nearly 25k words since a week ago Thursday. (that's two and a half dissertations for anyone doing a university degree in the UK) ;)
Quote:
Just wanted to say thanks to Bit-net for this series of articles, some great info contained with in it, easily the most definitive acrticle on the subject I have come across on the net to date, keep up the good work people!
Oh, btw Firehed, I had issues with my first A8N SLI, it wouldn't boot at all. It also took me a while to get the gigabyte up and running. The DFI took a couple of days to get running at 100%, but that was related to early BIOS IMHO.
Hmm. I'm just going by what I've read in the 100+ page anandtech thread on the A8N. A lot of people say they switched to the DFI because it had more overclocking potential, but was a lot harder to initially get going. The BIOS for the whole lot of SLI boards seems to be improving too, which is great becuse I think a lot of the initial BIOS versions were rushed because of the hype surrounding SLI.
And see what Mr. nVidia Taiwan thinks about that display scaling thing (just for moving the mouse between the screens, since cross-screen app dragging would get really freaky otherwise). Hard to explain in text but I'm sure if you've ever run multi monitors of different size you know what I'm referring to.
But yeah, grats on having, hands down, the best and most thorough SLI guide/review on the net. A little bit more about the PSUs would have been good (amp ratings? I know ASUS recommends 18a for 6600GTs and like 24a for 6800Us, and I've heard in most cases one huge 12v rail is better than dual smaller ones since in most situations, all of the mobo and graphics power would be drawn from one rail, and the molex's from the other), but other than that, great work.
I'm not sure if this is the appropriate thread to mentioned this seeing as the articles are all about SLI but as you mentioned recommended specs in the final one I thought I'd just ask what you'd recommend as a cooler for an Athlon 64bit 3500+. I'm mainly after a fairly simple, air cooled quiet cooling solution. I don't need to overclock the cpu so as long as it does a reasonable job at cooling I'll be happy. Many thanks.
Comments 26 to 50 of 66
Also, if SLI hasn't got a future, why are ATI releasing their own implementation of multi-GPU rendering?
I really don't see that they would do all that (above) and spend millions on marketing resurrecting the SLI name only to dump it a year later
As for X2 availability, I'd definitely put a finger on widespread availability of at least the 4200+ and 4400+ in June, given that they're floating around as we speak. I'm not so confident about the 4600+ and 4800+, though.
http://www.bit-tech.net/hardware/2005/05/18/nvidia_sli_pt3/1.html
SLI will definately be here a while. I'm pretty sure that the reason nVidia only put the golden fingers on the 7800GTX (ie G70 Ultra) is so people don't do what they're doing now and buy the lower-clocked-same-pipelined variant and OC the difference. I'm not happy about that, but I think they'd have been better off going x800 style and just using cards with damaged pipelines to have the GTX be 32 pipes and the GT be 24 (or something like that) but at the same clock speed, if possible.
I think ATI's going the right way about multi-GPU over PCIE to some extent... trying to make sure they don't shaft their current customers and making it so any card can use it.
I really wanna see games once they're multithreaded and take that CPU bottleneck out. Because even 1600x1200 maxxed in CS:S, it "feels" like it's a CPU limitation.
I think maybe that's why he ran the FX-55 @ 10 x 200
(might be wrong like)
heh..... look mom i gone invisible
I think everyone wants to know just how well the Bleeding Edge equipment runs. I dont want to know about midrange equipment on an SLI setup, because eh mid-range....there appears to be no such thing at the moment (Talking about graphics only here!). You can pretty much hazard a guess about where you would be with another CPU. But we all want an FX55 or atleast more of them :D
This SLI guide (and best on the net :o ) is huge, imagine eventual size when including additional processors
This is fantastic, great stuff
Thanks, I think the table will stay, as it helps to sum up the article/review a little bit easier too. :)
Check the highlighted specification - this is a direct paste from our article creation engine with no emphasis added:
;)
An A64 with 1MB L2 cache at 2200MHz is a 3700+, and an A64 with 1MB L2 cache at 2400MHz is a 4000+, so I think that an FX-55 at 2000MHz is roughly 3400-3500+ or there abouts. It's a fairly middle of the road CPU anyhow. If I had an Athlon 64 with 512K L2 Cache I would have used it for the third part, as I feel that there is no point in using a fast CPU when trying to find the best-playable settings on a GeForce 6600 GT, or GeForce 6600 GT SLI when using an FX-55, as it would not be a realistic system for the target user.
I've tried to make this as real-world as possible with the components I have available to me. Tomorrow's article should show you some more interesting findings. ;)
zr_ox, yes, we want to know the bleeding edge, but not everyone can afford that. For SLI 6800GTs and an FX55, you're looking at almost $2000 when you add in the mobo, which could get you two sets of mid-upper range mobo/cpu/graphics.
What i'm trying hard to come to terms with is that on two very similar machines (one Sli the other not) i can not see any difference at all in quality when playing HL2, both on max settings - 1600x1200 4xAA 16xAF Reflect All
First machine...
FX-53
Asus SK8V
Gainward 6800U Golden Sample
2x 512 OCZ 3200 ECC Reg (2-3-2-6)
Second machine
FX-55
Asus A8N Sli
2x Gainward 6800U Golden Sample
2x 512 OCZ 3200 EL Platinum Rev2 (2-2-2-5)
However on my work machine which is in line with the First Machine apart from CPU and all the junk running, i have to drop settings to 1024x768 with AA at 2x.
Work machine
3200+ 754
Lan Party UT nF3 250GB
Gainward 6800U Golden Sample
2x 512 OCZ 3200 EL Platinum Rev2 (2-2-2-5)
It could be as simple as down to bad eye sight or old age :D but would be interested to hear and invite any comments on this.
Irrespective, at the end of the day my two fav games are HL2 and NFS2 so your overall findings i'm happy with but would be interested on your thought and maybe add a couple of screenies to point me in the right direction as to see the benefits as i'll be damned if i can spot them
Oh yeah one last thing, does the OCZ PowerStream 520W have dedicated Sli connectors and would this make any difference as appose to the Noisetaker with dedicated lines??
Sorry for all that
I guess Part 4 must have answered as many questions as it did ask in my case lol.
By far the best review i've read. 10/10
The OCZ PowerStream 520W that I have has no SLI connectors, but there are newer versions that do, as far as I know. :)
http://www.bit-tech.net/hardware/2005/05/21/nvidia_sli_pt5/1.html
In your Doom 3 tests it shows average of just under 60.. thats because the default game config limits the FPS to 60 UNLESS you change it (max_fps If I remember correctly).
For non TFT users this command is the norm (put it to the monitor refresh rate) & once you do that you will see a difference between the CPU speeds, a big difference. It will give you results similar to HL2.
in both games you will also see that a higher av fps gives you a HIGHER MIN FPS, only reason i can think for this is its a % drop you get (especially in HL2 engine games), for example when i play HL2 or CS source with a capped fps of 60 the frame rate can drop to 40 or so in firefights, without capping or capping at 85 which I normally do it never goes below 60 as my FPS runs higher normally.
I would suggest trying those tests again with uncapped max FPS in both games & see what results you get.
The point about driver support is excellent, it really could use some work. I'm not worried about game compatibility so much as the ability to use multiple monitors. I'm severely dissapointed by the fact that you can't use multiple displays when SLI is enabled, while disabling it requires a reboot and not just unchecking a box. Overall I love it, don't get me wrong, but in the next release, I'd be thrilled to see a taskbar icon (or an option to have one) that will let you very easily enable/disable SLI on the fly, or better yet an option for "Smart SLI" where it only does multi-VPU-rendering when in 3d mode, and leaves it as two seperate entities when in 2d mode.
If rebooting doesn't bother you or you have no inclination to use two+ monitors (remember, you'll have 4 DVI ports on a 6800GT/U setup), it doesn't matter in the slightest, but having the extra real estate was a great feeling. I'd be even more impressed if they added a display-size-scaling thing, as it's not uncommon to have two dissimilar-sided displays, and you'll have a dead spot where the mouse won't go over to the other screen even though it "should" (ie, displays don't line up pixel-for-pixel, you should be able to just indicate on the main monitor where the second monitor's boundaries are and have it scale the pointer position accordingly)
my 5 cents :D
The games are left at their default settings, and Doom 3 has a capped FPS except for when you use the time demo function. I don't believe in benchmarking graphics cards with a time demo, because it does not represent true game play. The max_fps command you speak of does not work, and there are no commands that I can see relating to maximum frame rate in the console.
I don't believe in modifying the game from its default settings either, Doom 3 is capped at 60 fps as that is how ID Software designed the game to be. Half-Life 2 is totally uncapped, as was the developer's intention. There is no error, that is exactly how the end user will see the game if they install the game and play it IMHO.
I don't agree with the minimum frame rate statements you make either, it's dependant on any number of features, usually the speed of the video memory. If there is enough intensity in the graphics, you will see a massive drop, I've spent a great deal of time playing through all of the games I use to benchmark to be sure that I've found one of, if not the most graphic intense portions of the title for benchmarking purposes. The frame rates you see should be playable throughout the ENTIRE title, not just through my chosen section.
You see higher results when people use time demo's in Doom 3, but they are not realistic, and we all remember the cheating that went on during the GeForceFX era in 3DMark03.. using real games highlights exactly how your video card will perform when you play games. :)
Edit: Also, why would you want me to recommend that you spend even more money than you need to, when I've quite clearly proven that you can play these games on a slower CPU with SLI and still see very good performance? I don't get commissioned for every CPU or GPU sale I create, so I may as well advise you to spend your money wisely. That's what integrity is for.
Thanks for your kind comments ;)
And see what Mr. nVidia Taiwan thinks about that display scaling thing (just for moving the mouse between the screens, since cross-screen app dragging would get really freaky otherwise). Hard to explain in text but I'm sure if you've ever run multi monitors of different size you know what I'm referring to.
But yeah, grats on having, hands down, the best and most thorough SLI guide/review on the net. A little bit more about the PSUs would have been good (amp ratings? I know ASUS recommends 18a for 6600GTs and like 24a for 6800Us, and I've heard in most cases one huge 12v rail is better than dual smaller ones since in most situations, all of the mobo and graphics power would be drawn from one rail, and the molex's from the other), but other than that, great work.
I'm not sure if this is the appropriate thread to mentioned this seeing as the articles are all about SLI but as you mentioned recommended specs in the final one I thought I'd just ask what you'd recommend as a cooler for an Athlon 64bit 3500+. I'm mainly after a fairly simple, air cooled quiet cooling solution. I don't need to overclock the cpu so as long as it does a reasonable job at cooling I'll be happy. Many thanks.