Can this 16x SLI bencmark be right?

I mean 50% plus performance boost????

Why the hell is this not glowing neon Red on the front page of the [H] if those numbers are true?

I would presume this to be groundbreaking news as much as the 35% ATI driver fix was a few days back.
 
ill be ordering one the second they hit the retail channel. cant wait to see if it makes that big of a difference. i use SLI AA on everything.
 
PhoenixC5 said:
ill be ordering one the second they hit the retail channel. cant wait to see if it makes that big of a difference. i use SLI AA on everything.
Click to expand...


You better hold off and make sure that shit is right. LOL...would hate to see you was your money.
 
PhoenixC5 said:
http://www.abit-usa.com/products/mb/products.php?categories=1&model=310

or that...or the asus.

but the msi has decent audio and two slots between. the asus has a wierd setup with all that extra space between the video cards. i want to have aftermarket cooling on my cards, but dang, you could put an 18-wheeler in the space on the asus.
Click to expand...

For starters, the MSI and Asus boards have the exact same amount of room between the two slots. The difference is the Asus has two PCI, MSi has two PCIE 1x. And as far as it being to much room, most people prefer it that way. Gives dual slot cooling more room to breath, and its pretty much the only way to go if your watercooling. You can use low profile blocks on the other boards, but it gets tight with the screws coming up though the back of the card.
 
actually, you are right. the difference is how far down on the mobo the 16x slots start. i just dont like the layout at much as the MSI. but, as stated, i understand why the slots are there and also as stated, i want to use aftermarket cooling. ;) hence the msi choice over the abit (even tho i love abit :()
 
Ingonuts13 said:
I mean 50% plus performance boost????

Why the hell is this not glowing neon Red on the front page of the [H] if those numbers are true?

I would presume this to be groundbreaking news as much as the 35% ATI driver fix was a few days back.
Click to expand...

wow... don't know, but i remember anandtech guesstimating something between 5-10% max AFAIK. if these are legit nVidia shot ATI in the foot... again. :(
 
man that's awesome! I'm anxious to see more reviews on it now... although to me it seems kind of weird that they're only running at 1280X1024. Anyone know why they would do that? I mean can't one 6800 ultra run farcry at 1600X1200?
 
Double Jesus said:
man that's awesome! I'm anxious to see more reviews on it now... although to me it seems kind of weird that they're only running at 1280X1024. Anyone know why they would do that? I mean can't one 6800 ultra run farcry at 1600X1200?
Click to expand...


Not really on that 6800Ultra Question. I have one and it will run, but it's definitely chuggin aloing in areas and not quite as fast as one would like.

Also, I still do not see how SLI AA can saturated the 8x PCIE bus that much that it would gain such a peformance increas going with 16x.
 
I don't really think that can be right. I've had a SLi x16 board for quite some time. (nForce Professional) and I don't think my system's performance is any different in reference to 7800GTX performance vs any other motherboards.

I kind of doubt we are saturating the x8 PCI-E bus, much less the x16 bus if we still hadn't tapped out AGP's bandwidth yet. I seriously doubt PCI-E itself has become the bottleneck in less than two years time.
 
wizzackr said:
wow... don't know, but i remember anandtech guesstimating something between 5-10% max AFAIK. if these are legit nVidia shot ATI in the foot... again. :(
Click to expand...

It seems to me (assuming these benchies are correct) that nvidia didn't do jack shit besides make a solution that can only do x8 SLI. The board partners then went out and combined two chipsets onto one board to in effect obtain x16. nVidia didn't shoot ATI in the foot....the mobo makers did (if anyone).
 
I also have had an SLI x16 mobo (nforce pro once again) for a while, and im not seeing performance gains like that, then again, i dont have 7800gtx sli either, only 6800gt. But mine benches just like everything else (besides higher cpu score due to opteron)

Edit: Speaking of socket 940, nice to see a fellow enthusiast Sir Fragalot, ive had three opterons since they came out and have no intention of switching to 939 ever. And who can beat nforce pro :). I designate myself unofficial vice-emporor of the socket 940 club
 
Erasmus354 said:
It seems to me (assuming these benchies are correct) that nvidia didn't do jack shit besides make a solution that can only do x8 SLI. The board partners then went out and combined two chipsets onto one board to in effect obtain x16. nVidia didn't shoot ATI in the foot....the mobo makers did (if anyone).
Click to expand...

true. i seriously doubt these scores are anywhere near the performance gains we'll see anyways.
 
jbrukardt said:
I also have had an SLI x16 mobo (nforce pro once again) for a while, and im not seeing performance gains like that, then again, i dont have 7800gtx sli either, only 6800gt. But mine benches just like everything else (besides higher cpu score due to opteron)

Edit: Speaking of socket 940, nice to see a fellow enthusiast Sir Fragalot, ive had three opterons since they came out and have no intention of switching to 939 ever. And who can beat nforce pro :). I designate myself unofficial vice-emporor of the socket 940 club
Click to expand...

Welcome Vice Emporer. :cool:

I've got two dual Opterons, and I feel no need to go to S939 either. Back on topic, my benchmarks are consistant with FX-57 based machines running two SLi 7800GTX's. Granted I've not run alot of in game benches, only synthetic, but I seriously doubt there would be that gain. I wish I had a machine with a A8N-SLi Deluxe/Premium that I could bench, then upgrade to an A8N32-SLi Deluxe and see if there is an increase.
 
jbrukardt said:
I also have had an SLI x16 mobo (nforce pro once again) for a while, and im not seeing performance gains like that, then again, i dont have 7800gtx sli either, only 6800gt. But mine benches just like everything else (besides higher cpu score due to opteron)
Click to expand...


Well the benchmarks I was specifically referring to, were the SLI 7800GTX with AA turned on. According to how that reads, the reviewer is saying that turning on AA oversaturates a normal 8X PCIE bus to the point that it cannot send data fast enough. Those nuumbers shown indicate that when this happens, using 16X SLI can again you over 50% performance.

Your 6800GT does not push enough data for it to saturate the bus. I can see this making sense, it is just doesn't seem like it should be this much of an improvement.
 
well, if they ever come out, i have a baseline and can test it. but...still havent seen them in the retail channel yet. id love to know if its really that good, because i use SLI AA a lot
 
Good point until a reputable site puts the mobo to the test in games like Q4, FEAR, HL2 and that lame game FarCrap i wont believe it. It sounds too good to be true and if it is i would gladly spend money to get it :D
 
I would love to find out if this is true myself. I am betting that it isn't.
 
The question is, "Do you really think that the previous SLI setup was actually bottlenecked by the PCIe bus?"
 
Yeah but we must note the article is talking specifically about SLI-AA. What the hell is SLI-AA? They first tested the cards with and without AA/AF. But then without any introduction they discuss this new feature SLI-AA. How is it any better than normal AA?
 
Matrices said:
Yeah but we must note the article is talking specifically about SLI-AA. What the hell is SLI-AA? They first tested the cards with and without AA/AF. But then without any introduction they discuss this new feature SLI-AA. How is it any better than normal AA?
Click to expand...


SLI-AA was released a while back in drivers. [H] did an article on it, check the archives. To my understanding, in SLI-AA 8x, each card does 4x AA, same goes for SLI-AA 16x, each card does 8x. I could be wrong though, thats just off memory. Read the article if you want a more in depth answer...as well as benchmarks and such.
 
Who in God's name is going to notice the difference between 4xAA and 8xAA in any resolution above 1024x768 - I mean, that is just overkill, redefined. I don't think it's worth spending an extra $50-$100 on yet another new board...the existing SLI ones just started coming down in price, too.

Consider also what hothardware first had to say about SLI-AA:

"There are two side effects to these new SLI anti-aliasing modes. First, because AA is being applied to the frame being rendered on each GPU, and the frame needs to be offset and blended, additional memory is used as a buffer. So in some cases, the GPU may run out of frame buffer memory and have to swap into-and-out-of system memory, which will adversly affect performance. Secondly, due to the fact that each GPU is working on the same frame, there is no performance increase associated to running a pair of GeForce cards in a single system. SLI 16X AA and SLI 8X AA should be slower than a single card running in 4X mode because of this."

We already have the problem with SLI that 256MB + 256MB does NOT equal 512MB. So when new games come out that perform better on 512MB texture memory, anyone who tries this mode will be screwed even more than he already is.

So add to it this the new problem of zero performance increase with having SLI for this mode. If this is true then why the hell does only SLI get the option? This looks more like marketing gimmick than a real feature to me.

Let's face on thing: 7800 offers zero new features over 6800 line other than the extremely minor ability to have clean fences. So the company is keen to try to exaggerate and inflate what it can with this new mode to attract more buyers to the line.
 
Is it possible to slow down current PCI-e slots? Perhaps an 8x vs 4x/2x comparison could be done to see if there is any merit to this review. I highly doubt it though, they probably forgot to change a setting or something.
 
Talonz said:
Is it possible to slow down current PCI-e slots? Perhaps an 8x vs 4x/2x comparison could be done to see if there is any merit to this review. I highly doubt it though, they probably forgot to change a setting or something.
Click to expand...

Yeah, you can block the lanes with tape or something. Starting from the front and going back to emulate a slower PCI-E slot type. I saw a test done on this on the net at somepoint when PCI-E was just comming out. The cards at the time, the 6800GT being the fastest card tested was about the same until you got down to PCI-E x4. Then it started slowing down.
 
So what the hell then? This reveiew and what he is trying to show is basically useless they way it's written.
 
I am not %100 on the data I read. I read that article about a year ago. It was however, if I recall correctly, done to show that PCI-E wasn't yet needed.

Of course, no one argues what PCI-E can ultimately do or bring to the desktop and server platforms. It was just showing that there wasn't an advantage currently over AGP.

Now, with SLi Anti-Aliasing enabled, there could be the extra need since the cards have to do more than they ordinarily would.
 
The only way I would believe those results is if they were running SLI without the bridge connector. (which the new drivers do allow) If they did that, I can see how the huge card-to-card traffic of doing sli-aa could possibly saturate the pci-e bus. But if they used the sli bridge (and thus only card-to-system traffic is hitting the real bus) then I don't see how that is possible.

I definitely remember reading that article where they tape-modded a 6800GT from 16x all the way down to 1x mode, and even at 4x it was performing almost like full speed.
 
holy crap somone taped their video card? man that's hilarious if it would work... I thought i saw a setting in my bios to set the pci-e speed at one point could be wrong though and i don't wanna reboot to check haha
 
Double Jesus said:
holy crap somone taped their video card? man that's hilarious if it would work... I thought i saw a setting in my bios to set the pci-e speed at one point could be wrong though and i don't wanna reboot to check haha
Click to expand...

It's nuts, but the way PCI-E was designed, you can install PCI-E cards in any PCI-E slot, however if you install a x16 card in a x4 slot, it can't operate at full speed, but it will work. That is why on the A8N-E, they provide a PEG retention clip, even though all that is available is a x4 socket.
 
Doesn't anyone remember the first fiasco with people unlocking non-SLI boards into SLI capability with a pencil or something? So with one PCI-E 16x and one PCI-E 1X, you could get 95% of the performance of whatever the "real" SLI board was...shouldn't that tell us something?
 
Matrices said:
Doesn't anyone remember the first fiasco with people unlocking non-SLI boards into SLI capability with a pencil or something? So with one PCI-E 16x and one PCI-E 1X, you could get 95% of the performance of whatever the "real" SLI board was...shouldn't that tell us something?
Click to expand...

Well to me that says the PCIE bus is plenty wide enough to handle any traffic being sent to it by the Video Cards. The problem is that according to how this review is posted from Hothardware, the bus seems to be lagging and overburdened at only 8x vs 16x when using SLI AA. (Which btw, I am not even sure is worth using in general since there seems to be a huge performance hit from the article.)
 
Matrices said:
Doesn't anyone remember the first fiasco with people unlocking non-SLI boards into SLI capability with a pencil or something? So with one PCI-E 16x and one PCI-E 1X, you could get 95% of the performance of whatever the "real" SLI board was...shouldn't that tell us something?
Click to expand...

Those were x4 slots, not x1 slots if I recall correctly. There was a few x1 solutions, but they were a little less than 95% effective.

SLi AA without the bridge might be the only way that bus would get saturated.
 
Well now Anandtech also finds this to be true: http://www.anandtech.com/mb/showdoc.aspx?i=2580&p=15

"It fully supports NVIDIA SLI with two approved NVIDIA based video cards in full x16 operation and with the 81.85 drivers, it will provide you with significant performance enhancements over the x8 products at higher resolutions or when utilizing SLI-AA settings."

==>Lazn
 
Lazn_Work said:
Well now Anandtech also finds this to be true: http://www.anandtech.com/mb/showdoc.aspx?i=2580&p=15

"It fully supports NVIDIA SLI with two approved NVIDIA based video cards in full x16 operation and with the 81.85 drivers, it will provide you with significant performance enhancements over the x8 products at higher resolutions or when utilizing SLI-AA settings."

==>Lazn
Click to expand...
Yeah, I'm having a hard time believing that hothardware blew things that badly. . . and I can't be sure, but a lot of people (not all, but quite a few) seems to be overlooking that the claim of performance increase is specific to SLI-AA performance (SLI-AA is not the same as regular AA that happens to be running on an SLI-enabled system). But, I can't see where Anandtech actually compared SLI-AA performance in that review! :(
 
Hurin said:
Yeah, I'm having a hard time believing that hothardware blew things that badly. . . and I can't be sure, but a lot of people (not all, but quite a few) seems to be overlooking that the claim of performance increase is specific to SLI-AA performance (SLI-AA is not the same as regular AA that happens to be running on an SLI-enabled system). But, I can't see where Anandtech actually compared SLI-AA performance in that review! :(
Click to expand...

http://www.anandtech.com/mb/showdoc.aspx?i=2580&p=10

As far as I can see, only in the first two graphs on that page.

edit:
"We benchmarked F.E.A.R. with the newly released NVIDIA 81.85 WHQL driver set, based upon recommendations from NVIDIA about further optimizations for SLI-AA and Dual Core processors that would show marked improvements for the x16 product at higher resolutions over the x8 product line. The Asus P5N32-SLI Deluxe indicated an almost 3% gain over the MSI P4N Diamond in the previous 1280x960 benchmark. As the resolutions increased in the standard AT benchmark settings, the ability of the MSI P4N Diamond with its x8 SLI configuration fell behind the Asus P5N32-SLI Deluxe by upwards of 11% in this application. Once we changed the standard benchmark settings to include 2x AA and 16X AF, the benchmarks ended up favoring the Asus P5N32-SLI Deluxe by 25%. Based upon these results, we can conclude that the additional 8GB/second of bandwidth afforded by the additional 16 PCI Express lanes and the 81.85 driver optimizations allow a great deal of headroom potential at the higher resolutions with today's hardware. "

==>Lazn
 
You must log in or register to reply here.
Back
Top