nVIDIA's "SLI" solution

from a power consumption and heat standpoint I think these will get some AWFUL reviews from hardcore overclocking sites (especially for heat if they actually put the damn things in a case instead of leaving the MB on the benchtop).

Not even close to being a good bang/buck, even with the possibility of a future upgrade by getting a second card. Maybe some further developments will shed some light onto this but as far as I'm concerned right now this is a neat-o thing for workstation-class computers and means zilch for end/home users.

fun to watch though. :) 8)
 
trinibwoy said:
Joe DeFuria said:
The situation with Voodoo2 SLI, in contrast, was much different...every mobo, chipset, and CPU on the market would have support for more than 2 PCI slots at the introduction of V2 SLI. It was a much less risky option to "buy one now and a second one 2 later" for an upgrade path.

Well considering that this news just broke today I wouldn't be so hasty to pass judgement on it's viability.

Am I allowed to post my opinion on it?

If we see an Nforce4 in six months that supports dual PEG with a reasonable premium then I am sure lots of people will jump on this.

Define "lots." I personally don't know many people who would spend $1000+ for the graphics sub-systems. Of course there are some that will. (Again, I applaud nvidia for introducing it...the more choice we have the better).
 
trinibwoy said:
Huh? He meant that the PCIe 16x slots on all mobos are the same. How does buying one 6800 and a dual PEG mobo limit your upgrade options? It only expands them.....

Huh? You're not getting what I'm saying.

Your mobo upgrade options are going to be limited because of the dual PEG-16x slot requirements.

If you're planning on "buying one SLI capable" GeForce card now, and then "a second one" down the road, then at some point, you will have to buy a dual PEG-16 mobo. You are going to be limiting yourself to some selection of motherboards that are dual PEG-16 capable.
 
I like the whole SLI multi card concept. Its just cool.

I tell you what I Would like to see NVidia Do... If it were possible. Come up with a Way to have one card run 4x MSAA and the Other Card run 4x SSAA and merge the two images. That would just rock. Even if it was a Temporal AA Kind of thing. Sort of like an AFR thing. Card 0 does 4x FSAA on the top half of the screen on its first pass.. then Card 1 hits it with 4x SSAA the second pass and the Two alternate back and forth top/botom this way..

They are each only doing half the screen at a time Should result in decent FPS at some truely stunning AA Quality.
 
And dual PEG boards are going to lack other functionality, though perhaps only minor things--such as fewer PCI slots. You can only pack so much onto a PCB of a particular form factor.
 
Joe DeFuria said:
Your mobo upgrade options are going to be limited because of the dual PEG-16x slot requirements.

If you're planning on "buying one SLI capable" GeForce card now, and then "a second one" down the road, then at some point, you will have to buy a dual PEG-16 mobo. You are going to be limiting yourself to some selection of motherboards that are dual PEG-16 capable.

Oh that's what you meant. Firstly, I'm not going to buy a card based solely on it's SLI capability but if I did plan on going SLI and there was a hint of consumer boards I would wait for those boards so a board upgrade is a moot point. This would be a long-term decision and some planning would be involved.

I don't get the limiting yourself bit though. Are you insinuating that dual PEG-16 boards will be inferior to regular boards somehow? If I were to buy a dual-PEG board and a single GPU now what exactly is the limiting factor? A dual-PEG board *should* have all the features and support of a single slot board. If I decided to forego SLI and just get a single next-gen card I would have that option just like anyone else.

This is all hypothetical but I really do see it as being viable. I think the deciding factor will be the cost of mainstream dual-PEG boards if they ever surface.
 
John Reynolds said:
And dual PEG boards are going to lack other functionality, though perhaps only minor things--such as fewer PCI slots. You can only pack so much onto a PCB of a particular form factor.

don't forget that even a dual pci-e x16 board will cost more than a standard one even using nforce 4 chips.

Its not the chips that cost the money its the logic for running the pci-e slots and the traces . If it was free or cheap to add on dual pci-e x16 boards then everyone would be releasing them. but its not . It will be like the 4th ramslot . Much more expensive

This is cool don't get me wrong but to me it seems like a last ditch effort for bragging rights .

By the time all is said and done your going to spend at least 2k on a system like this .

I can get an athlon 64 3000+ , 1 gig of ram , 6800ultra , 200 gig hardrive , dvd rw drive , an audigy 2 zs for less than 1.5 k . and in 2 years i can then up grade to a card with dx 10 features and more performance and come in at the same price as the dual system just for the basics of the system.

Would have been nice if they did it with two chips on one board and sold it for 700ish or 800ish. At least hten it could be done in standard boards and equipment with the only upgrade being a power supply
 
Joe DeFuria said:
Am I allowed to post my opinion on it?

Your opinion is welcome as always ;)

Define "lots." I personally don't know many people who would spend $1000+ for the graphics sub-systems. Of course there are some that will. (Again, I applaud nvidia for introducing it...the more choice we have the better).

I don't know where people get these magic numbers from. I might be a little optimisitic but let's say that a dual-PEG board costs $20 more than a regular one. That is the only barrier to entry here. Nobody is forcing you to buy two cards immediately and who says you have to buy two 6800U? Personally this setup is perfect for those who don't have wads of cash to blow at once. Buy a $250 card now, buy another $250 in a year. For $20 bucks extra I think that's a great option.
 
I dunno, the way I see it if you're going to spend the money on going the SLI route I think you'd just want to go the whole 10 yards and build your whole system around it.

I'm NOT saying it's a bad thing, this is cool as hell tech to me....I just think it's definately going to be extremely not cheap.

It will be powerful as hell though, and I would love to play with it. :)
 
Quitch said:
joe emo said:
am I the only one who thinks it's extremely funny that the Geforce6800 is still rarer than an clean prostitute, and NVIDIA is already giving us the option to buy two of them at the same time? :LOL: :LOL: :LOL:

LOL!

How will this ratio method hold up to the old scanline method? Better? Worse? Sounds inferior to me, but the IP for SLI is held by someone else, right?

It's not really infringement since

1) they use "Scalable Link Interface" for SLI instead of "Scan Line Interleave

and

2) because what nVidia has is digital SLI, and not analog SLI; which an individual owns.

And I know he's not giving it up to nvidia anytime soon.

Anyways, analoge SLI is better for two or more chips on a single PCB.

So maybe this is the best nVidia can do with the "SLI" they've got.
 
I wasn't thinking infringement, but I think SLI is a little misleading considering what I'm sure springs to everyones mind... actually,I bet "Two boards working" together is what comes to mind, so scratch that :)
 
trinibwoy said:
I don't get the limiting yourself bit though. Are you insinuating that dual PEG-16 boards will be inferior to regular boards somehow?

Yes and no.

Because of the limited market, there will be fewer models of motherboards with dual slots. Think about dual CPU motherboards. They are not "inherently inferior" to single CPU motherboards, but there is a MUCH smaller variety making them:

1) More expensive
2) More likely that you'll end up comprimising one thing or another, because no motherboard meets your "needs" exactly. (Chipset you want, on-board features you want, form factor, overclock / tweakability, etc.)

A dual-PEG board *should* have all the features and support of a single slot board.

Do TODAY'S dual PEG boards have all the features and support of single slot boards? Do dual CPU boards have all the features and support of single CPU boards? How about AGP-Pro motherboards? Do they have all the "features and support" of standard AGP motherboards?

This is all hypothetical but I really do see it as being viable. I think the deciding factor will be the cost of mainstream dual-PEG boards if they ever surface.

Well, then we really don't differ that much in our opinions. ;) Just that I don't see the cost and availability of dual PEG mobos being all that great for at least a year. (I didn't say EVER.) As I said, perhaps after that things might be different. Again, at this point, I'm more interested in perhaps the NV50 generation of cards (12+ months away), and that they continue to support SLI, as I feel that if dual PEG will ever really be more than a very small niche, it won't be until then.
 
trinibwoy said:
Joe DeFuria said:
Am I allowed to post my opinion on it?

Your opinion is welcome as always ;)

Define "lots." I personally don't know many people who would spend $1000+ for the graphics sub-systems. Of course there are some that will. (Again, I applaud nvidia for introducing it...the more choice we have the better).

I don't know where people get these magic numbers from. I might be a little optimisitic but let's say that a dual-PEG board costs $20 more than a regular one. That is the only barrier to entry here. Nobody is forcing you to buy two cards immediately and who says you have to buy two 6800U? Personally this setup is perfect for those who don't have wads of cash to blow at once. Buy a $250 card now, buy another $250 in a year. For $20 bucks extra I think that's a great option.

IMO, people who don't have extra money are never better off spending more money up front for the hope of upgradeability, because more often than not that will wind up being money wasted. He would likely be better off saving the $20 or investing it in better components (faster cpu/gpu/ram) at the time of purchase, then selling off his gpu in a year and buying a new one if he wants to upgrade.
 
By the way, Tomshardware did a write up on nVidia's new SLI solution. It's a pretty good read, but things are still up in the air about actual performance figures (I don't think it'll ever reach an actual doubling of performance, but it can come close).

Anyway, did we really need The Inquirer to tell us that nVidia's next nForce chipset would support two 16x PCI Express connections? I mean, seariously, how flippin' obvious can it be?

I'd say the main reason to get a SLI (Scalable Link Interface) solution from nVidia this generation would be to make use of one of the various supersampling FSAA modes.

Edit:
Looks like Anandtech and HardOCP have their own writeups (there's probably more sites, but I didn't bother to look), and don't forget to check out Beyond3D's own front page.
 
trinibwoy said:
I might be a little optimisitic but let's say that a dual-PEG board costs $20 more than a regular one.

How much more than NO new motherboard at all?

I currently have a P4 3.2 Northwood. I may or may not want to upgrade my CPU / Mobo / Ram next year.

If I buy "with intnet to SLI" next year....I may be forced to upgrade my whole system to accomodate it, when I wouldn't be upgrading anything at all other than the latest single slot GPU otherwise.

That is the only barrier to entry here.

Yes and no. You are assuming that anyone buying a new video card next year would ALSO be in the market for a new MOBO, and that the MOBO will support his current RAM and CPU.
 
Well, Joe, keep in mind that new platforms are coming out soon, and I expect a number of enthusiasts to upgrade relatively quickly to PCI Express platforms. If the dual-PEG motherboards aren't much more expensive than the singles, there may be a lot of SLI-ready motherboards available next year.

So yes, obviously the need for a new motherboard will be a significant barrier to many, particularly since they probably won't appear until a few months after the first PCI Express motherboards, but it won't be a big barrier for everybody.

Personally, I just don't see a need for that much fillrate, so I'm not planning on getting one. I am, however, planning on purchasing a PCI Express system platform before long.
 
Joe DeFuria said:
You are assuming that anyone buying a new video card next year would ALSO be in the market for a new MOBO, and that the MOBO will support his current RAM and CPU.
You don't think that in ~6 months time the majority of people upgrading will be grabbing an LGA775/S939 PCIE board anyway?
 
Fodder said:
You don't think that in ~6 months time the majority of people upgrading will be grabbing an LGA775/S939 PCIE board anyway?

No.

Why would they?

Do the "majority" of people upgrade for the sake of upgrading? I personally tend to upgrade "my system" (CPU / memory / mobo), when the "next to most expensive" system I can buy is 2x of what I currently own. This used to be about every 18 months. CPU "innovation" seems to have slowed to a crawl the past year, and doesn't seem to show signs of qickening any time soon. When's the equivalent of a 6 Ghz CPU set to arrive?

Until then, I'll just be upgrading components on a regular basis...graphics cards, faster burners, a bump in memory capacity, etc.

I certainly don't see myself getting a new mobo / CPU / memory for at least another 18 months from today. I'll definitely be getting at least one new graphics card during that time though.

Of course, everyone's on different upgrade cycles and timing. But I don't see why the "majority" of people buying a new graphics card inthe next 6 months will ALSO be buying a new motherboad / CPU (and possibly ram.)
 
Back
Top