Nvidia went SLI because they can't compete?

[Geo]

If I was going to dabble in SLI. . .

You can afford to buy (and desire to purchase) a new $4-500 videocard every year? Just to "keep up?" Incredible.

Afford? Sure, if I wanted to. Desire? Well, note the conditional in my original. As a practical matter I'm still running an AGP mobo with a single 6800GT, and because I have a bit of price/performance sensibility it'll be at least another year before I consider a PCIe mobo. As to the pricing, the conditional model I projected only had me buying the $4-500 card every second year. . .the year in between I'd be buying a $250 card (or so). But even then, I haven't done it. And probably won't unless the IHV show me significant added IQ for doing so.

But that's just me. I have no doubt at all that there are tens of thousands (and possibly low hundreds of thousands) of the hardcore cash-flush types that will drop $800-1,000 EVERY year (rather than every other) on the top-end SLI setups. SLI didn't go away originally for market reasons . . .it went away for technical reasons (AGP limitations), since removed, and now its back. The market is still there and likely always will be.

 

[Dave Baumann]

I think I also read that Nvidia has updated their SLI profiles. Anyone know of any benchies for these newly updated SLI profiles/drivers? I would be curious to see how some of those other games are benefitting from SLI (if they are at all).

I'm currently in the middle of our (long time coming) SLI article. I'll be taking a look at what profiles have been added (but only testing those in our current test suite).

 

[Frank]

They simply don't need a new part as they have a sm3.0 part on the market .

In how far is their SM3.0 functional in real life? I think it very much depends, like the SM2.0 support of the FX 5x00 series.

But the 6x00 are very nice chips and they rock under Linux.

And the most popular card seems to be the FX5200 at the moment. DX 9.0?

How about two FX 5200 in SLI, when the single one is losing it? I mean, the majority of all cards is still DX 7.0, so speed would be the important thing, right?

 

[ANova]

SLi would be good if it weren't for the fact that a single next gen card will equal or best a pair of 6800Us and have more features for an overall less cost while working with all games.

In which case, you'd pair up two next-gen cards. Why the strawman comparison of two old gen SLI cards vs next-gen. For me, the comparison is two next-gen cards vs a single next-gen card.

As for features, are you sure ATI's next card refresh is going to offer any features beyond DX9, nor any that will be exposed and taken advantage of? I'm expecting SM3.0, in which case, they'd be on par with the previous gen. Just about the only new features I can see them offering over that would be more frame buffer formats and AA, such as antialiased HDR framebuffers, anisotropic/trilinear filtered HDR textures, FP, etc

Why? Because I don't believe in blowing money. Money is hard to come by, just because you may have some doesn't mean you have to spend it like the world is ending tomorrow. Updating every year is bad enough but spending $1000 every year just on graphics is in the area of rediculous. Most people buy SLi because they use the excuse that they'll buy one card now and another later on down the line. Why spend $500 plus $200 for an SLi motherboard now, then another, say $200-$250, in a year when you can get an R600/NV60 for $500 in 1-2 years that performs as well as the two 6800Us you spent $750 on and offers WGF instead of SM3 and works with all games, unlike SLi. It just doesn't make any sense. Yeah, you'll be able to run Doom 3 at 100 fps in 16x12 and show off to all your friends for 1 year, only to be outdone the next despite spending a boatload.

If you compare two current gen with one current gen your still spending double the money for a 50% average performance increase.

 

[DemoCoder]

plus $200 for an SLi motherboard now

You don't need to buy a new SLI motherboard every year, nor is the price differential $200. Since most people need to upgrade to PCI-E from AGP, they could go straight to SLI MB for next to nothing, or whenever they next upgrade their MB. Otherwise, you should factor in the cost of an PCI-E mainboard when running the scenario "User upgrades from 9700/9800 or X800 to R600"

Your theory and comparison fails for the following reasons:

a) some people don't want to wait an indeterminate 2 years for a hypothetical next-gen part

b) the hypothetical next-gen part may not perform as well you you think. 2x performance increase every 2 years is not a law of nature.

c) $500 in two years on a top-end is more costly than $200 in two years. In other words, if I spend $500 today on a topend, then spend $200 in two years, I will have bought my SLI for a total of $700. But if I spend $500 today for a non-SLI card topend, I'd have to handout another $500 to pay for my upgrade in two years, yielding a total of $1000. The SLI user's cost amounted to $700, the $500 GPU non-SLI user today, paid $1000 by the time he is done.

But again, we are talking about the scenario of early adopters buying top-end newly released cards, which are usually uber-expensive, because of low volume and because they haven't usually been tweaked yet to increase yields and margins. The mid-range and value cards always come later (2.5 years)

I just don't see people strapped for cash buying $500 GPUs and top of the line , FX-57 or Intel EE systems. The people I see buying these systems, of which I know many, generally have real jobs, and are not working in the college bookstore.

How do you justify buying the newest, latest and greatest top-end next-gen GPU as not "blowing money", since you know, if you simply way a few months, a "refresh" tweak will be out that is far cheaper. Where is your righteous indignation and condemnation over Platinum Edition, Extreme Edition, and Ultra Edition-style products?

There are lots of purchases in life that seem to be money "blown", like ridiculously expensive cars, or hyper-expensive stereo equipment. Just because you don't have the money to throw away, doesn't mean there isn't a market for these products that people *want*

 

[jvd]

ow do you justify buying the newest, latest and greatest top-end next-gen GPU as not "blowing money", since you know, if you simply way a few months, a "refresh" tweak will be out that is far cheaper. Where is your righteous indignation and condemnation over Platinum Edition, Extreme Edition, and Ultra Edition-style products?

Refresh 10-20% speed diffrence


So 500$ most likely a single or dual slot cooling unit and great performance

vs sli

Mabye a new power supply 75-150$ for a good supply with good raills

4 slots used for the high end card .

Double the price



Some people would buy it sure , really hardcore gamers but i don't think it will be a sustainable upgrade path for anyone unless they have so much cash to burn and don't have a life to spend it on

 

[Geo]

One of the old-timer PC Magazine types had a Law he described as "The PC you Really Want costs $5,000". It held true for a long time, but look around now. And that Law started twenty years ago. . .with inflation that would look more like $7,000 now. There are quite a few people who can afford to do it (not all, of course, but the IHV's aren't counting on *all* --how many units of the top-end do you have to sell to be a significant growth addition? I suspect another 50,000 units per year at the top would get their attention) --even with $1k in gpus the overall system prices isn't all that great compared to historically, and most of those other pieces don't need to change every year. Plus, there's ebay --if you're staying at the top-end all the time you can get a significant portion (50%?) of the cost of the cards back in the resale market after one year to help finance the next round.

 

[rwolf]

How about two FX 5200 in SLI, when the single one is losing it? I mean, the majority of all cards is still DX 7.0, so speed would be the important thing, right?

Except the FX5200 is only AGP.

 

[rwolf]

I have severe doubts that NVIDIA intends to compete with ATI's next high end offering with just NV40 SLi.

I wouldn't qualify the R520 as the next high end offering for ATI, it's more of an in-between. I think ATI decided to go this route because of what you mentioned about Longhorn being delayed for at least another year and a half. Naturally it makes sense for ATI to introduce their own SM3 card since it will be lasting them quite awhile and I doubt we'll be seeing more then a 30% speed improvement. I don't expect anything from nvidia except maybe a higher clocked NV40 with 512 MB ram.

Well isn't 90nm half the size of the 130nm? They should be able to almost double the number of transistors. And the clock speed should be 550Mhz to 750Mhz. I suspect more than 30% improvement. If some of the math units on the chip are using Intrinsity's dynamic logic those parts of the chip could scale up to 4Ghz althought I doubt it would scale that high on a graphics chip.

 

[bigz]

ow do you justify buying the newest, latest and greatest top-end next-gen GPU as not "blowing money", since you know, if you simply way a few months, a "refresh" tweak will be out that is far cheaper. Where is your righteous indignation and condemnation over Platinum Edition, Extreme Edition, and Ultra Edition-style products?

Refresh 10-20% speed diffrence

ATI's last refresh yielded a 4% performance increase.

 

[BRiT]

ATI's last refresh yielded a 4% performance increase.

That depends on what section you're looking at. The "refresh" in the $200 - $300 range has produced a significantly higher performance boost [X800/X800 XL versus the R9800/X700Pro]. Hell, the X800 XL vs the X800 Pro shows a huge performance increase while being a good deal cheaper too!

Also, I beleive the last refresh of the high-end was more about availability than anything else.

 

[DemoCoder]

Refresh 10-20% speed diffrence

So 500$ most likely a single or dual slot cooling unit and great performance

We're talking about someone owning a non-SLI $500 GPU upgrading to a next-gen $500 GPU 2 years later. Total cost = $1000. Vs someone with an SLI $500 GPU upgrading to a a second GPU 2 years later when it has fallen to $200, for a total of $700.

Saying that buying a $500 non-SLI GPU every two years is somehow a vastly more affordable upgrade path is pushing it.



Someone owning a buying a $500 GPU and then waiting 2.5 years, instead of 2 years, and buying the next-gen GPU when the refreshs are out will do much better. As you can see, buying when the next-gen hits, owning the top of the line always costs you significantly more than waiting and buying the midrange.

Simply put, to own the top-of-the-line always costs you a premium, and to suggest that buying cherry picked $500-800 cards is somehow more economically rational is just dumb.

If anything, the depreciation in value of a non-SLI card is alot steeper. Your $500-800 card will lose value much more quickly as process is tweaked or supply changes, and since it has no useful lifespan when the next-gen hits, it's resale value is less than the resale value of a card which can be used to augment the primary GPU of an SLI system beyond its useful lifespan.

 

[Dave Baumann]

Why wouldn't future boards be SLI capable?

 

[ondaedg]

I think I also read that Nvidia has updated their SLI profiles. Anyone know of any benchies for these newly updated SLI profiles/drivers? I would be curious to see how some of those other games are benefitting from SLI (if they are at all).

I'm currently in the middle of our (long time coming) SLI article. I'll be taking a look at what profiles have been added (but only testing those in our current test suite).

Cool beans!

Haven't said that in a while...

 

[Ailuros]

I have severe doubts that NVIDIA intends to compete with ATI's next high end offering with just NV40 SLi.

I wouldn't qualify the R520 as the next high end offering for ATI, it's more of an in-between. I think ATI decided to go this route because of what you mentioned about Longhorn being delayed for at least another year and a half. Naturally it makes sense for ATI to introduce their own SM3 card since it will be lasting them quite awhile and I doubt we'll be seeing more then a 30% speed improvement. I don't expect anything from nvidia except maybe a higher clocked NV40 with 512 MB ram.

If you'd qualify anything as an in between release it would rather be R420; I'd rather speculate that any roadmap changes are as old and ATI did have more than one reason to cancel the original R400.

Furthermore how high can you clock a NV40 to match the estimated clockspeed of R520?

 

[Ailuros]

Democoder,

We're talking about someone owning a non-SLI $500 GPU upgrading to a next-gen $500 GPU 2 years later. Total cost = $1000. Vs someone with an SLI $500 GPU upgrading to a a second GPU 2 years later when it has fallen to $200, for a total of $700.

That line of reasoning might not be bad, but playing the devil's advocate I'd tell you that I'll sell after timespan X former high end board for Y amount, whereby I'll have chances to get to an almost equal final sum. Most high end boards have a 3 years warranty (with an expired warranty used HW is always harder to sell). Would you suggest that I would get less than let's say 250-300$ for a 6800Ultra in mid 2005 hypothetically, if I want to upgrade to a faster high end future board?

 

[trinibwoy]

Some people would buy it sure , really hardcore gamers but i don't think it will be a sustainable upgrade path for anyone unless they have so much cash to burn and don't have a life to spend it on

I wish people would stop using this argument. $1000 a year is chump change for many people (i spend 14k on rent alone ). If you're strapped for cash fine - but calling those who have the extra dough 'dumb' and 'ridiculous' doesn't really help your cause. (not directed at you jvd, just used your quote for reference)

 

[Ailuros]

Ail, if your question was about geometry scaling then it will scale in AFR mode in non-CPU limited situations (since they are working on entirely different frames). In SFR mode Transformation/Vertex Shading won't scale but will from the clipping stage (since screen space is known from here).

I'm unsure if the driver will scale geometry performance in a professional application, w/o f.e. the requirement for a special profile for those.

It would be interesting to see from someone with a SLi system, how it behaves in professional applications, when the GPUs are turned into Quadros via Rivatuner.

 

[Dave Baumann]

Errr, it'll be up to the user to decide and make the relevant profile for the application and enable the best choice for that app.

I've just noticed that 3dsmax has a predefined single board value for the standard desktop drivers.

 

[jvd]

We're talking about someone owning a non-SLI $500 GPU upgrading to a next-gen $500 GPU 2 years later. Total cost = $1000. Vs someone with an SLI $500 GPU upgrading to a a second GPU 2 years later when it has fallen to $200, for a total of $700.

Your arugement makes no sense .

You buy a geforce 6800ultra today for 500$ in 2 years you buy another 6800ultra for say 200$ That 700$ and you will get sub par performance and a sub par feature set

Vs someone who buys a 500$ gpu , sells it 2 years later for 100 and buys a new generation one .


Sli is only a good benefit if you buy it in a short period of time from the other gpu .

If you go to the lower end its still not as wise . Buy a 200$ 6600gt now , 2 years later buy a 75$ 6600gt and have 6800gt performance. Where as for the intial upfront cost of a sli mobo you could get a 6800non ultra or 6800gt . Not to mention that at the 2 year mark you could most likely get a dx10/wgf 2.0 card for 200$ that will offer more performance than the 6800ultra .

Look at last gen . The ti 4600 was the old high end card and a year later you got a 9500pro for 200$ that was faster than the 4600ti