ATI tries to downplay SLI

[apoppin]

Originally posted by: DAPUNISHER

WTF?

ROTFL

Considering you use fifty emoticons in every post, use caps so much it's painful to read, and reiterate the same point over and over, you might want to STFU now

sorry man . . . i just never found such a collection of cliches in a single post B4 . . . had to laugh - not at the "meat" of the post; it was thoughtful as usual.

As to my "style" . . . it is "unique"

sorry to cause you such . . . "pain"

(you actually read my posts?)
( . . . why?)

:Q

 

[Rage187]

Originally posted by: DAPUNISHER
RE: Apoppin's comment above:

Yep, ATi found one hell of a bread&butter tech but to paraphrase Bilbo "it's stretched thin, like butter over too much bread" It's tapped out and no amount of spin can hide that fact. nV is leading in tech innovation now after trailing ATi for a couple years, and ATi is adopting a stratgey similar to what Intel has used in the face of AMD doing the same to them. Wave off the tech as unneeded or too soon, that is until you can get your own version out the door, then, and only then, promote it like the 2nd coming.

Ati however, doesn't have the advantages Intel does of being a household name and dominating market share. Consequently they will find themselves hard pressed to win "hearts and minds" when the time comes.

exactly!!!


ATI is trying to pull an Intel. Saying that its too early for SM3/PS3, and that its all about the Mhz.

But, just like the athlon 64, the Nv40 is not too early.


So you can either keep believing the marketing team who knows diddly, and is only trying to save face in light of new technology which they do not have. OR Look at reality and realize that the future is here and the time is now.


and the fact that the LOTR RTS is only weeks away is even better.

 

[CaiNaM]

Originally posted by: Insomniak
Because most software isn't written to USE multiple CPUs. However, SLI is a performance bump that's entirely hardware dependant. No changes to software needed.

i'll have to consider this further and get back to you.

Well, getting back to the basics, let's not forget that NV's single card solutions are every bit the best on the market, along with ATi's. You can't get better engineering at the moment. SLI is an option, not a solution. I think that's where the distinction lies. Now, whether or not it's a needed or useful option for most people is what we've been discussing here. Whilst multiple views can be taken on that issue, I think it's obvious that having the option available for the obscenely wealthy or those looking to get the best $$$ deal out of the midrange doesn't hurt.

arguably, but i'd agree they have an edge in "second-tier" high end, and certainly ati has nothing to compete with the vanilla 6800 & 6600GT (but to keep things in context this card is not out yet).

as far as engineering "at the moment", everyone seems to neglect the fact that sli is not here "at the moment" either.

further, as i stated in an earlier comment, i did indicate more options are a good thing, but that's not really the point here. what were discussing is how viable that option actually is

I think you meant earlier.

I myself am running a GeForce 4 Ti. I specifically like the idea of SLI because it gives you the "faster" with the "expensive". I again refer to my example before of picking up a 6600GT, and then throwing in another for less $$ later down the line when you need the performance boost...it's more economical than buying a 6800GT now and hoping for future proof....

why? how much is the 6600GT? let's say by the time you can get it, it's $200. let's be conservative and assume your ps is fine, and the mb is $150. that's $350; about the cost of a GT, for far less performance (the cost of a GT right now with a "deal"; i ordered a PNY for $336 plus shipping).

lets say a year from now (which would be about 6-9 months, once again with an assumption, this time that we can get a dual pcie for a reasonable price in the next 3 months) the next gen cards are out, but rather than paying top dollar for a new gen, you add a second 6600 GT for performance, at hmm.. $150 (reasonable assuming the card was $200 at launch). that's $500, when you could've had that performance or better at less the cost. how is that sensible? and who's to say a GT woudn't be avail at that time for $250-300? how smart does that $500 sli setup look now?

furthermore, too many people assume an sli will double the performance. it won't. more likely 50% performance gained for the most part, while higher in certain situations.

 

[CaiNaM]

Originally posted by: DAPUNISHER
CaiNaM,

You are talking economics when you reference how many people use older hardware. That is exactly right too, but the workstation market isn't on a tight budget and as I've suggested to you before, it makes perfect financial and marketing sense to expand the number of potential markets for your technology. They will already be manufacturing the hardware so it is savy for them to push it into the high-end gamers market *a group that is growing rapidly despite your pronuoncements about it being a tiny niche'* and the "more money than brains" crowd.

Heck, look at what an Apple costs, yet they have a core group of users that helps keep them in business, high cost of ownership not withstanding. the PC market has a decidely larger group with the financial resources to enjoy the expensive high-end hardware. Alienware is a prime example of how to maintain a successful business model by catering to those who want the fastest and are willing to pay for it. Dell is now in the same game with AW not due to some hair brained idea, but because of extensive analysis and research pointing to it as a untapped market Dell can profit by targeting Dell sees the advisability of catering to the expensive high-end crowd, why can't you?

i've stated many times this solution makes far more sense for pro 3d gfx stations than it does for gaming

 

[apoppin]

as far as engineering "at the moment", everyone seems to neglect the fact that sli is not here "at the moment" either

isn't the alienware "sli" solution "available" - now?

we do know it works as advertised . . . we are just waiting for the MBs and "bridge" from nVidia . . . and i do think just ONE mb is available at the moment that would work for a workstation . . . more are on the way for the desktop in the Nforce4 by x-mas . . . it is no longer a matter of "IF" . .. just "when".

SLI appears to give 80-90% perf increase . . . 50% is way low by most educated estimates.

 

[CaiNaM]

Originally posted by: apoppin
you are still missing the point . . . performance doesN'T double with each generation - sli will "fix" this.

r420 is not the slightest bit impressive (compared with r300/350/360) in it's INCREMENTAL approach to performance.

OTOH, nv40 IS (comparted with NV30/35) . . . and they WILL have to (nearly) double transistors to double performance . . . that's why r300-r480 has run out of gas . . .

uh huh... first of all, sli will not double performance. furthermore, i'm not sure how sli will fix cpu limitation..... i don't think it's me who's missing the point here.

 

[Insomniak]

Originally posted by: CaiNaM
why? how much is the 6600GT? let's say by the time you can get it, it's $200. let's be conservative and assume your ps is fine, and the mb is $150. that's $350; about the cost of a GT, for far less performance (the cost of a GT right now with a "deal"; i ordered a PNY for $336 plus shipping).

And the cost of a GT + Mobo is....$480 at best? Plus, with the GT pulling more power, I may need a new PSU as soon as I got it, instead of waiting to get my second 6600 and picking up a PSU just as powerful for less $$$....

Originally posted by: CaiNaM
lets say a year from now (which would be about 6-9 months, once again with an assumption, this time that we can get a dual pcie for a reasonable price in the next 3 months) the next gen cards are out, but rather than paying top dollar for a new gen, you add a second 6600 GT for performance, at hmm.. $150 (reasonable assuming the card was $200 at launch). that's $500, when you could've had that performance or better at less the cost. how is that sensible? and who's to say a GT woudn't be avail at that time for $250-300? how smart does that $500 sli setup look now?

This argument hinges on your fuzzy math that I corrected above. Yes, $500 or so dollars for the same performance (in theory) as the 6800GT, which if I'd gotten it to begin with would've run about....$500.

The difference is I get to pay that $500 over time, and if I don't need/want to pay part of it down the line, I don't have to.

People put things on payment plans all the time, and effectively, that's what SLI looks like. It's like saying "I may need all that performance some day, but for right now, I only need 50-60% of it...so I'll only buy 50 to 60% of it. If I need more later, I'll come back and pick it up. If there's another, better option, then I'll go with that and not have wasted any cash..."

Originally posted by: CaiNaM
furthermore, too many people assume an sli will double the performance. it won't. more likely 50% performance gained for the most part, while higher in certain situations.

This is probably very true, which is why I've been saying all this time that this hinges on IF what Nvidia says is true. Go back and look at my posts.

 

[CaiNaM]

Originally posted by: Insomniak
You missed the boat there.

You're not doubling the power requirements. Let me draw this picture once more:

*other blabbering*

already answered well by jiffylube

 

[Insomniak]

Originally posted by: CaiNaM

Originally posted by: Insomniak
You missed the boat there.

You're not doubling the power requirements. Let me draw this picture once more:

*other blabbering*

already answered well by jiffylube

And consequently taken care of. I would still love to hear thoughts on the mid-range and 6800NU cards, as there appear to be no barriers what-so-ever to SLI there.

 

[apoppin]

Originally posted by: CaiNaM

Originally posted by: apoppin
you are still missing the point . . . performance doesN'T double with each generation - sli will "fix" this.

r420 is not the slightest bit impressive (compared with r300/350/360) in it's INCREMENTAL approach to performance.

OTOH, nv40 IS (comparted with NV30/35) . . . and they WILL have to (nearly) double transistors to double performance . . . that's why r300-r480 has run out of gas . . .

uh huh... first of all, sli will not double performance. furthermore, i'm not sure how sli will fix cpu limitation..... i don't think it's me who's missing the point here.

well, time will tell.

and sli DOES nearly double performance (on the order of 80-90%); your 50% has no foundation. . . . some of the details of CPU bottlenecking was addressed by someone in GH more qualified than me (i'd have to search) . . . but - anyway, we expect to see dual-core CPUs next year - that'll take care of current bottlenecks.

 

[Insomniak]

Originally posted by: apoppin
and sli DOES nearly double performance (on the order of 80-90%); your 50% has no foundation. . . . some of the details of CPU bottlenecking was addressed by someone in GH more qualified than me (i'd have to search) . . . but - anyway, we expect to see dual-core CPUs next year - that'll take care of current bottlenecks.

Hmm, I dunno. Dual core CPUs won't be clocked much faster if at all...in fact, they may be clocked slower...
I don't think it'll remove bottlenecks so much as just widen them a bit...

And I'm with Cainam here, I don't see how SLI will do anything to resolve CPU limiting...

 

[apoppin]

Originally posted by: Insomniak

Originally posted by: apoppin
and sli DOES nearly double performance (on the order of 80-90%); your 50% has no foundation. . . . some of the details of CPU bottlenecking was addressed by someone in GH more qualified than me (i'd have to search) . . . but - anyway, we expect to see dual-core CPUs next year - that'll take care of current bottlenecks.

Hmm, I dunno. Dual core CPUs won't be clocked much faster if at all...in fact, they may be clocked slower...
I don't think it'll remove bottlenecks so much as just widen them a bit...

And I'm with Cainam here, I don't see how SLI will do anything to resolve CPU limiting...

but you DO (somehow) - along with Cainam - see that it will benefit gfx rendering workstations?

:roll:

 

[Insomniak]

Originally posted by: apoppin

Originally posted by: Insomniak

Originally posted by: apoppin
and sli DOES nearly double performance (on the order of 80-90%); your 50% has no foundation. . . . some of the details of CPU bottlenecking was addressed by someone in GH more qualified than me (i'd have to search) . . . but - anyway, we expect to see dual-core CPUs next year - that'll take care of current bottlenecks.

Hmm, I dunno. Dual core CPUs won't be clocked much faster if at all...in fact, they may be clocked slower...
I don't think it'll remove bottlenecks so much as just widen them a bit...

And I'm with Cainam here, I don't see how SLI will do anything to resolve CPU limiting...

but you DO (somehow) - along with Cainam - see that it will benefit gfx rendering workstations?

:roll:

Uh, yeah...

 

[lordtyranus]

I don't see much use for SLI. I had a 9700 pro for 2 years or so, and it ran games great until Farcry came out in march. I guess SLI would be an option at that point (if it could be done with 9700s), but why bother when a few months later you can get a 6800 or x800 card?

 

[Ackmed]

"Everyone" didnt get SLI in the past. A very small percentage of computer users did. Just as only a very small percenrage of computer users get a top end card, be it a X800 or 6800.

I think you mistake how small of a piece of the overall pie that we really have.

That being said, I would love the option to have SLI in my rig. I had it before with V2's, and I would again. But thats because I spend a lot on my hardware, most people dont.

He is right that it would produce a lot more heat, noise, and power problems. The V2's were fanless, and didnt put out much heat, or use much power. Things have changed. I watercool, and have what is probably the best PSU out, so its not a concern to me. But it will be for most people.

 

[DAPUNISHER]

Apoppin

I believe it would be more accurately to call them metaphors, not cliches. I used the tongue icon so you would know I'm messin' with you since we are acquainted, and I think you are as cool as the other side of the pillow The "painful" remark was purely melodrama

CaiNaM,

i've stated many times this solution makes far more sense for pro 3d gfx stations than it does for gaming

Yes sir, you have made that clear to me in the past, and while I still readily concede the point, I find it difficult to understand why you don't also see the good sense in expanding sales by expanding the the number of groups being marketed to The fact that the workstation market benefits is the primary justification for it's existance, making sales outside of that market is a sweet ancillary benefit though, don't you think?

Rage187,

I'm glad I'm not the only one who sees it that way! :beer: Now, a question for you, and anyone else with the answer. I have been watching some cool stuff on the history channel where they recreate old battles using quote "video card technology" I presume this is done thanks to the geometry instancing the LOTR game uses correct? and that the battle scenes are done on nV quadro because ATi doesn't have geometry instancing?

 

[apoppin]

Originally posted by: lordtyranus
I don't see much use for SLI. I had a 9700 pro for 2 years or so, and it ran games great until Farcry came out in march. I guess SLI would be an option at that point (if it could be done with 9700s), but why bother when a few months later you can get a 6800 or x800 card?

why?

letsee, let's say, right now you have a single sli-enabled 9700p and a new game comes out that chokes it . . .

you COULD buy a new x800 or 6800GT for $400 or spend ~$125 for another 9700p to get you by until the NEXT generation entirely whereby you start with ONE card until IT get schocked by a killer app.

:roll:

"Everyone" didnt get SLI in the past. A very small percentage of computer users did. Just as only a very small percenrage of computer users get a top end card, be it a X800 or 6800.

in the "past" there was always the likelyhood a much faster GPU would get released in 18 months that would double performance . . . with the nv40 we are getting over 220 million transistors . . . imagine the BEAST that will replace it (doubling performance - what 400 million transistors)

and the heat and the noise . . . sli becomes "cost effective" . . . soon

ANYWAY, with gfx workstations LEADING the way in sli . . . we can't be that far behind . . . and AMD and Intel will catch-up . . . even THEY are thinking multi-core (2 slower clocked cooler chips to do the 180% of the work of one hot complicated one)

DAPUNISHER, yep - you're right - on all counts - "metaphor" was the word i couldn't think of.

 

[classy]

I agree with ATI. SLI will not be the hit it was back in Voodoo days. And the V2 changed gaming. While these newer cards are faster and while they offer improved image quality, nothing to the degree of when V2 hit the scene. And your talking about $1000-1200 bucks. Only a very limited few will pay that kind of cash.

 

[CaiNaM]

Originally posted by: Insomniak
And the cost of a GT + Mobo is....$480 at best? Plus, with the GT pulling more power, I may need a new PSU as soon as I got it, instead of waiting to get my second 6600 and picking up a PSU just as powerful for less $$$....

a GT doesn't require a new motherboard.

This argument hinges on your fuzzy math that I corrected above. Yes, $500 or so dollars for the same performance (in theory) as the 6800GT, which if I'd gotten it to begin with would've run about....$500.

there's not fuzzy math above, and your correction was erroneous as it wasn't needed, nor was it relevant.

The difference is I get to pay that $500 over time, and if I don't need/want to pay part of it down the line, I don't have to.

again incorrect. sure, you can pay $500 over time, but why would you want to when you could have a helluva alot more performance now for roughly the same cost ($350). suffering with mediocre performance now in order to spend more money later for (arguably still less performance) is not something i consider a good thing.

People put things on payment plans all the time, and effectively, that's what SLI looks like. It's like saying "I may need all that performance some day, but for right now, I only need 50-60% of it...so I'll only buy 50 to 60% of it. If I need more later, I'll come back and pick it up. If there's another, better option, then I'll go with that and not have wasted any cash..."

see above.

This is probably very true, which is why I've been saying all this time that this hinges on IF what Nvidia says is true. Go back and look at my posts.

so basically ignore everything you've stated previously?

at any rate, they've already stated it's not double the performance. nvidia states (and remember, this is their OWN test, not an impartial one done by a third party) early tests have shown a 77% performance increase, but they state they'll achive closer to 90%.

as a matter of fact, they state it's the pcie bus which allows performance, when that's ridiculous as we're nowhere near saturing the agp bus. this doesn't even take advantage of the pcie bus: synchronous data transfer, as video data is only one way.

also curious as to how their "dynamic" load balancing works. since this is done in software, it will take some cpu cycles. while it's possible this could be insignificant, how will it affect games whose performance is already suffering from a cpu bottleneck? again, not necessarily a point of concern, but it's still another ananswered question - as most questions are regarding sli.

it's early, and i haven't seen anything to make me think it's a marketing ploy (and a good one), but nothing more, and while it's certainly too early to come to this conclusion, it's also too early to proclaim it as the next biggest thing in performance as many are doing.

 

[lordtyranus]

why?

letsee, let's say, right now you have a single sli-enabled 9700p and a new game comes out that chokes it . . .

you COULD buy a new x800 or 6800GT for $400 or spend ~$125 for another 9700p to get you by until the NEXT generation entirely whereby you start with ONE card until IT get schocked by a killer app.

Well, I got my 6800 GT for $300 and sold my 9700 Pro for $140. So it was a $160 upgrade. I guess I could spend $125 and get another 9700 Pro, but then I have 2 worthless cards in a couple years, whereas the 6800 GT will be a good mainstream card then.

 

[Insomniak]

Originally posted by: CaiNaM
a GT doesn't require a new motherboard.

Uh, yes it does if we're talking PCIE flavor.

And I need a new motherboard, regardless, as I'm on a two year old Socket A beast right now. It's not like I'm not in the market for one right now.

Originally posted by: CaiNaM
again incorrect. sure, you can pay $500 over time, but why would you want to when you could have a helluva alot more performance now for roughly the same cost ($350). suffering with mediocre performance now in order to spend more money later for (arguably still less performance) is not
something i consider a good thing.

What's mediocre about it? the 6600GT is supposedly 20% faster than a 9800 Pro, which itself is more than fast enough to play everything out there.

When the performance becomes mediocre is precisely when you buy the second half of your SLI setup.

Originally posted by: CaiNaM
so basically ignore everything you've stated previously?

Everything WE'VE stated. I thought it was obvious we were talking in theory here?

Originally posted by: CaiNaM
at any rate, they've already stated it's not double the performance. nvidia states (and remember, this is their OWN test, not an impartial one done by a third party) early tests have shown a 77% performance increase, but they state they'll achive closer to 90%.

Indeed, and this is the big if. If in fact they don't get the performance gains they're talking about, then SLI is all but useless to the desktop user. Workstations will love it, but we'll keep buying our single cards.

Originally posted by: CaiNaM
it's early, and i haven't seen anything to make me think it's a marketing ploy (and a good one), but nothing more, and while it's certainly too early to come to this conclusion, it's also too early to proclaim it as the next biggest thing in performance as many are doing.

Same, but speaking on the assumption (Ass U Me) that the numbers being bandied about are accurate (which they probably aren't, but who knows)....

 

[CaiNaM]

Originally posted by: Insomniak

Originally posted by: CaiNaM

Originally posted by: Insomniak
You missed the boat there.

You're not doubling the power requirements. Let me draw this picture once more:

*other blabbering*

already answered well by jiffylube

And consequently taken care of. I would still love to hear thoughts on the mid-range and 6800NU cards, as there appear to be no barriers what-so-ever to SLI there.

i dunno.. it could possibly be a "sweet spot" for sli, then again while it did well in d3, hl2 performance is lacking... and you're talking in the area of $700 and up, which would give me concern as certainly superior performance could be had in the top end card of next generation. that's how i see things as they stand now, as there are simply too many questions regarding sli technology at this point to warrant the enthusiasm many have shown. i mean, the possibility is certainly exciting, but just reading this forum i'm amazed at how so many have already crowned this the king of performance solutions...

so much of it depends on the costs and quality of the components (both for sli as well as single card solutions) avail in the future that to me, caution seems the more appropirate reaction than buying into they hype.

 

[apoppin]

Originally posted by: CaiNaM

it's early, and i haven't seen anything to make me think it's a marketing ploy (and a good one), but nothing more, and while it's certainly too early to come to this conclusion, it's also too early to proclaim it as the next biggest thing in performance as many are doing.

am just stating my opinion as you are stating yours . . . i DO think it will be the "next BIG thing" for both CPUs and GPUs . . .

we'll see (and i WILL (hopefully) be here to remind you)

here's a preemptive - "toldjaso"



i full expect to do my upgrade next year - a Nforce4 MB w/sli; A-64 CPU and a single nVidia nv50 (or cheap 6800u) . . . and i will have spent no premium for this futire gaming system.

i then have the OPTION to add a second identical GPU - at a future date - IF i choose.

Options are good.

 

[CaiNaM]

Originally posted by: apoppin

Originally posted by: CaiNaM

Originally posted by: apoppin
you are still missing the point . . . performance doesN'T double with each generation - sli will "fix" this.

r420 is not the slightest bit impressive (compared with r300/350/360) in it's INCREMENTAL approach to performance.

OTOH, nv40 IS (comparted with NV30/35) . . . and they WILL have to (nearly) double transistors to double performance . . . that's why r300-r480 has run out of gas . . .

uh huh... first of all, sli will not double performance. furthermore, i'm not sure how sli will fix cpu limitation..... i don't think it's me who's missing the point here.

well, time will tell.

and sli DOES nearly double performance (on the order of 80-90%); your 50% has no foundation. . . . some of the details of CPU bottlenecking was addressed by someone in GH more qualified than me (i'd have to search) . . . but - anyway, we expect to see dual-core CPUs next year - that'll take care of current bottlenecks.

only if you want to believe nvidia pr/hype. certainly everything they've stated in the past has held true (5800 after all was the king of gfx cards). :roll:

while it can certainly be argued i might be overly critical of sli at this early juncture, at the same time there's been nothing revealed which deserves all the adolation nv fans seem to want to adorn it. simple logic shows this is an expensive solution (at least at this point) that only makes sense in specific circumstances.

 

[TStep]

Why not promote SLI? It may not be for everyone, nor even a large percentage, but neither is $500 video card (as evident by Intel's leadership in the graphics market).

Do we all not have motherboards with open slots that allow additional memory upgrade possibilities without scrapping the memory we are currently using? Take for example the four slots on an 875 board. I like having the option of going from 2X256 to 4x256 to get my gig, rather than having to sell what I have and find some 2x512. How much performance is gained? Maybe 3% to 10% tops, correct? How much is spent? At least $100, but more than likely nearer $150 for a relatively minor gain.

If it were possible w/agp and SLI was already the norm on boards (which if pushed now, would be the case in the second gen boards) look at what that same $150 (or a wisker more) would get: a used 9800pro. Add a little overclocking and get somewhere near 9800XT levels and look at this graph of how a 9800XT and X800pro scale with the P4:

1600x1200 4XAA & 8XAF, gpu bound, since if you are anything but a benchmark freak, you will wind up being gpu bound, and the eyecandy/hi res is what these cards are all about.

X800pro 36 fps
9800xt 22 fps
theoretical sli 9800xt (at 70% efficiency) 31 fps
theoretical sli 9800xt (at 90% efficiency) 40 fps

Much more performance per dollar spent than a memory upgrade for the essentially same cash. Not everyone bought a gig of DDR a year ago, but I'm sure consideration was given to have additional slots in the motherboard to upgrade in the future. why not think in this capacity with SLI? Why not have more than one way to skin the cat, so to speak? Like everything else, a premium will be paid for the first boards with SLI tech, but if accepted, it will become mainstream. If I could do the SLI thing right now with my 9800, I would in a heartbeat