GT300 Benchmarks

[Wreckage]

Originally posted by: BFG10K
Twice as fast as a 5870? That would mean it?d have to be about three times as fast as a GTX285, but the specs don?t really allow that. Hmm, I smell bullshit.

Why don't the specs allow that? It's an entirely new architecture using MIMD. It could well exceed those benchmarks

Odd.

 

[ilkhan]

Originally posted by: Grooveriding
I think we all know it's a given the GT300 will be faster than a 5870. Don't kid yourself it will be twice as fast though.. We'll likely see exactly what we saw with 4890 vs 285. GT300 will probably come in around 10-20% faster depending on the game over 5870.

pretty much.

The raw 512/240 vs 1600/800 tells us that. Id say these results are plausible, regardless of their authenticity. Playing 1920x1200 maxed out in crysis is about right for this generation, right in line with those GT300 numbers at 1680.

 

[Hauk]

Originally posted by: ZimZum
There is a also a clip on youtube of Godzilla fighting Mothra where can actually see the sneakers of the guy in the Mothra suit. That video is still far more believable than this one.

:laugh:

 

[MagickMan]

So, the NV guys are saying it's 2x faster and the ATI guys are whipped up into a frenzy saying there's now way and the info is fake.


My prediction? The top end GT300 will be 40-60% faster than a 5870, the mid-range will be 20-30% faster, and if there's a low end it'll just about match it. All GT300s will be pricey, probably $500, $400, and $300 at each price point. It'll be a great little war and I look forward to seeing the carnage!

 

[LCD123]

ATI will just come out with the hd5870x2 or hd5890 to match that GT300

 

[IlllI]

i don't doubt that GT300 will be faster, but at what cost? i mean price-wise

 

[MagickMan]

Originally posted by: LCD123
ATI will just come out with the hd5870x2 or hd5890 to match that GT300

I know, it'll be great. We need the war to continue!

 

[T2k]

Jesus in heaven,

what kind of newbie-wannabe-shill-zoo this forum has become?

Seriously, people, look at these hilarious posts:

Originally posted by: Wreckage
http://www.youtube.com/watch?v=10xKBSEEvwQ

Not sure how accurate those benchmarks are, but if they are true. :shocked:

Originally posted by: nitromullet
The results actually are probably pretty plausible IMO. I'm sure NVIDIA will put out a card significantly faster then the 5870, it's a matter of when and for how much $$$.

Originally posted by: Grooveriding
I think we all know it's a given the GT300 will be faster than a 5870. Don't kid yourself it will be twice as fast though.. We'll likely see exactly what we saw with 4890 vs 285. GT300 will probably come in around 10-20% faster depending on the game over 5870.

Originally posted by: Wreckage

Originally posted by: BFG10K
Twice as fast as a 5870? That would mean it?d have to be about three times as fast as a GTX285, but the specs don?t really allow that. Hmm, I smell bullshit.

Why don't the specs allow that? It's an entirely new architecture using MIMD. It could well exceed those benchmarks

Odd.

Originally posted by: ilkhan

Originally posted by: Grooveriding
I think we all know it's a given the GT300 will be faster than a 5870. Don't kid yourself it will be twice as fast though.. We'll likely see exactly what we saw with 4890 vs 285. GT300 will probably come in around 10-20% faster depending on the game over 5870.

pretty much.

The raw 512/240 vs 1600/800 tells us that. Id say these results are plausible, regardless of their authenticity. Playing 1920x1200 maxed out in crysis is about right for this generation, right in line with those GT300 numbers at 1680.

Originally posted by: MagickMan
So, the NV guys are saying it's 2x faster and the ATI guys are whipped up into a frenzy saying there's now way and the info is fake.


My prediction? The top end GT300 will be 40-60% faster than a 5870, the mid-range will be 20-30% faster, and if there's a low end it'll just about match it. All GT300s will be pricey, probably $500, $400, and $300 at each price point. It'll be a great little war and I look forward to seeing the carnage!

ABSOLUTELY
P-R-I-C-E-L-E-S-S
posts.

 

[alyarb]

Originally posted by: MagickMan
So, the NV guys are saying it's 2x faster and the ATI guys are whipped up into a frenzy saying there's now way and the info is fake.

dude, you are way out of line saying that each of the ~two dozen (so far) skeptics who have responded to this blatantly unsubstantiated show of numbers superimposed upon stock photographs accompanied by a metal soundtrack bearing no resemblance whatsoever to any nVIDIA presentation ever (including a retarded personalized intro animation) are each and every one of them, an ATI fanboy. this place isn't that exreme politically, and to claim that everyone here is irrationally loyal to ATI, in this thread is to imply that the video is credible and from a credible source who has shown to be reliable in the past. This video really is crap, by any standards. If nvidia really did release a video like this, i'd probably be thinking "they sure gayed everything up for GT300 didn't they" in my head while watching it. Everybody's been through this. nvidia demoed a fake card at a presentation less than a month ago. Everybody was skeptical then and the same group remains that way. Yes, some fanboys have posted in this thread, who cares. Coming in and just throwing that card at everyone is even lamer. Anyone in their right mind can see that this video is fake.

We aren't saying the GT300 won't be significantly faster; it almost certainly will based on what we already know. That's not the point. The point is that this video is merely a poorly executed optimism based on what little preliminary data we have, and the quality of the production and time of its release make that obvious to those who have experience reading bullshit all day on the internet. It is also obvious to us that your comment is prejudicial and uncalled for. Nobody is in a frenzy (and if someone is, he isn't in any more of a frenzy than he is normally) and the data in this video is most likely fake and far less likely to have, by any other means than luck, any correspondence whatsoever to the actual behavior of the consumer GT300 cards, particularly on a machine of undisclosed specifications.

 

[T2k]

I'm with you, alyarb - except this one:

We aren't saying the GT300 won't be significantly faster; it almost certainly will based on what we already know.

This is pretty far from certain, more likely not true at all, given that the extra trannies in GT300 went for purely GPGPU purposes.

As a matter of fact I expect that first GT300 gaming results will be pretty disappointing - remember, 58050X2/5870X2 will be on the market for a couple of months by then...

 

[Ben90]

Dude screw this lame GT300 crap, i just aquired myself a leaked GTX 480 (yea bet u didnt even know they were releasing this for like another year)

Anyways basic rundown:
GPU: GT400
Release Date: November 20th 2009
MSRP: like 50ish bucks
Idle Power Usage: 2.2 watts
Max Power Usage: 8 watts

Benchmarks:
Top Secret Benchmark Leaks

 

[MagickMan]

Originally posted by: T2k
Jesus in heaven,

what kind of newbie-wannabe-shill-zoo this forum has become?


ABSOLUTELY
P-R-I-C-E-L-E-S-S
posts.

I dunno, the ATI shills are screaming like the NV fanboys are stealing their candy. You gonna cry us a river too?

I'm not on either "side", I just want the fighting between ATI and NV to escalate.

 

[T2k]

Originally posted by: MagickMan

Originally posted by: T2k
Jesus in heaven,

what kind of newbie-wannabe-shill-zoo this forum has become?


ABSOLUTELY
P-R-I-C-E-L-E-S-S
posts.

I dunno, the ATI shills are screaming like the NV fanboys are stealing their candy. You gonna cry us a river too?

I'm not on either "side", I just want the fighting between ATI and NV to escalate.

Dude,

your comment was among my examples - do I have to say anything more?

 

[MagickMan]

Originally posted by: alyarb

Originally posted by: MagickMan
So, the NV guys are saying it's 2x faster and the ATI guys are whipped up into a frenzy saying there's now way and the info is fake.

dude, you are way out of line saying that each of the ~two dozen (so far) skeptics who have responded to this blatantly unsubstantiated show of numbers superimposed upon stock photographs accompanied by a metal soundtrack bearing no resemblance whatsoever to any nVIDIA presentation ever (including a retarded personalized intro animation) are each and every one of them, an ATI fanboy. this place isn't that exreme politically, and to claim that everyone here is irrationally loyal to ATI, in this thread is to imply that the video is credible and from a credible source who has shown to be reliable in the past. This video really is crap, by any standards. If nvidia really did release a video like this, i'd probably be thinking "they sure gayed everything up for GT300 didn't they" in my head while watching it. Everybody's been through this. nvidia demoed a fake card at a presentation less than a month ago. Everybody was skeptical then and the same group remains that way. Yes, some fanboys have posted in this thread, who cares. Coming in and just throwing that card at everyone is even lamer. Anyone in their right mind can see that this video is fake.

We aren't saying the GT300 won't be significantly faster; it almost certainly will based on what we already know. That's not the point. The point is that this video is merely a poorly executed optimism based on what little preliminary data we have, and the quality of the production and time of its release make that obvious to those who have experience reading bullshit all day on the internet. It is also obvious to us that your comment is prejudicial and uncalled. Nobody is in a frenzy (and if someone is, he isn't in any more of a frenzy than he is normally) and the data in this video is most likely fake and far less likely to have, by any other means than luck, any correspondence whatsoever to the actual behavior of the consumer GT300 cards, particularly on a machine of undisclosed specifications.

You should pray to God and sunny Jesus that it's true, it can only be good for the PC gaming community if NV comes out swinging.

 

[MagickMan]

Originally posted by: T2k

Originally posted by: MagickMan

Originally posted by: T2k
Jesus in heaven,

what kind of newbie-wannabe-shill-zoo this forum has become?


ABSOLUTELY
P-R-I-C-E-L-E-S-S
posts.

I dunno, the ATI shills are screaming like the NV fanboys are stealing their candy. You gonna cry us a river too?

I'm not on either "side", I just want the fighting between ATI and NV to escalate.

Dude,

yours were one of the posts - do I have to say anything more?

Duh? Go back and read my posts. :roll:

 

[solofly]

Originally posted by: ShadowOfMyself
Wreckage probably made the video himself under another name or something

HAHAHAHAHAHAHAHAHAHAHAHAHHAJHAHAHAHAHAHAHAHAHA!!!

This line made my day...

 

[alyarb]

Originally posted by: T2k
I'm with you, alyarb - except this one:

We aren't saying the GT300 won't be significantly faster; it almost certainly will based on what we already know.

This is pretty far from certain, more likely not true at all, given that the extra trannies in GT300 went for purely GPGPU purposes.

As a matter of fact I expect that first GT300 gaming results will be pretty disappointing - remember, 58050X2/5870X2 will be on the market for a couple of months by then...

i'm not saying the card will be 60% faster like in this video, but to say that it's not certain is to say that GT300 is likely to be slower than 5870 despite having more GPU power and faster memory. i'd say 40% max, any faster than that wouldn't really make much sense unless the thread handling techniques are unimaginably improved such that linear performance-per-shader could be achievable (and even this wouldn't make MagickMan's dream come true), and we really haven't heard much other than the number of shaders is 512, and that the cache is really quite large. is the 384-bit bus even confirmed?

 

[extra]

I predict that it will be 15-33% faster than the 5870 and cost approx 33% more and whether or not it will be worth buying will depend on the games you play. At least, this seems plausible from the supposed specs. It's also entirely possible it will lose to the 5870 in some games though. Nvidia better hurry up and get it released though they are getting their asses kicked right now clear across the board. If they don't have at least partial availability by Christmas it's going to be pretty rough for them.

 

[T2k]

Originally posted by: alyarb

Originally posted by: T2k
I'm with you, alyarb - except this one:

We aren't saying the GT300 won't be significantly faster; it almost certainly will based on what we already know.

This is pretty far from certain, more likely not true at all, given that the extra trannies in GT300 went for purely GPGPU purposes.

As a matter of fact I expect that first GT300 gaming results will be pretty disappointing - remember, 58050X2/5870X2 will be on the market for a couple of months by then...

i'm not saying the card will be 60% faster like in this video, but to say that it's not certain is to say that GT300 is likely to be slower than 5870 despite having more GPU power and faster memory. i'd say 40% max, any faster than that wouldn't really make much sense unless the thread handling techniques are unimaginably improved such that linear performance-per-shader could be achievable (and even this wouldn't make MagickMan's dream come true), and we really haven't heard much other than the number of shaders is 512. is the 384-bit bus even confirmed?

So far nothing is confirmed except one thing: Nvidia suddenly did a 180 and started talking about nonexisting, future product(s), despite its long-held opposite stance.

 

[T2k]

Originally posted by: extra
I predict that it will be 15-33% faster than the 5870 and cost approx 33% more and whether or not it will be worth buying will depend on the games you play. At least, this seems plausible from the supposed specs. It's also entirely possible it will lose to the 5870 in some games though. Nvidia better hurry up and get it released though they are getting their asses kicked right now clear across the board. If they don't have at least partial availability by Christmas it's going to be pretty rough for them.

Dude,

where have you been? Except Nvidia's full admission it's been reported by AT, H and several others that NV has literally no chance to launch it (real, en masse shipping, not some vaporlaunch with 50 preproduction review cards) before early next year...


...it's going to be a very tough year for NV, I'm saying this for a while now.

 

[Grooveriding]

Originally posted by: Nemesis 1

Originally posted by: Grooveriding
I think we all know it's a given the GT300 will be faster than a 5870. Don't kid yourself it will be twice as fast though.. We'll likely see exactly what we saw with 4890 vs 285. GT300 will probably come in around 10-20% faster depending on the game over 5870.

OH really got proof . Larabee will be 3x what 385 is . LOL do i need proof . NA I said so it has to be true LOL.

I'm not some raving fanboy type. I just went from a GTX 285 to 5870 CF. Before my 285, I had a 4870.

I just think it's a safe bet nvidia will have a single gpu faster than 5870, after all, they are going to release well after ati has.. if they have any sense, they'll make sure it is faster by some margin.

It's great to have video cards with more power, but I feel we have more than enough right now, unless you have a 30" monitor, there is nothing you can't handle now at 1920x1200.

Even with the single 5870, I could run Crysis 1920x1200 4xaa everything enthusiast and it felt smooth enough at 35 or so fps.. now at 48 or so it's even better. Clear Sky is also smooth at the same settings.

There are no new game engines on the horizon that look to rival cryengine 2. So really, unless you have a 30" screen, there is enough power available in gpus now to play any game at it's highest settings smoothly.

Guess it is time to get a 30" monitor...

 

[alyarb]

no, there is a lot more than nothing that is confirmed with the architecture.
http://www.nvidia.com/content/...itectureWhitepaper.pdf

the dual warp scheduler dispatches double the number of instructions per clock and there are now two of them per SM, and they support out of order execution. The L1 cache (one per SM which is one per 32 shaders now) is four times as large as GT200, the L2 cache is totally new and also very big considering that it is shared globally and 20x lower latency than a 150GB/s memory bus, and of course, the number of shaders is more than doubled than GT200. turns out the 384-bit bus is also confirmed. I think it's clear that this card is going to be faster. These improvements are not exclusively for the benefit of GPGPU, not by a long shot. It was strategic for them to paper launch their GPGPU product before the consumer product. Wait until we hear about ROP/TMU.

 

[T2k]

Originally posted by: alyarb
no, there is a lot more than nothing that is confirmed with the architecture.
http://www.nvidia.com/content/...itectureWhitepaper.pdf

the dual warp scheduler dispatches double the number of instructions per clock and there are now two of them per SM, and they support out of order execution. The L1 cache (one per SM which is one per 32 shaders now) is four times as large as GT200, the L2 cache is totally new and also very big considering that it is shared globally and 20x lower latency than a 150GB/s memory bus, and of course, the number of shaders is more than doubled than GT200. turns out the 384-bit bus is also confirmed. I think it's clear that this card is going to be faster. These improvements are not exclusively for the benefit of GPGPU, not by a long shot. It was strategic for them to paper launch their GPGPU product before the consumer product. Wait until we hear about ROP/TMU.

Well, sorry to rain your parade but they have yet to show a working piece.
Until then it's all PR/paperware, sorry.

 

[novasatori]

Originally posted by: T2k

Originally posted by: alyarb
no, there is a lot more than nothing that is confirmed with the architecture.
http://www.nvidia.com/content/...itectureWhitepaper.pdf

the dual warp scheduler dispatches double the number of instructions per clock and there are now two of them per SM, and they support out of order execution. The L1 cache (one per SM which is one per 32 shaders now) is four times as large as GT200, the L2 cache is totally new and also very big considering that it is shared globally and 20x lower latency than a 150GB/s memory bus, and of course, the number of shaders is more than doubled than GT200. turns out the 384-bit bus is also confirmed. I think it's clear that this card is going to be faster. These improvements are not exclusively for the benefit of GPGPU, not by a long shot. It was strategic for them to paper launch their GPGPU product before the consumer product. Wait until we hear about ROP/TMU.

Well, sorry to rain your parade but they have yet to show a working piece.
Until then it's all PR/paperware, sorry.

http://dictionary.reference.com/browse/opinion
http://dictionary.reference.com/browse/speculation

 

[alyarb]

Originally posted by: T2k

Originally posted by: alyarb
no, there is a lot more than nothing that is confirmed with the architecture.
http://www.nvidia.com/content/...itectureWhitepaper.pdf

the dual warp scheduler dispatches double the number of instructions per clock and there are now two of them per SM, and they support out of order execution. The L1 cache (one per SM which is one per 32 shaders now) is four times as large as GT200, the L2 cache is totally new and also very big considering that it is shared globally and 20x lower latency than a 150GB/s memory bus, and of course, the number of shaders is more than doubled than GT200. turns out the 384-bit bus is also confirmed. I think it's clear that this card is going to be faster. These improvements are not exclusively for the benefit of GPGPU, not by a long shot. It was strategic for them to paper launch their GPGPU product before the consumer product. Wait until we hear about ROP/TMU.

Well, sorry to rain your parade but they have yet to show a working piece.
Until then it's all PR/paperware, sorry.

it's not a parade (again why does everyone have to throw the fanboy card?), i'm just rehashing the facts for us to consider in the context of this thread, or in the context of what this thread is likely to become.

whether there are working cards right now or not doesn't matter. nvidia is not going to go back on their word and release an architecture that doesn't do what they say it does (especially in the Tesla whitepaper), nor are they going to hype up their new product and then go out of business. they are going to release a new GPU sooner or later and it's going behave like a GPU of specifications mentioned in the whitepaper would behave. i'm not promising abstract things like GT300 will "nail AMD to the wall and paint them with turds" or "bring your gameplay to new levels of realism" or "visualize your world in a PhsX orgy." those are things for wreckage to promise when he rebuts all the skeptic posts in the morning. i'm talking about discrete instruction handling and throughput as it pertains to plain old performance. the details i mentioned are improvements to what was otherwise last generation's architecture (disregarding the MUL unit), taken right from the pdf. it is highly unlikely that the architecture will be a step backward in any way (has this ever happened to any GPU?) just as it is unlikely that all aspects of performance will go up 100% or even 70%. If you don't believe something as concrete as dual warp scheduling taken from a source pdf then you may as well not believe the card is Dx11 or OpenCL compliant either. You may as well not believe anything.