Nvidia GT200 sucessor tapes out - Interesting Artice , GT200 is so expensive to make

[largon]

(...) the 4870 single is reported to be that much faster (65-70%) than the 3870x2 atm.

Where would one find such reports that claim RV770 >> R680?

some say 96 shaders by 480sp. but looking at the early leaked die pics. it seems more like 160shaders by 480sp

Also, where are these die shots?


Hmmm...
I better quote the whole post just in case.

the graph is showing the gt200's performance relative to the 3870x2 thus in crysis the gtx 280 is projected to be 80% (.8 times) faster than the 3870x2. the reason certain benchmarks show higher than others is because the more demanding games (like crysis) have more room for improvement. so if you were gettin 40 frams in crysis on the 3870x2 they're saying you'd be getting 72 frames with the gtx280. now in cod4 the 3870x2 already has high frames so the system run will definetly bottleneck a card that is supposed to be of that magnitude. all in all it could be bad drivers on the 3870x2, but this is the performance we'd expect of a real next gen.

and the guy at xs is a moron if he thinks the 4870x2 will only be 65% faster than the 3870x2. the 4870 single is reported to be that much faster (65-70%) than the 3870x2 atm. which places the 4870 quite nicely against the gtx260 and the x2 seems like it will stomp the gtx280 but time will tell. It all depends on the number of shaders on the 4870. some say 96 shaders by 480sp. but looking at the early leaked die pics. it seems more like 160shaders by 480sp.

 

[yogurt_21]

from teh wizzard himself.

http://forums.techpowerup.com/showpost.php?p=790539&postcount=36

if he says 160, I'm most likely going to believe 160.

based on that number you have to take a look at complex shaders vs complex shaders so 160 on the rv770 vs 128 on the r680 and 64 on the rv670. so we're talking 25% more complex shaders clocked 40% faster combined with 100% more tmu's per core (so equal amount 32 on the x2 and 32 on the rv770) and faster clocked still 256-bit gddr5 (which will increase bandwidth for whatever that's worth lol) and you come up a theortetical performance increase (1.25 complex shaders times 1.4 clock speed) of 75% on the rv770 from the 3870x2. now of course this will drop to 10-15% in areas where the 3870x2 already thrives, but will likely hit that 65% mark in heavy dx10 games with the details turned up where the r680 struggles. as without the lack of complex shaders like the rv670 and r680 had, I doubt the 4870 will have trouble with aa or dx10 for that matter.

when you compare the rv770 to the rv670 you get 160 complex shaders vs 64, a 150% increase (and thats the number to watch as that was the major weak point of the 600 series architecture) 1050 MHZ shader clock vs 750MHZ, a 40% increase, 32 tmu's vs 16, 100% increase you get a rather inflated theoretical performance increase of 700% lol (2.5 complex shaders times 1.4 clock speed times 2 tmus) which will probabaly become more like 2-2.5 times more performance average (50% faster in titles the rv670 already thrives in, 150% faster in areas it's performance is abysmal)

not sure why you wanted to make sure to quote my whole post, I'm not about to edit anything. this is all speculation and theory at this point, so my theory stands just as anybody elses will until actual release. we'll see where the rubber hits the road. but the theoretical performance has already been had out on the 600 series so saying a 65% increase over the x2 doesn't seem too far fetched to me based on the numbers i'm seeing. which like I said places it nicely against the gtx260 and makes the r700 look quite nice against the gtx280 (but at this point in the specs it appears the 4870x2 will be heavily bottlenecked and suffer from driver problems in the begining)

@people who don't understand the graph
for the last time the 3870x2 is being used as a benchmark (look it up it doesn't mean 3dmarks lol) thus it's performance will always be 1 in the graphe because its performance is being used as the control for their tests. so you take anything on top and compare based on that lol

 

[imperialreign]

Crysis uses Cryengine 2, which nv had boosted spending many hours working with
Oblivion uses Gamebryo. I still don't see this engine used in newer titles
COD4 uses COD4 engine
COH uses Essence 1.0 engine
Prey & ET:QW uses id Tech 4 (found in games like Doom, Quake 4, etc)
Fear uses the Lithtech: Jupiter EX (haven't seen this engine used since 2002 or so)
UT3: U3 engine
WiC uses Masstech Game Engine

hmm, I found some of this odd in the graph earlier as well - one of the first tip offs for me that the graph was fabricated was the results for FEAR and Call of Juarez

Reason being, Call of Juarez is one of the few games where ATI worked with the developers, and we've routinelly seen ATI cards staying ahead of nVidia's, even with DX10 - hell, a 3870 can hang neck and neck with an 8800 GTX; and crossfire performance with this game also tends to tromp SLI setups (I still wonder why this game isn't included more often with new hardware reviews . . .).

Also, FEAR is another one of the few games where ATI did work with the devlopers as well, and although it's a TWIMTBP game, ATI cards have stayed on par with nVidia's, both companies offering negligible performance differences over the other.


I mean, if you're going to attempt to fabricate a chart like that, at least try to make it look reasonable across the board.

 

[EastCoasthandle]

hmm, I found some of this odd in the graph earlier as well - one of the first tip offs for me that the graph was fabricated was the results for FEAR and Call of Juarez

Reason being, Call of Juarez is one of the few games where ATI worked with the developers, and we've routinelly seen ATI cards staying ahead of nVidia's, even with DX10 - hell, a 3870 can hang neck and neck with an 8800 GTX; and crossfire performance with this game also tends to tromp SLI setups (I still wonder why this game isn't included more often with new hardware reviews . . .).

Also, FEAR is another one of the few games where ATI did work with the devlopers as well, and although it's a TWIMTBP game, ATI cards have stayed on par with nVidia's, both companies offering negligible performance differences over the other.


I mean, if you're going to attempt to fabricate a chart like that, at least try to make it look reasonable across the board.

Pretty much my thoughts on the situation. We gotta take these charts with a grain of salt (or 2) even if it did come from nv.

 

[magibeg]

If the charts did come from NV thats even more reason not to trust them. Companies are only going to show their cards in the best possible conditions against their competition. I'm pretty sure theres a conflict of interest when it comes to reviewing your own stuff.

 

[Deleted member 3]

An article by the Inq with five links to other articles by the Inq. No other sources. Interesting.

 

[Deleted member 3]

Again, take this with a grain of salt. I believe this is photoshoped.

There is a good possibility that they use Photoshop to make graphs. At TPU we do it as well.

I think what you mean to say is that the numbers are made up.

 

[Megasty]

I guess they were trying to use the 3870x2 as a baseline. But that still don't make any GD sense because it appears that the 3870x2 does the same in all games. They're thinking too much :shadedshu

The numbers are fecked up & the graph is completely out of proportion anyway - like you can get 400+ fps in COJ I bet they didn't even test the cards themselves

After looking at this BS I think I'll try that 360 rumblepad too

 

[magibeg]

The numbers are fecked up & the graph is completely out of proportion anyway - like you can get 400+ fps in COJ I bet they didn't even test the cards themselves

Well thats the thing though. We have no idea what settings they were using or anything. So its most likely the best possible settings for their cards and the worst possible settings for ATI. Like if you crank the settings just right and at 2560x1600 with everything maxed to high hell and the 3870x2 gets 5fps and the GTX280 gets 12fps thats a big difference on the graph.

 

[eidairaman1]

Cyrsis is the one game that apparently has Problems over the Resolution of 1920x1200, on Dual Videocards from either Company, Nvidia Claims a certain FPS at a Certain Resolution but after that top is when the card starts losing performance.

 

[DarkMatter]

It's amazing how only like ONE person noticed the chart shows relative performance, and thus the X2 has always a value of one.

COD4, UT3 and FEAR have already extremely high frames on current midrange 8800 and HD3000s, and are already bottlenecked by the platform. That's why the improvement over the X2 is smaller. This has already been explained, but people has insisted in the thing so...

Also I don't think the results are fabricated, but also I am not saying Nvidia has not used the best faborable scenarios.

 

[[I.R.A]_FBi]

the worst thing about the chart i dunno if af and aa are on, what res, nothing .. corporate horseshit.

 

[Megasty]

the worst thing about the chart i dunno if af and aa are on, what res, nothing .. corporate horseshit.

Would that corporate horseshit be thoroughbred or donkey

I prefer the donkeyshit cause the thoroughbred variety would just end up costing more The darn 280 is already near $650 so I don't know if it can pull any more horseshit out of its ass if you would call it a thoroughbred :shadedshu

But I do think that nvidia used the worse case scenarios for the card tests such as the highest res & 3-4 fps results. Results ending up with garbage going off the charts tells all

 

[HTC]

Dunno if this has been posted before or not:

GTX280 CRYSIS 1920*1200 VH Average FPS Reached to 36.81!

Japan IT Media website today brings us the CRYSIS 1920*1200 VH test result of NVIDIA next generation flagship -GeForce GTX 280 Graphics Card.

According to IT Media said, NVIDIA and an anonymous motherboard manufacturer hold a secret presentation to show the performance of GTX 280 outside the Computex 2008.

The visitors said that the demonstration room is very dim lighting. In addition to show the performance of GTX 280 graphics card, the secret presentation also shown parts of the motherboards which are compatible with GTX 280, they simply had been placed on the windowsill of the room.

IT Media site had the opportunity to run GPU-Z, CPU-Z and Crysis Benchmark on the GTX 280 demo system. From the photos, we can clearly see that NVIDIA GTX 280 presentation system used Intel Core 2 Quad four-core processor, the frequency is 2.66GHz, the Crysis Benchmark with 1920 x1200 VeryHigh settings indicated that the average fps of GTX 280 graphics card reached 36.81!

Source: PCzilla but originally found @ PCDig@ (Portuguese site).

EDIT

Had posted this in the wrong topic, by mistake!

 

[wolf2009]

impressive !

 

[Megasty]

impressive !

That's an understatement. That's damn good

But why wouldn't they give actual gaming frames instead of that benchmark. Even my system is a bit faster in the benchmark than the actual game

Anyway, it looks sweet don't it: http://www.vr-zone.com/articles/Detailed_Geforce_GTX_280_Pictures/5826.html

 

[DarkMatter]

That's an understatement. That's damn good

But why wouldn't they give actual gaming frames instead of that benchmark. Even my system is a bit faster in the benchmark than the actual game

Anyway, it looks sweet don't it: http://www.vr-zone.com/articles/Detailed_Geforce_GTX_280_Pictures/5826.html

She's sexy.

 

[newconroer]

Ah, always pleasant to see good folk like the Inquirer, continuing their pilgrimmage to bloat the airwaves with nonsense before products are even released - only to slink back under their rock once those same products hit the market.

 

[eidairaman1]

Ah, always pleasant to see good folk like the Inquirer, continuing their pilgrimmage to bloat the airwaves with nonsense before products are even released - only to slink back under their rock once those same products hit the market.

Inquierer is a Rumor Finder, thats all they are, its funny even you take them seriously.

 

[newconroer]

They take themselves seriously, that's enough.