Mintmaster
Veteran
NVIDIA GF100 & Friends speculation
- Thread starter Arty
- Start date
No Catalyst 10.3...I'm sure they followed nV reviewer guidelines to the letter
No HD5970...
The GTX480 has very good performance... a Full GF100, 512 sps, +> 750MHz, X 2 TMUs(?) could match even the HD5970!
I agree.. It will be a big disappointement..
The physx on tests and demo benchmarks are a good joke.
And no power and temperature. It seems they realy folowed the nV reviewer guidelines.No Catalyst 10.3...
No HD5970...
The GTX480 has very good performance... a Full GF100, 512 sps, +> 750MHz, X 2 TMUs(?) could match even the HD5970!
What? What are you smoking? The numbers in that link are a complete wash!
Wow .. would not be surprised if AMD raised prices. Last time the GTX260 atleast kept pace with the 4870, this time round that job is being done by the GTX480.
No Catalyst 10.3...
No HD5970...
The GTX480 has very good performance... a Full GF100, 512 sps, +> 750MHz, X 2 TMUs(?) could match even the HD5970!
It's far from bad performance. Trouble is, this review tries so desperately hard at making nV look better it's not even funny anymore.
Srsly. Radeon always at the bottom of the graph, even when faster. PhysX always ON and forcing AA on Radeons in Batman: AA. Mixing DX10 with DX11 in the same graph so that even GTX285 is twice as fast as HD5870. Using 2 card SLI for nV but refusing to use HD5970. I mean, puh-leas
Mintmaster
Veteran
And Catalyst 10.2 is used. Rather crappy leak...All Physx games only benchmarked with it enabled
Can you say "nVidia PR review"?
Why?
1. 2x GTX480 for SLI benches
2. They actually got a GTX470
3. All GPU PhysX games tested with GPU PhysX enabled only
4. Catalyst 10.2 used instead of 10.3
5. No HD5970 in sight
5b. No HD5870 CF against GTX480 SLI
Also, worth noting, is that GTX285 is now a DX11 card (DiRT2
)edit:
Does anyone have link for their HD5870 review? Just for sake of some number comparisons
Almost correct, only Catalyst 10.3 missing ... and 64-bit OS/Drivers
Can you say "nVidia PR review"?
Why?
1. 2x GTX480 for SLI benches
2. They actually got a GTX470
3. All GPU PhysX games tested with GPU PhysX enabled only
4. Catalyst 10.2 used instead of 10.3
5. No HD5970 in sight
Also, worth noting, is that GTX285 is now a DX11 card (DiRT2
Actually guys use a translator and read the article
, instead of looking at the pretty graphs. They did these tests, they have the cards in house and have photos of tempurature readings of the cards, and power consumption numbers, and everyone talking about Cat 10.3, it doesn't give that much performance over 10.2 maybe in a game or two but thats it.
Last edited by a moderator:
Bad thing is that architectural slides are known to bend the reality a bit from time to time. Besides that, do you think engineers draw such shiny and stylish graphs?
Certainly not, but they do draw the "blueprints" of the chip and if those 4 big blocks exist and they were given the "GPC" name (is it by PR or by the engineers themselves), then I'm going to call them GPCs
Actually guys use a translator and read the article
Who said they didn't do the tests
What we are arguing is, this is nV reviewers guideline review to the letter.Who said they didn't do the tests
hah I have seen the press kit, no its not
Actually guys use a translator and read the article
Well, in the other leak, 10.3 actually gave LESS performance.
Plus, this is the usual preview, always with much fluff from the product's company. The Radeons DX10.1 capabilities a while back, were also tested against DX10 GeForces. I don't agree with that and neither do I agree that PhysX is enabled also for the Radeons. The PhysX tests should be done separately.
Outside of those tests everything else seems fine. Probably just ran out of time to do all the tests, and 10.3 was released when? thats why you won't see all the reviewers using them.
Certainly not, but they do draw the "blueprints" of the chip and if those 4 big blocks exist and they were given the "GPC" name (is it by PR or by the engineers themselves), then I'm going to call them GPCs
That was not his point. He was saying that there's a *lot* of stuff that is not shown on the graphical architecture slides. And others have been saying that some of that logic makes Fermi (atleast somewhat) harder to scale down than other gpu's.
And have you seen "GPC" on any of those "blueprints"? Or have you seen the blueprints at all?Certainly not, but they do draw the "blueprints" of the chip and if those 4 big blocks exist and they were given the "GPC" name (is it by PR or by the engineers themselves), then I'm going to call them GPCs
Maybe I should reiterate my question. When nvidia can combine the raster units with an arbitrary number of SMs, what actually makes a GPC a GPC?
I think it's more a "virtual" structure than a real one just invented to have something resembling somehow the old TPCs
. It would make more sense (just think of the load balancing issue between GPCs) if this structure simply does not exist in the way it is drawn in the slides.
Last edited by a moderator:
That's not actually true... especially in the DX11 workloads I've been running I've seen performance improvements north of 20%, and these are for apps that are not public and ATI could not have optimized specifically for... Still, given the time frame of those being released, that's less offensive than the PhysX on and 285 in DX11 benchmarks
I can't even interpret what the running of the DirectX SDK demos is doing... what are the numbers they are reporting there (2/card)?
Similar threads
- Deleted member 2197
- Architecture and Products
