Yeah it does have one, but it's the same as the one in R600 isn't it ? The one that suffered quite a bit of changes in Evergreen's family...
NVIDIA GF100 & Friends speculation
- Thread starter Arty
- Start date
Yeah it does have one, but it's the same as the one in R600 isn't it ? The one that suffered quite a bit of changes in Evergreen's family...
You got me there. Can you let us know what games were used in that graph mao5 posted? Because that's what we're discussing in case you didn't notice....
Eyeballing it I get 5870 = 1.64x 4890 and 480 = 1.56x 285.
I didn't known we even had established it was games being used in that chart?
Sure but that's an exception not the norm (presumably really shader heavy?). Average is more like 60%. Which isn't all that great for a chip with generally twice the number of units and more than twice the number of transistors (granted that's not entirely fair since memory bandwidth is only marginally bigger, but compared to Juniper it's also not great).
nVidia claimed 4x more FPS in their Water Demo with 1,6 billion triangles in a second or 48 million triangles in a frame.
10% would be a disaster. About 15-20% increase in clocks on cypress would make up that defict.
Considering ati has had 6 months to work on a refresh I don't see hw 10% would be a good place for nvidia to be at. 30% would be much better and simple clocks would not allow cypress to catch up.
HOpefully tho we get some price drops !
But you've picked the one which scales best - by far.
Also it's without AA actually - IIRC the performance increase it typically a bit less with AA (unless the score without AA would be cpu limited of course), probably due to the not that much increased memory bandwidth.
Maybe, but with some unreleased demo written by the vendor itself specifically written to highlight their architecture I'm not really impressed...
You can always come up with some synthetic scores easily showing one architecture blowing the other away. Just look at vantage perlin noise, and the HD5770 is nearly twice as fast as a GTX285, and that one wasn't even written by AMD, just shows the theoretical alu throughput difference...
Last edited by a moderator:
Jawed
Legend
Beware Vantage nonsense
Courtesy of an xtremesystems forum poster:
http://img686.imageshack.us/img686/6831/vei74ghz58701gb8501200c.jpg
http://forum.beyond3d.com/showpost.php?p=1410931&postcount=4345
GPU test one is 5% faster on HD5870, while GPU test two is 4% faster on GTX480. The actual GPU score is faster on HD5870.
PhsyX for the win, eh?
Jawed
Courtesy of an xtremesystems forum poster:
http://img686.imageshack.us/img686/6831/vei74ghz58701gb8501200c.jpg
http://forum.beyond3d.com/showpost.php?p=1410931&postcount=4345
GPU test one is 5% faster on HD5870, while GPU test two is 4% faster on GTX480. The actual GPU score is faster on HD5870.
PhsyX for the win, eh?
Jawed
I'm vastly more interested in actual game benchmarks.
When considering tesselation and its use in games you also have to consider that basically you have products ranging from sub $100 to $400 with the same per-clock sub division rate with the AMD products. When NVIDIA starts producing their derivative products their tesselation/geometry performance is going to be scaled back - lets say that GF104 is half the shaders engines (there is history to suggest that this may be the case) then it has half the tesselation/geometry performance, and the products further down the line have even less.
So, do we really think it makes sense for developers to solely focus on one single high-end product?
So, do we really think it makes sense for developers to solely focus on one single high-end product?
Jawed
Legend
Tessellation is conceptually the same, the API and execution model is what differs comparing the two. Oh and XB360's tessellator doesn't have the same range of amplification. Similar to the way that the consoles can't support 16384x16384 textures, they have to make do with less.
Jawed
30% over Cypress is still ~30% on average behind Hemlock
There's a lot of wiggle room between "good place" and "disaster". I don't believe it's going to be only 10%, just saying that these predictions of doom are unfounded. IIRC the 285 is only 10% faster than the 4890 and didn't earn a disaster badge.
Jawed
Legend
You made a throwaway comment "same ballpark as Cypress' (in)efficiency relative to RV770" about the performance of Cypress, just thought you needed some correction on that specific point. The performance is there, but most games aren't.
For what it's worth I expect Metro 2033 to be much faster than that on GTX480, excluding PhysX. Including PhysX I presume it'll be yet faster.
Jawed
Nvidia doesn´t have sony´s deep pockets.
285 didn't come half a year after 4890.
And there is insufficient gain. People don't get "locked in" to nVidia parts in the same way as they get locked in to a storage format.
What was throwaway about it? Reviews support it. And so does mao5's graph. You put up results from a single game as a rebuttal and that's obviously not going to work...
I considered pre-emptively addressing this because I knew someone would raise it. The consumer buying something on April 15th doesn't care how long it's been on the shelves. Prior delays don't mean squat on the purchase date. Now if that delay means the consumer is now going to wait for AMD's followup well that's a different story.
Last I heard they had ~$1.3 in the bank and a healthy profit margin, what do you base your assumption on?
Similar threads
- Deleted member 2197
- Architecture and Products
