next gen kepler to support dx 11.1, also take a year to rollout all cards

[Rollo]

Originally Posted by shadow001

This is from Vr-zone about 3 months ago, and pray it isn't accurate at all






We see where the new chips slot in the performance scale of things relative to the old generation, and what comes out first isn't even the GK104 but lower end parts first, and the GK110 is basically a dual GPU GK104 based card, with the real high end only being released in early 2013....Pray that this also isn't accurate at all and that VR-zone is talking bull**** too, just like charlie is as well.


http://vr-zone.com/articles/report-n...map/14067.html

Seriously? Pray? LOL. I don't care if Nvidia releases a part this year beyond I like to try new video cards. My 3gb 580s give me 80% of what ATi's new pooch could give me in framerate, runs my games fine.

[shadow001]

Originally Posted by Rollo

Seriously? Pray? LOL. I don't care if Nvidia releases a part this year beyond I like to try new video cards. My 3gb 580s give me 80% of what ATi's new pooch could give me in framerate, runs my games fine.

I'm happy that you're happy, but it means that AMD owns all of 2012 if that chart is correct...Remember your constant argument about market share and who has more of it, so you can guess what it'll do to it if they come second best for the entire year.


And the fastest dual GPU HD7990 has yet to be released, and if that chart is accurate(big if here), what will Nvidia have to counter that monster?, single card versus single card...Nothing.

[ninelven]

Originally Posted by shadow001

The theoretical maximum that the GF110 is capable of is 37.06 Gigapixels/sec

No, unless you are using AA, the max for GF110 is 24.83 Gigapixels/sec. Tahiti is 29.6 GP/sec by the same math.

So who's more efficient and does this count as supporting evidence enough?

No. It doesn't.

The numbers are 9.75 for Fermi and 13.3 for Tahiti. Why isn't Tahiti getting 29.6? Because the test is bandwidth limited. Now, the 7970 has 37% more bandwidth than the 580, so if the test is bandwidth limited we should expect the 7970 to perform 37% better than the 580. 13.3 / 9.75 = 1.364. SHOCKING. The 7970's advantage over the 580 is exactly equal to its bandwidth advantage in a bandwidth limited test... imagine that.

This is the third time I have explained the result, and I really don't think I can make it any simpler or more clear. If at this point you can not understand that the test is completely bandwidth limited, then I must conclude you are either 1) intentionally trolling and not worth having a conversation with or 2) genuinely intellectually incapable of understanding and not worth having a conversation with.

[shadow001]

Originally Posted by ninelven

No, unless you are using AA, the max for GF110 is 24.83 Gigapixels/sec. Tahiti is 29.6 GP/sec by the same math.


No. It doesn't.

The numbers are 9.75 for Fermi and 13.3 for Tahiti. Why isn't Tahiti getting 29.6? Because the test is bandwidth limited. Now, the 7970 has 37% more bandwidth than the 580, so if the test is bandwidth limited we should expect the 7970 to perform 37% better than the 580. 13.3 / 9.75 = 1.364. SHOCKING. The 7970's advantage over the 580 is exactly equal to its bandwidth advantage in a bandwidth limited test... imagine that.

This is the third time I have explained the result, and I really don't think I can make it any simpler or more clear. If at this point you can not understand that the test is completely bandwidth limited, then I must conclude you are either 1) intentionally trolling and not worth having a conversation with or 2) genuinely intellectually incapable of understanding and not worth having a conversation with.

The test wasn't using AA at all, so that doesn't come into it to begin with, and i fully know that the test is bandwith limited, but it does to show that theoretical maximums mean nothing if it can't be delivered in real terms.


To make matters worse, you stated that AA use doesn't drop the theoretical maximum fillrate on the HD7970, while it does on the GF110 to 24 Gigapixels so score one more point for AMD there, and above all else, those 48 rops on the GF110 might actually take up more room on the die than the 32 used in the HD7970.


It's about efficiency in the end so adding more of them in kepler is pointless as it just takes up more die space and there isn't enough memory bandwith to feed them all anyhow....Add fewer and make them more flexible in not dropping fillrate when AA is used like you mentioned for the HD7970.

[ViN86]

Thread cleaned. Keep it on topic and don't resort to name calling ('kid', 'asshole', etc.) and it will stay open.

[ninelven]

Originally Posted by shadow001

and i fully know that the test is bandwith limited

So you were being intentionally dishonest in your assessment. Ok.

Originally Posted by shadow001

you stated that AA use doesn't drop the theoretical maximum fillrate on the HD7970, while it does on the GF110 to 24 Gigapixels so score one more point for AMD there

No, that is not what I wrote. You need to re-read what I wrote more carefully. AA actually improves the results for Nvidia because they get to use the 16 extra ROPs that are otherwise under-utilized.

Originally Posted by shadow001

those 48 rops on the GF110 might actually take up more room on the die than the 32 used in the HD7970

And being within 10 feet of a 7970 might cause you to spontaneously burst into flame. I have no actual evidence of this, but it is much safer to not take that chance.

Originally Posted by shadow001

in not dropping fillrate when AA is used like you mentioned for the HD7970

They don't drop fillrate when AA is used, they actually gain effective fillrate.

[john19055]

If nvidia comes out with a $299 card that is just as fast maybe even a little faster ,then that would put the hurt on the HD7970.That 576 cuda cores could be pretty close to what it is .That would be 12-Sm cluster units with 48 on each unit and add 8 texture unit for each cluster that would be 96,Run the core at 950 and memory at 5000.Even being just a 256bit interface ,it would be close.If that confriguration is even possible.

[Ninja Prime]

Originally Posted by john19055

If nvidia comes out with a $299 card that is just as fast maybe even a little faster ,then that would put the hurt on the HD7970.That 576 cuda cores could be pretty close to what it is .That would be 12-Sm cluster units with 48 on each unit and add 8 texture unit for each cluster that would be 96,Run the core at 950 and memory at 5000.Even being just a 256bit interface ,it would be close.If that confriguration is even possible.

The price tells you everything you need to know. Its not faster than the 7970. Its probably not even faster than the GTX 580. Unless NV all the sudden decided it doesn't like money, its performance is somewhat better than a current $299 card. So maybe 6970 level performance or slightly better. This was already rumored for months, NV launching mid and low end first, I don't know why people are suprised all the sudden. I'm smelling another FX 5800 so far, rumors of NV pulling out every dirty marketing trick in the book to try and make them look better...

[shadow001]

Originally Posted by ninelven

.


No, that is not what I wrote. You need to re-read what I wrote more carefully. AA actually improves the results for Nvidia because they get to use the 16 extra ROPs that are otherwise under-utilized.

.

So let me get this straight, there isn't enough memory bandwith as it is when running a fillrate test using no AA at all, and when using AA, wich puts even more pressure on whatever memory bandwith is available, those extra 16 rops on fermi are somehow going to become usefull in that scenario?...I find it hard to believe it.

[ninelven]

Originally Posted by shadow001

those extra 16 rops on fermi are somehow going to become usefull in that scenario

No, that isn't what I said. If you are bandwidth limited to that point, then it really doesn't matter. In normal gaming scenarios, you are not 100% bandwidth limited 100% of the time.

Tell you what... get a HD7970, downclock the memory to GTX 580 levels, downclock the core to GTX580 levels and then compare the results. If the HD7970 truly has more effective ROPs, then it should still win by a significant amount. If you don't have a 7970, ask someone who does to run the test for you. Then you will at least have reasonably sound data on which to base a conjecture.

EDIT: *You can also look at the 7970 vs the 6970. The 7970 has a 50% bandwidth advantage over the 6970, it performs 50% better than the 6970. Amazing what bandwidth can do in a completely bandwidth limited synthetic test.

[john19055]

I would'nt be surprise if it is just as fast as a GTX 580.It is Build on the 28nm chip ,Plus a GTX 560 TI is a pretty fast card in it on right ,and it is build with a 256bit interface.So if this is a new chip ,I don't see it being slower then a GTX 560 Ti,unless it is build useing 128bit interface.I already new that they were going to release the slower modles first ,or least that has been the rummor for a while now.I just saying if they put out a 660Ti ,it should be faster then the old GTX 560Ti and if they sell it for $299 it should be a pretty good deal,I heard it supposed to be a replacement for the GTX 560 Ti ,and the GTX 560 TI was quiet a bit faster then the GTX 460 1gig card. .But who knows it all rummors right now,it could be thier lowest end card ,besides the HD7970 is a good card just to pricey ,but If I was AMD I would be selling them at a premium too,since the GTX 580 is still around $520 .

[shadow001]

Originally Posted by ninelven

No, that isn't what I said. If you are bandwidth limited to that point, then it really doesn't matter. In normal gaming scenarios, you are not 100% bandwidth limited 100% of the time.

It is a known fact that AA will use up more memory bandwith period, so adding the gaming scenario variable only adds more potential bottlenecks both on the hardware side of the system(latency and I/O throughtput), and software bloat especially in the direct 3D API itself that basically hides it.

Tell you what... get a HD7970, downclock the memory to GTX 580 levels, downclock the core to GTX580 levels and then compare the results. If the HD7970 truly has more effective ROPs, then it should still win by a significant amount. If you don't have a 7970, ask someone who does to run the test for you. Then you will at least have reasonably sound data on which to base a conjecture.

Got 4 of them on the way and this is to replace a triple GTX580 SLI cooled on water, driving 3 2560*1440 panels in surround(7880*1440 pixels, bezel compensated), so i think i'm in good position to really push fillrate, texturing rate and memory bandwith of any video card to the limit, and even more so with AA enabled at that crazy resolution thanks and trying to keep it playable....

EDIT: *You can also look at the 7970 vs the 6970. The 7970 has a 50% bandwidth advantage over the 6970, it performs 50% better than the 6970. Amazing what bandwidth can do in a completely bandwidth limited synthetic test.

And makes the point of having more Rops kinda irrelevant now doesn't it, as even in AA figures, the GTX580 gets left behind even with the extra rops in it, and given that i have overclocked the memory on my own cards( i have 3 of them right now), it doesn't make a difference in any benchmark or game i've ever ran, even though i'm adding more memory bandwith than what they have stock....Increasing the core speed had a big improvement though...


So where's the bottleneck?