[Silent_Buddha]

That seems to imply that GTX 560 is going to have 69xx power levels. Stock might be slightly less than 6950, while OC'd version might be more than 6970. Wouldn't surprise me if the super OC versions came close to 300 watts.

Regards,
SB

[NathansFortune]

Reviews should be out by now no?

The card is on sale and not a single review from any of the major websites, in fact not even any performance indications other than one forum guy who bought one.

[TKK]

Originally Posted by NathansFortune

Reviews should be out by now no?

The card is on sale and not a single review from any of the major websites, in fact not even any performance indications other than one forum guy who bought one.

afaik NDA ends at 3 p.m. in Germany, which is in about 50 minutes.

[NathansFortune]

OK, I thought it was 12PM GMT, but that turned out to be the expiry time for the 6950 1GB NDA...

[psolord]

Originally Posted by NathansFortune

OK, I thought it was 12PM GMT, but that turned out to be the expiry time for the 6950 1GB NDA...

So we are going to see double reviews today? Both the 6950 1GB and the 560? Yeepee!

[NathansFortune]

http://www.anandtech.com/show/4135/n...the-250-market

[Blazkowicz]

I've checked again whether there was info on a GF116 chip and found it

In addition, the NVIDIA GF116 core based GeForce GT 550/GT 530 will come in Q1 2011.

The GeForce GT 550 and GT 530 are to be based on the GF116 core - the successor of the GF-106 Core on which the previous models of these cards were built. Number of stream processors should still be up to 192 (four groups of SM), the core size will not change, but some people say which may be increased to five SM 240 stream processors, but no confirmation.

http://www.downloadatoz.com/driver/a...n-q1-2011.html

nice!
the new 530GT is the missing board I was wishing for and the incremental improvement of GF11x doesn't hurt.

I couldn't way and bought a gt240 gddr5 instead, though.

[psolord]

Excellent work from Nvidia on the GTX 560. Nothing not expected really, but still impressive.

According to Computerbase, two of these babies are around 35% faster than a GTX 580 (for normal resolutions up to 1920k). Considering they cost almost the same, that's just awesome.

[Dr Evil]

Originally Posted by psolord

Excellent work from Nvidia on the GTX 560. Nothing not expected really, but still impressive.

Yeah solid card with excellent OC room. So Gigabyte did bring that 1GHZ card priced at 269$, that'll cause some havoc...

[NathansFortune]

I think Nvidia priced it wrong. If they had priced it at $239 with 2GB and $219 with 1GB it would be getting much more positive reviews, but I don't know whether Nvidia's margins would allow for such aggressive pricing. When they released the 460 they were in a much weaker position, the 480/470 were much less successful and Nvidia had lost a lot of market share. This time maybe such aggressive pricing is something Nvidia can afford not to do.

[Pressure]

Graphic card launches are becoming less exciting these days and the Geforce GTX 560 Ti just seems to hit an empty slot with the expected performance. Nothing wrong with the part, just hard to get really rilled up about it. The same can basically be said about the Radeon HD 6950 1GB card.

The mobile space is where all the action is these days

[GZ007]

It will be interesting what will hapen to the gtx460 OC cards. Because you could find 750 MHz cards and sometimes even more for almost the same price or a litle more (probably they will change this very soon). This time noone included OC cards in the review as with the 6870 test.

[DarthShader]

I was reading the Anandtech review, to see what kind of architecture improvements they included in the GTX560 comparing to GTX460 and I am left disappointed. They just reworked some transistors. Not even the Z-cull improvemnts of GF110? Or are they already present in the 460? That's almost like rebrand and they give the "Ti" suffix to it.

It's still a good card with good thermals and low noise, but one would expect a bit more from +100 to name...

[mczak]

Originally Posted by DarthShader

I was reading the Anandtech review, to see what kind of architecture improvements they included in the GTX560 comparing to GTX460 and I am left disappointed. They just reworked some transistors. Not even the Z-cull improvemnts of GF110? Or are they already present in the 460?

Some reviews said these were already in GF104 (don't know if it's true but I always had the suspicion it could be the case). I didn't find any quote indeed that there are ANY functional changes at all (things like minor bugfixes).
But imho the clock gains possible with even slightly less voltage (it seems - difficult to tell really with the multiple vids) are quite impressive. Ok it has a bit less overclocking potential (so operating closer to its limit) but still the improvement is very notable, also improving perf/w.

[psolord]

When it comes to averages, Computerbase and Hardware.fr have become my choice as of late. 6% difference from TPU to Computerbase may not sound much, but it is. It paints a quite different picture, ie it's different to say the 5870 is 5% faster than the 560 instead of the 560 is actually faster than the 5870 (having the 1920k +filters in mind here).

Also Tweaktown has shaped up considerably, although I mostly use it to compare cards of the same IHV, since most of their suite is kinda Nvidia favorish.

Speaking of Tweaktown, they benched the Gigabyte 560 SOC today, which runs at 1Ghz by default. According to their own TPR rating, the 560 SOC is on par with the 6970 and slightly behind the 570. If that is not impressive, I don't know what is.

Yes this card will come with a price premium, increased power draw, blah blah blah, but what I am admiring here, is the potential of the card. Even vanilla 560 users, will at least have the option to OC if need be and have a taste of 570 like performance. I am not talking about senseless 24/7 OC. Just a helping hand when is needed.

All that, in conjunction with a lower price tag than a 570, spells impressive in my book.

[DarthShader]

Originally Posted by mczak

Some reviews said these were already in GF104 (don't know if it's true but I always had the suspicion it could be the case)

The full speed FP16 texture filtering appeared indeed first in th GF104, but when reading GF110 reviews I remember the improved Z-cull being described as a new feature. I expected it to appear in the GF114 too. Would proably require some specific tests to find out.

Less OC headroom is rather a good sign to me - means the engeneers extracted the most of their design at stock and it works as planned and I don't have to do risky stuff to extract more performance.

[swaaye]

Originally Posted by Dave Baumann

And the vapor chambers are excellent for the products they were designed for.

I am rather curious about how they compare to the more "traditional" heat pipe designs with respect to efficiency. The Cayman chips dump a lot of heat and I imagine that it's challenging to manage cheaply in the 2 slot cooler format. My only qualm would be the blower noise, but whatcha gonna do when you need to move air in that direction.

[Arty]

Originally Posted by NathansFortune

I think Nvidia priced it wrong. If they had priced it at $239 with 2GB and $219 with 1GB it would be getting much more positive reviews, but I don't know whether Nvidia's margins would allow for such aggressive pricing. When they released the 460 they were in a much weaker position, the 480/470 were much less successful and Nvidia had lost a lot of market share. This time maybe such aggressive pricing is something Nvidia can afford not to do.

Agreed, they had the chance to kill AMD in this segment but seems like the mutual agreement is that both of them can make some money ..

[NathansFortune]

Originally Posted by Arty

Agreed, they had the chance to kill AMD in this segment but seems like the mutual agreement is that both of them can make some money ..

If so expect AMD to whistleblow at some point. The weaker company always blows the whistle first to avoid fines...

[NathansFortune]

I was just thinking about it, I take it back. Nvidia haven't missed the opportunity, they have left it for another time. The next card they will introduce is the GTX555 which will have the same GF114, but with the chips that couldn't make the higher clock. They will use harvested GF114 parts to compete with Barts until they get GF116 out of the door or even after depending on how competitive that will be.

[Silent_Buddha]

Originally Posted by NathansFortune

If so expect AMD to whistleblow at some point. The weaker company always blows the whistle first to avoid fines...

Unlikely if both are making money. In a case like this where one company isn't dumping product below cost, it'll be up to a 3rd company that can't make a competitive product or a class action suit to blow the whistle. And in a case of price fixing for profit, a 3rd competitor wouldn't blow the whistle as it's likely the price fixing will allow their less performant products to command a higher margin than otherwise. So it'd end up having to be some sort of class action suit or government instigated investigation.

But considering the prices aren't unreasonable (don't appear inflated) and one company did have to drop prices and/or introduce a competitive product in response, it's unlikely anything would be done even if some 3rd company claimed price fixing.

Regards,
SB

[gongo]

more likely to see GTX 560 without the suffix...but with disabled CUDA cores? Confusing Charlie crap again...

I am impressed with the longevity of Fermi...after all that was said wrt to Nvidia forgetting about gamers cue white-noise ..GF114 has been in a class of its own! It is smaller than Cayman yet it competes well...i dont think AMD has an answer for it...Barts is near its clock limits...Cayman Pro has OC-marketing potential but CCC locked...role reversed! AMD failed to keep up RV700(perf/w/Celsius was bad) and Cypress(bad tess/AO/SSAO perf) fps/$ charge...

Partly to blame are AMD shitty drivers ...11.1a does no promised fix for VLIW4...another month to wait....For Cayman to lose out to Cypress in non TWIMP games like F1/Civ5/Dirt...you add in TWIMP games...the average fps/$ becomes worse...than it should have been...it is also starting to seem AMD DX11 perf/features is not as strong as Nvidia.....early bird is not always good it seems..

[mczak]

Originally Posted by gongo

more likely to see GTX 560 without the suffix...but with disabled CUDA cores? Confusing Charlie crap again...

Hmm I'm confused too with the lineup including 560 TI and old 460 for now. So what will 560 non-Ti be? Lower clock? And 555 7 SMs and less clock (that is, pretty much rebadged 460 1GB) ?

I am impressed with the longevity of Fermi...after all that was said wrt to Nvidia forgetting about gamers cue white-noise ..GF114 has been in a class of its own! It is smaller than Cayman yet it competes well...

I think the "forgot about gamers" part was really mostly only attributed to GF100 (and GF110), which have the uber-fast (but disabled for most products) DP, and things like ECC, not the whole Fermi family. And by the looks of it that stuff doesn't hurt too much neither.
I don't know how you came up with "class of its own". Cayman is not even 10% bigger yet 10% faster (not the HD6950 certainly, but the chip in its full version). Looks like quite same class to me.
If anything, Barts is in a class of its own, being a lot smaller yet not that much slower (or in other words, GF106 is just slightly smaller than Barts yet can hardly compete with Juniper...)

i dont think AMD has an answer for it...

Well the HD 6950 1GB at similar price looks like a nice answer to me. It's not like that's a whole lot more expensive to manufacture, the die size difference is quite minimal, they use the same memory chips, and the pcb probably has similar complexity (given the power draw is near identical).
Certainly though, GF114 is the most competitive chip in Nvidias GF1xx lineup - you could quite say it's the only chip which offers comparable performance to competing similarly sized AMD chips.

[CarstenS]

I seem to remember Dave saying that a very high memory speed also involves some PCB complexity.

[mczak]

Originally Posted by CarstenS

I seem to remember Dave saying that a very high memory speed also involves some PCB complexity.

Yes, this makes sense. More careful routing is probably needed, and high frequency noise might be more of a problem. It is unclear though to me if the pcb is actually more complex in this case (I was primarily basing this on power draw) but it could be.