NVIDIA Maxwell Speculation Thread

[Deleted member 2197]

Arwin,
Is that your video? If so you running 3 monitors with single 970 or SLI 970's?

 

[swaaye]

I went from a 6950 to a 970. I think that's around 300% improvement, plus having double the RAM to work with.

Oh and the card is like 2/3 the size of the 6950.

 

[Arwin]

Arwin,
Is that your video? If so you running 3 monitors with single 970 or SLI 970's?

Yes, mine ... Single GTX 970 (MSI 4G), 16Gb and i7 vpro 3rd gen, resolution is 5080x1050 or something like that. Seems to still manage a synced 60fps except with a lot of cars right in front of me. Very playable. These games also aren't that demanding though - I'm sure it would get worse with more modern titles.

 

[Deleted member 2197]

 

[pjbliverpool]

$600? I'd be very surprised if so.

 

[Alexko]

$600? I'd be very surprised if so.

It all depends on Fiji's performance and price.

 

[A1xLLcqAgt0qc2RyMz0y]

How much effort would it be for Nvidia to put GM200's on a single board (like the Kepler's on the Titan-Z)?

Do you think Nvidia would do this to counter AMD's Fiji VR dual-chip?
http://www.fudzilla.com/news/graphics/37566-two-amd-fiji-cards-coming-in-june

 

[lanek]

How much effort would it be for Nvidia to put GM200's on a single board (like the Kepler's on the Titan-Z)?

Do you think Nvidia would do this to counter AMD's Fiji VR dual-chip?
http://www.fudzilla.com/news/graphics/37566-two-amd-fiji-cards-coming-in-june

Not much effort. If AMD can do a dual gpu with chips as Hawaii, i think everything is possible... I will absolutely not been surprised to learn that Nvidia have something like that at work.

But the question is still how perform FijiXT and how could perform an dual FijiXT.. Im not sure they want it end as the TitanZ vs 295x2.. I can imagine, this time they will not do a Titan x2, but more a 980TI x2.

 

[Deleted member 2197]

NVIDIA Quadro M6000 Review: Maxwell Goes Workstation

Last July, we noted that the FirePro W9100 was a much more competitive card than the W9000 had been and congratulated AMD on its positioning and overall performance. The Quadro M6000 doesn't completely erase that fact and AMD's competitive value proposition, but it definitely puts some holes in it. Against the K6000, the W9100 had a solid, near-universal price/performance win -- it wasn't as fast in a number of tests, but it was definitely competitive on price. The new NVIDIA Quadro M6000 weakens that argument significantly.

There are now cases -- and more than just in SpecViewPerf 12 -- where the M6000 is a better deal than the FirePro W9100, even on a price/performance basis. Our results suggest that AMD will have to cut the price of its 16GB behemoth if it wants to stay competitive with NVIDIA's Quadro line. At $2000 - $2500, the FirePro W9100 would be more competitive again with the Quadro M6000. For now, however, NVIDIA clearly has the high-end workstation market locked down. We'll have to wait and see if AMD's new architecture can challenge it later this year.

http://hothardware.com/reviews/nvidias-quadro-m6000-maxwell-goes-workstation?page=1

 

[Frenetic Pony]

$600? I'd be very surprised if so.

Eh, speculation based on speculation of the 390x pricepoint. Should be flattered that speculation from these forum threads has been picked up as more "official" rumors, but then sites like WCCF and Fudzilla are hardly journalistic bastions. After all they're just copypasting people on here shooting the breeze and idle speculation.

And I don't doubt for a second it's just them trawling these forums and copypasting these threads. I remember a similar site picking up a rant from a former GPU firmware developer. It was literally just a forum post on how API's related to game development, and how bad DX and OpenGl APIs can end up making shipping code. They of course made it sound like some kind of official and super important "insider" thing.

 

[silent_guy]

At this point I only consider videocardz.com to be trustworthy. They've copy a lot of the other garbage, of course, but they've been remarkably accurate with their own original content.

 

[homerdog]

Since this is the Maxwell thread I was wondering, how likely is it that NVIDIA would release some kind of GTX970v2 with a full set of ROPs and L2? They've made minor updates to their lineups before so it wouldn't be unprecedented (e.g. GTX275).

Also I thought there were supposed to be 8GB 980s and 970s but so far I haven't seen them. Do they exist?

 

[Malo]

Since this is the Maxwell thread I was wondering, how likely is it that NVIDIA would release some kind of GTX970v2 with a full set of ROPs and L2?

Isn't that a 980?

 

[Blazkowicz]

I thought they save them up for the next AMD launches. Although, a GTX 970 with 8GB would have people bitching that it's missing a full GB.
I also wonder how you would sell a "fixed up" 970 : okay, here's the defective one and this one is fully working, we promise!

 

[Deleted member 13524]

It'd be ridiculous to sell a 970 with 8GB. At most, they'd sell a 7GB version.

 

[homerdog]

Isn't that a 980?

Not quite. The 980 still has 3 more SMs than the 970.

I thought they save them up for the next AMD launches. Although, a GTX 970 with 8GB would have people bitching that it's missing a full GB.
I also wonder how you would sell a "fixed up" 970 : okay, here's the defective one and this one is fully working, we promise!

They would surely just call it the GTX975 or something like that, with slightly higher clocks and maybe another SMM.

 

[Deleted member 2197]

GeForce GTX 980 Ti rumours ...

On the basis of reports by Baidu there are too recent rumors to the GeForce GTX 980 Ti conflicting information. Thus, to the alleged NVIDIA plans now to await the introduction of the Radeon R9 390X, but to provide previously another template. The presentation of the GeForce GTX 980 Ti will take place in May 2015 - more precisely in the period between 16 and 26 May

But more: the GTX 980 Ti is in a complete and non circumcised GM200 chip (GTX Titan X) and secure even compete with higher clock speeds than the GeForce GTX Titan X. That would inevitably mean that the GTX 980 Ti is faster to works than the Titanium X, which currently their new owners expected for expensive 1200 euros. The only restriction is the main memory. This will "only" be the GTX 980 Ti 6 gigabytes, in the GeForce GTX Titan X he is 12 GB in size, which will make in practice hardly noticeable.

If this is true then it would of course be a severe blow to the face of the recent "early adopters" who have already received for such a princely sum X Titan have purchased. Because it is clear that the GTX 980 Ti is probably wrestle with the R9 390X, and will start in the price range from 700 euros - at least if one believes recent rumors.

http://translate.googleusercontent...._2015/&usg=ALkJrhjLxXLDmrBL-dQJmf9JkYcGvx5rtA

 

[RedVi]

Nice translation.... lol

 

[Grall]

"non circumcised"...

 

[Alexko]

"non circumcised"...

I guess in German, chips are not castrated but circumcised (unless it's a mistranslation). I don't know about you, but the fact that a loose translation can lead to confusion between circumcision and castration makes me feel uneasy.