DavidGraham
Veteran
So the 2.1GHz is either an OC'ed speed or an error in the frequency reading?
This is their way of saying "we don't want to stop making money with g-sync yet"
I think both products are very well priced. The difference between the Founders Edition may be the boost clock since Jen Hsun said it would have crazy overclockability .... I think the demo may have been using the GTX 1080 Founders Edition with the 2114 MHz clock.
firstminion
Newcomer
I think Founders Edition will be the first cards available, when availability gets better the MSRP will drop and they will drop that name.
gamervivek
Regular
OCed otherwise just 30-40% over 980Ti with 2.1Ghz and 2.5k shaders would be pretty awkward.
A1xLLcqAgt0qc2RyMz0y
Veteran
Not at all. Adaptive Sync will always be an optional part of the DP spec: there is no reason to require it for business monitors etc. I think it's because the spec is too new for certification testing. If DP is anything like other telecom specs, a major part of certification testing is interop testing. I don't think there are DP 1.3 monitors on the market right now. Let alone DP 1.4.
AMD can send all its high-end stuff to the local municipal dump. The $379 price is devastating for Fury X.
I think we can also let go of the idea that 16nm is not yet mature enough: it'd be insane to price this aggressively without being able to supply the market, because Nvidia just made most of its Maxwell chips irrelevant.
I think we can also let go of the idea that 16nm is not yet mature enough: it'd be insane to price this aggressively without being able to supply the market, because Nvidia just made most of its Maxwell chips irrelevant.
Last edited:
2.1GHz without liquid cooling. So much for 2GHz being a stretch...
What? Only 20% faster at almost 100% more clock rate? Is that a joke?
Time to bargain for a cheap 980Ti in wait for the real GeForces.
The guaranteed clock rates are about 50% higher. And the Titan X has 3072 cores vs 2560 cores for the 1080.
2560/3072*1.5=1.25 ends up more or less where you'd expect it to be.
Ext3h
Regular
The performance numbers published vary a bit.
In the resources published by NV online, it says 2x Titan X for VR applications (mostly due to some type of "Multi-Projection" support?), and 1.3-1.4x for existing titles.
Perf announced on stage at the event is plain out 2x without any differentiation. (While the slide in the background has "VR performance" on the axis.)
Does anyone know what this "Multi-Projection" support actually include? Sounds like that batches geometry for stereo projection prior to rasterization, so even the 2x perf is probably just absolute best case for a synthetic, geometry limited scene? Probably achieves the savings only during tesselation and geometry shader evaluation?
In the resources published by NV online, it says 2x Titan X for VR applications (mostly due to some type of "Multi-Projection" support?), and 1.3-1.4x for existing titles.
Perf announced on stage at the event is plain out 2x without any differentiation. (While the slide in the background has "VR performance" on the axis.)
Does anyone know what this "Multi-Projection" support actually include? Sounds like that batches geometry for stereo projection prior to rasterization, so even the 2x perf is probably just absolute best case for a synthetic, geometry limited scene? Probably achieves the savings only during tesselation and geometry shader evaluation?
Specs are a perfect match of what was previously speculated.
To be confirmed still 160 TMUs, 64 ROPs, 40 SMs.
A question arises what will be the big graphics Pascal, given the 15B transistor budget.
Will it be based on the P100 ie 60 SMs, 3840 cores, 240 TMUs.
or will it be 2xGP104 ie 80 SMs, 5120 cores, 320 TMUs.
I'm hoping for the latter.
To be confirmed still 160 TMUs, 64 ROPs, 40 SMs.
A question arises what will be the big graphics Pascal, given the 15B transistor budget.
Will it be based on the P100 ie 60 SMs, 3840 cores, 240 TMUs.
or will it be 2xGP104 ie 80 SMs, 5120 cores, 320 TMUs.
I'm hoping for the latter.
Transponster
Newcomer
Wouldn't it be *a little bit* oversized? With "HBM2" and everything...
Ext3h
Regular
HBM2 isn't on-die, and the HBM2 is reportedly smaller than a GDDR5 controller. So the die size would be still somewhat plausible, when cutting on the register files compared to GP100.
GeForce can render same triangle to multiple viewports without GS amplification costs. See GL_NV_viewport_array2 and
GL_NV_viewport_swizzle. Maxwell can do this to 9 viewports, Pascal can do this to 16.
Similar threads
- Poll
