Some interesting stuff.
And yes, they know it isnt made for silly gaming but actually important stuff
People will be buying these to replace Titan XPs or GTX 1080 Tis in compute workstations. Gamers should not be buying this card at all.
Doesn't matter, it's been solidified in the consumer mind that Titan cards are ultra enthusiast level gaming hardware for those with money to blow. The Ti is obviously a just as good/better choice price/performance wise but that doesn't mean anything too those who the Titan has targeted for years now.
except the 780ti beat the 700 titanOriginally Posted by Echoa
Doesn't matter, it's been solidified in the consumer mind that Titan cards are ultra enthusiast level gaming hardware for those with money to blow. The Ti is obviously a just as good/better choice price/performance wise but that doesn't mean anything too those who the Titan has targeted for years now.
I don't get this "This Titan isn't a gaming one" thing when we all know good and well that's a moot point at this stage. The Titan REP is set in stone now, the number of times I've heard "You'd need a Titan to run that on Ultra" trumps any other card.
The 780 Ti lost to the Titan Black
Exactly what I was going to say, 100% agree.Originally Posted by Charcharo
The problem is, the Kepler Titans were prosumer cards. Sure they were awesome for games, but that was not their forte. Their real value was as HPC, CUDA, scientific and prosumer cards. Which was great, they made sense and were actually CHEAP for the target audience.
But Maxwell and Pascal titans are just gaming cards. Nothing more.
This Titan V though seems like a return to form - an actual prosumer product. Not just some gaming thingy.
the 780 ti came 9 months after the original titan and it had a bigger core.
This is not a fully enabled part, it has 5120 SPs and a 3072-bit memory bus. GV100 has 5376 SPs and 4096-bit.Originally Posted by jarble
Pure speculation on my part but if they are getting a fully enabled part out the door (retail not closed shipments) it would seem to indicate that the yield on this design is fairly high. Compared to the previous titan releasing with a cutdown chip to start then followed up with a fully enabled chip much later.
Would 3 Vega 64 greatly outperform Titan V on machine learning tasks other tasks that Titan V is designed for?
Thanks to AMD providing competition a bit giving semi pro drivers to Vega Frontier, Nvidia did the same thing and Titans are actually work horses for cad programs and good values now.Originally Posted by Charcharo
The 780 Ti lost to the Titan Black
The 1080 Ti loses to the updated, full Titan Xp.
The problem is, the Kepler Titans were prosumer cards. Sure they were awesome for games, but that was not their forte. Their real value was as HPC, CUDA, scientific and prosumer cards. Which was great, they made sense and were actually CHEAP for the target audience.
But Maxwell and Pascal titans are just gaming cards. Nothing more.
This Titan V though seems like a return to form - an actual prosumer product. Not just some gaming thingy.
No, you're wrong in several ways.
None of what you said is relevant at all to my post or the question I was answering. For the record, I didn't even say why it wasn't fully enabled, and I could not care less about your being flustered with how HBM is marketed.Originally Posted by prjindigo
No, you're wrong in several ways.
The first of which is that this is a damaged bin part not an under-enabled part. nV can charge 2x as much for fully functional parts. These cards have a bad HBM2 stack on them or they'd be 16gb.
The second is that each HBM2 stack is only 128 bits wide and cannot be addressed with larger than 64bit data, it has two 64bit channels into it not 1024.
The parallelism of the ram in the HBM2 stack is immaterial.
These cards are 384bit, are missing an HBM2 stack, may have damaged nVlink parts and are useless in arrays because of it.
I've described elsewhere how HBM and HBM2 stacks work but basically the "ram controller" for the stack is in the bottom and has access channels to the PCIe and the GPU, the GPU communicates with 2x64bit bi-directional channels per stack at very high speed.
Claiming that the HBM is 3072 or 4096 wide is like claiming that 4-channel boards with 8 slots populated with 8gb sticks of ram are 1024bit wide memory buses. There may be 1024 bits of memory port address in there but it's still only 4x64 bit channels to the CPU for 256.
Don't propagate marketing twaddle like "4096" and "x86"... they're lies.
Its the first card with the "TITAN" imprint on the side of the shroud instead of Geforce, isnt that clear enough?