Can't you see that people here are talking about something more important (leather packaging) and no one else cares about the number of TFLOPS some random computer part spits out? Shoo with your specifications! Shoo, I say!
They haven't actually published most of the specs so I think jabbadap's explanation makes the most sense... if the tensor number is accurate, the other compute numbers probably aren't.
SXM2 version of V100 has 125TFlops of tensor power. So maybe it has same clocks and same tdp as sxm2 V100. So real specs would be 7.8Tflops fp64, 15.7Tflops fp32 and 31.4Tflops fp16. And thus gpu clocks(boost maybe) would be 1.533GHz.
"The only realistic ways to get 125 TFLOPS is to increase clock speeds or increase amount of SMs"
Not if they are memory-speed-limited in the previous design. Which might be, especially in the case of 8-bit encoded data where computation is essentially free (there is no need to actually add or multiply anything, just a direct read from a 64KB table with multiple read ports).
The current popular rumor is late July/early August. There's a related rumor that the consumer cards may be branded Ampere instead.
As always: Nothing is confirmed until the official press release. Actual results may vary. Past performance is not a prediction of future results. Professional driver on a closed course. Do not try this at home.
If he wanted to make a meaningful giveaway, he should have announced nVidia putting their GTX video cards back to their original MSRP's, and selling them direct from nVidia for a week. I'm tired of seeing GTX 1070's retailing for over $450, especially when I picked up my GTX 970 new for $100 less. And buying a used 970 is a crap-shoot since half of them are cooked from being used by Bitcoin miners.
It's called market, son. If you don't like it, you may choose between Cuba, DPRK and Venezuela. I suggest you DPRK - you will get nice badge. Absolutely free! And for the rest of your life :D
I was referring to the rumored launch price of next gen, now that current GPU prices have returned to near-MSRP levels. If they do launch *above* current GPU pricing, there's no justification for that besides greed + lack of competition.
I will grant you that the past year might've shown more strength of demand for GPUs at higher price points, and that will only embolden them to further raise launch prices.
Their last-gen GPUs aren't going to vanish right away. They're going to continue to sell them for a while, and as long as they are selling them, these new GPUs will slot in above them. Even if AMD releases something powerful, it won't substantially impact the market as long as miners are snapping them up.
At least the mid-range cards are affordable again. That's all I ever buy anyway.
I was expecting a 32GB Titan V. Now that we have an upcoming AMD Vega 20 7nm with 32GB, there will be an data center version (to replace MI25 accelerator at the top end) to compete with Nvidia Tesla V100, and Nvidia Titan V 32GB to go head-to-head with the recently benchmark tests spotted AMD Radeon Pro Vega 20 7nm for workstations. The only question is whether the Titan V 32GB will have ECC enabled (disabled ECC on the current 12GB Titan V may have been a potential source of the calculation errors recently reported and a higher 32GB HBM2 might fix that (with or without enabled ECC))
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
38 Comments
Back to Article
MrSpadge - Thursday, June 21, 2018 - link
I would have expected them to pack the cards in black leather.eurico - Thursday, June 21, 2018 - link
I wish I could up-vote that comment!ZeDestructor - Thursday, June 21, 2018 - link
Ahh, you, good sir, have made my day :)ImSpartacus - Thursday, June 21, 2018 - link
Jesus, that would've been perfect.Or at least a leather wrap around part of the box.
Alexvrb - Thursday, June 21, 2018 - link
As long as they attach the leather with wood screws.Spunjji - Friday, June 22, 2018 - link
Our hearts are still burnin' for Thermi after all these years :DR12 - Wednesday, April 3, 2019 - link
You seem to be misusing Jesus name. He is my friend/Friend, my Savior and my Lord.SuperiorSpecimen - Thursday, June 21, 2018 - link
I'm dense.Can someone explain?
mode_13h - Friday, June 22, 2018 - link
Nvidia's CEO always makes his PR appearances in a black leather jacket.Cuz he's a bad boy like dat.
Bulat Ziganshin - Thursday, June 21, 2018 - link
The only realistic ways to get 125 TFLOPS is to increase clock speeds or increase amount of SMs, so FP32/FP64 flops should be changed proportionallyPeachNCream - Thursday, June 21, 2018 - link
Can't you see that people here are talking about something more important (leather packaging) and no one else cares about the number of TFLOPS some random computer part spits out? Shoo with your specifications! Shoo, I say!Alexvrb - Thursday, June 21, 2018 - link
They haven't actually published most of the specs so I think jabbadap's explanation makes the most sense... if the tensor number is accurate, the other compute numbers probably aren't.jabbadap - Thursday, June 21, 2018 - link
SXM2 version of V100 has 125TFlops of tensor power. So maybe it has same clocks and same tdp as sxm2 V100. So real specs would be 7.8Tflops fp64, 15.7Tflops fp32 and 31.4Tflops fp16. And thus gpu clocks(boost maybe) would be 1.533GHz.Bulat Ziganshin - Thursday, June 21, 2018 - link
125/4=31.25 and so on, OTOH 31.4*4=125.6 so it may just a matter of roundingI just realized that 8 such GPUs has a nice 1 PFlop total speed, which is probably what they will market for their DGC-8.5 workstations
mode_13h - Friday, June 22, 2018 - link
Because they're rated at 300 W, whereas the PCIe version is only 250 W.Spunjji - Friday, June 22, 2018 - link
That makes sense - set a 300W rating on the card and boom, more clock headroom. What's one more 8 pin connector among friends?mode_13h - Friday, June 22, 2018 - link
Yeah, the standard Titan V has a 6-pin + 8-pin and a 250 W rating.peevee - Thursday, June 21, 2018 - link
"The only realistic ways to get 125 TFLOPS is to increase clock speeds or increase amount of SMs"Not if they are memory-speed-limited in the previous design. Which might be, especially in the case of 8-bit encoded data where computation is essentially free (there is no need to actually add or multiply anything, just a direct read from a 64KB table with multiple read ports).
Bulat Ziganshin - Thursday, June 21, 2018 - link
memory speed is completely separate topic. these TFLOPS are ALWAY computed from raw ALU powerBulat Ziganshin - Thursday, June 21, 2018 - link
and if you mean on-chip RAM (or rather register pool), it scales by frequency/amount with SMs, since it's part of SM itselfajlongman - Thursday, June 21, 2018 - link
Does this mean my Titan Xp is now obsolete?mode_13h - Friday, June 22, 2018 - link
https://www.anandtech.com/comments/12170/nvidia-ti...mode_13h - Friday, June 22, 2018 - link
Oops, that was a link to the comment thread.https://www.anandtech.com/show/12170/nvidia-titan-...
peevee - Thursday, June 21, 2018 - link
Anybody has any idea when Volta-based consumer GPUs are going to be announced?DanNeely - Thursday, June 21, 2018 - link
The current popular rumor is late July/early August. There's a related rumor that the consumer cards may be branded Ampere instead.As always: Nothing is confirmed until the official press release. Actual results may vary. Past performance is not a prediction of future results. Professional driver on a closed course. Do not try this at home.
peevee - Thursday, June 21, 2018 - link
So it might be wrong to buy a used 1060 now shortly before people started dumping them for a latest and greatest?Bulat Ziganshin - Thursday, June 21, 2018 - link
absolutely. OTOH, new cards probably will not be cheap, nor produced in large amountknightspawn1138 - Thursday, June 21, 2018 - link
If he wanted to make a meaningful giveaway, he should have announced nVidia putting their GTX video cards back to their original MSRP's, and selling them direct from nVidia for a week. I'm tired of seeing GTX 1070's retailing for over $450, especially when I picked up my GTX 970 new for $100 less. And buying a used 970 is a crap-shoot since half of them are cooked from being used by Bitcoin miners.Bulat Ziganshin - Thursday, June 21, 2018 - link
It's called market, son. If you don't like it, you may choose between Cuba, DPRK and Venezuela. I suggest you DPRK - you will get nice badge. Absolutely free! And for the rest of your life :Djabbadap - Thursday, June 21, 2018 - link
Well they actually sell them directly, that gtx1070FE is $399 on their web store. Though personally I would put $50 more and get that gtx1070ti FE.https://www.nvidia.com/en-us/geforce/products/10se...
mode_13h - Friday, June 22, 2018 - link
Don't blame Nvidia. Maximizing profit is simply their nature. It'd be like blaming a hungry tiger for eating a small child.Rather, blame AMD for not releasing any new GPUs this year...
Alexvrb - Saturday, June 23, 2018 - link
Neither one of them is to blame, for this particular cluster. Supply. Demand. Altcoins. It's not sorcery.mode_13h - Sunday, June 24, 2018 - link
I was referring to the rumored launch price of next gen, now that current GPU prices have returned to near-MSRP levels. If they do launch *above* current GPU pricing, there's no justification for that besides greed + lack of competition.I will grant you that the past year might've shown more strength of demand for GPUs at higher price points, and that will only embolden them to further raise launch prices.
Alexvrb - Monday, June 25, 2018 - link
Their last-gen GPUs aren't going to vanish right away. They're going to continue to sell them for a while, and as long as they are selling them, these new GPUs will slot in above them. Even if AMD releases something powerful, it won't substantially impact the market as long as miners are snapping them up.At least the mid-range cards are affordable again. That's all I ever buy anyway.
mode_13h - Friday, June 22, 2018 - link
Does nobody else see that this is merely a Quadro GV100 with a Titan V shroud (and maybe slightly tweaked clocks)?Unless they start selling these for less than the GV100, it's just a publicity stunt.
sunty - Sunday, June 24, 2018 - link
I was expecting a 32GB Titan V. Now that we have an upcoming AMD Vega 20 7nm with 32GB, there will be an data center version (to replace MI25 accelerator at the top end) to compete with Nvidia Tesla V100, and Nvidia Titan V 32GB to go head-to-head with the recently benchmark tests spotted AMD Radeon Pro Vega 20 7nm for workstations. The only question is whether the Titan V 32GB will have ECC enabled (disabled ECC on the current 12GB Titan V may have been a potential source of the calculation errors recently reported and a higher 32GB HBM2 might fix that (with or without enabled ECC))sunty - Sunday, June 24, 2018 - link
ok..forgot about the Quadro GV100. So this will be GV100 minus the ECC, the whole point of Titan Vmode_13h - Tuesday, June 26, 2018 - link
Thanks. It's like everybody forgot about the Quadro GV100.Not all Quadro's have ECC, but it seems GV100 does (or so the datasheet claims).