Nvidia Titan V

ontariotl

Limp Gawd
Joined
Oct 2, 2008
Messages
242
SLI is dead.
you don't need an sli bridge for multi card cuda/computing support since they work independently.. these cards aren't meant for gaming.
While that maybe true, even the very first Titan GPU (and Z) had SLI bridge connectors and some of these cards Nvidia stated they were not for gaming (even though later models pretty much were). It's just good to see that Nvidia is finally serious with this Titan V as a compute gpu.

Besidea mgpu compute does not need SLI - all messaging occurs thru PCIe

sli games also don't need that much processing power as GPU's evolve - they are legacy games and a single card outperforms any SLI older gen anyways at any game


future games will use MGPU compute
I agree, but it's good to see Nvidia finally move away from the SLI bridge when AMD did away with their crossfire bridge it a few gens back.

Since this card is supposed to be a compute card, it is possible SLI could still be implemented on the gaming side of Volta/Amphere.
 

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
56,392
honestly there is room for AI in gaming and get rid of the dumb enemies

so you may get the tensor core for consumer
This won't happen. We won't see any kind of AI in games that needs to offload to GPU's like that. AI will remain dumb because it's got to run on integrated graphics and those ultra cheap bottom end GPU's with more VRAM than they can effectively use.
 

Araxie

Supreme [H]ardness
Joined
Feb 11, 2013
Messages
6,439
This won't happen. We won't see any kind of AI in games that needs to offload to GPU's like that. AI will remain dumb because it's got to run on integrated graphics and those ultra cheap bottom end GPU's with more VRAM than they can effectively use.
AI always run on CPU because they need to be really precise.. we don't only need way more advanced CPUs before we can have more advanced AI, we also need better programmed engines suited for that task, a proper AI taking care of the *enemy intelligence* isn't cheap for gaming.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
14,488
SLI is dead.
With as much work as engine developers are putting into getting it working on DX12 and Vulkan, I don't see that quite yet. This card lacking the SLI connector isn't that big of a deal, it's clearly meant more for development work, with a bone thrown to wealthy gaming enthusiasts.
 

Araxie

Supreme [H]ardness
Joined
Feb 11, 2013
Messages
6,439
Never mind, I misread the clock speed of the v100 Press Release. For some reason I thought it was a couple hundred faster than this beauty..

Titan V will have the SAME boost clock as the fastest version of v100 Tesla, so the only castration will be the memory bandwidth.

You are RIGHT, this will be at least 40% faster :D

It boost higher than Titan Xp.. that only by itself together with the 128 ROPS and 5120 shaders mean a lot of performance gains without factor architectural improvements.

I think 40% it's a good stimation nut it certainly can go up to about 50% in my numbers.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
14,488
Are we looking at 50%+ gaming performance above the current Titan XP?

Then would a pure gaming part, i.e. GV102, be looking at potentially +100% over a 1080 Ti?
 

razor1

[H]F Junkie
Joined
Jul 14, 2005
Messages
10,155
Are we looking at 50%+ gaming performance above the current Titan XP?

Then would a pure gaming part, i.e. GV102, be looking at potentially +100% over a 1080 Ti?

Well depends on how high they clock based on power consumption and what the node can handle voltage wise, the chip itself, should be around 65% of the size of the gaming version. So power consumption should be much lower than the full V100.

Also V100 does have x2 the compute performance over P100, so yeah maybe able to reach that as long as its got the bandwidth.
 

Armenius

Fully [H]
Joined
Jan 28, 2014
Messages
22,822
Damn it. Was so ready to pull the trigger when I saw the e-mail, but $3k is way more than I am willing to spend. Will continue to wait for GV102.
 

realworld

Limp Gawd
Joined
Apr 24, 2016
Messages
297
I can see their marketing trick from a mile away. Release super expensive world beater, then follow up with a much cheaper card with similar gaming performance next year. Everyone will think they’re getting a sweet deal. Profit.

I would’ve jumped in at $1000~1400
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
14,488
I can their marketing trick from a mile away. Release super expensive world beater, then follow up with a much cheaper card with similar gaming performance next year. Everyone will think they’re getting a sweet deal. Profit.
The one they've been doing for years?

Not really a marketing trick- you want the best first, you pay for it.
 

polonyc2

Fully [H]
Joined
Oct 25, 2004
Messages
17,994
what's up with all the AI calculations??...is this the beginning of Skynet?
 

Brackle

Old Timer
Joined
Jun 19, 2003
Messages
7,409
Don't get me wrong I think the price is way too high, but you are paying for a new Node (12nm) First Nvidia card with HBM memory (HBM2), and it has over 5100+ cuda cores.

I think it some ways it is worth $3k. I can't afford it, but I can tell you that this card will play anything at 4k60hz easily.

If people are willing to spend $1300 on a new phone, then this card will sell like hot cakes.

Now the the kicker is, who is willing to spend $6000 for NVLINK Sli. With the NVLINK costing $599!!
 

realworld

Limp Gawd
Joined
Apr 24, 2016
Messages
297
It’s a shame they won’t tell us if it’s using HDMI 2.0 or 2.1. Here’s to hoping Ampere uses 2.1.
 

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
With as much work as engine developers are putting into getting it working on DX12 and Vulkan, I don't see that quite yet. This card lacking the SLI connector isn't that big of a deal, it's clearly meant more for development work, with a bone thrown to wealthy gaming enthusiasts.
I hope your post was sarcastic since there is essentially no work on it and no demand ;)
 

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
Oh I'm getting this card if its the full V100, just for that reason lol. Cause its the biggest chip ever made that actually works!
Only broken part is a stack of HBM2/1024bit controller. Quite funny considering the hype of that.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
14,488
I hope your post was sarcastic since there is essentially no work on it and no demand ;)
Epic made some AMD fans mad for getting SLI working first...

DICE has it working on Frostbyte for EA...

Those are a few big name examples. Honestly, I just think we're in a dry spell, which could be predicted with Vulkan and DX12 coming online, and AMD and Nvidia have set it aside as engine developers catch up. Reality is that the fastest ain't fast enough, and mGPU is relatively cheap.
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
14,488
Well maybe 2 in NVLINK will do it!

TBH I would think at that point we would also need faster CPU's to try to sustain 120hz in 4k LOL
2x1080 Ti's can do it, an they're not the fastest we've got; and if platform bandwidth is an issue, Intel's HEDT CPUs can be properly overclocked, for a penny. Supposedly they'll only get faster when they get around to getting their next architecture out.
 

Fixall

[H]ard|Gawd
Joined
May 17, 2011
Messages
1,384
Well maybe 2 in NVLINK will do it!

TBH I would think at that point we would also need faster CPU's to try to sustain 120hz in 4k LOL
Shit, I'd settle for 1440 @165hz (glares at his stupid Maxwell Titan X).
 

Nightfire

2[H]4U
Joined
Sep 7, 2017
Messages
2,411
I think it some ways it is worth $3k. I can't afford it, but I can tell you that this card will play anything at 4k60hz easily.
An Xbox 1X plays some game a 4k60. I would hope to christ his could.

No doubt there will be an 1180ti in a couple of months for half the price or less.
 

Nightfire

2[H]4U
Joined
Sep 7, 2017
Messages
2,411
Well maybe 2 in NVLINK will do it!

TBH I would think at that point we would also need faster CPU's to try to sustain 120hz in 4k LOL
No. 120 fps in 4k is really no more intensive on the cpu than 120 fps in 1080p.

144hz/165hz at 1080p or 1440p require the most cpu power.
 

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
Epic made some AMD fans mad for getting SLI working first...

DICE has it working on Frostbyte for EA...

Those are a few big name examples. Honestly, I just think we're in a dry spell, which could be predicted with Vulkan and DX12 coming online, and AMD and Nvidia have set it aside as engine developers catch up. Reality is that the fastest ain't fast enough, and mGPU is relatively cheap.
EPIC games are?

And the BF series at this point is dead.

The point is gaming is moving to mobile and small form factors besides that people dont want the issues of SLI/CF and having a useless GPU in 99.9% of all games.

Indie is king, AAA at this time outside a very few is just a joke. How much did the lastest AAA games sell? 200-300K units?
 

IdiotInCharge

[H]F Junkie
Joined
Jun 13, 2003
Messages
14,488
EPIC games are?

And the BF series at this point is dead.

The point is gaming is moving to mobile and small form factors besides that people dont want the issues of SLI/CF and having a useless GPU in 99.9% of all games.

Indie is king, AAA at this time outside a very few is just a joke. How much did the lastest AAA games sell? 200-300K units?
I mean, you can attack EA, but they sell games, and you can attack Epic, but studios license their engine- and those are just two.
 
Joined
Oct 16, 2016
Messages
535
Well now, just when I thought prior Titan cards were insanely priced... at the very least, this proves that the wait for Volta in consumer hands is technically over. Bring on the benchmarks.

I held on to my $315 GTX 980 (that was basically GTX 970 pricing at the end of 2015), resisted the allure of Pascal, and held off for Volta. I might not have to wait much longer, and I'm gonna need all the single GPU performance I can get for VR's sake. (1080p120 is already something I don't have too much trouble with in most games, but VR is something else entirely.)

What I'm really wondering, at our sorta consumer level, is how much Volta has to offer over Pascal on an architectural level. They kept hyping up Pascal's enhancements over Maxwell for VR in particular; I'm expecting some additional big stuff for Volta that could widen the gap a bit more.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
5,193
Impressive! Gotta love Nvidia for this one. While expensive it probably packs a very huge punch. Is this a fully functional V100 chip and exposed with the gaming drivers? Limitations? The price for this is actually rather kick ass for what you are getting - just not sure what is if anything is being restricted.

For those that guess that Volta will be available for the consumer this year - my hat is off to you. There is probably not a monitor made that this thing cannot drive at full speed. :D
 

The Cobra

2[H]4U
Joined
Jun 19, 2003
Messages
2,721
After I spent $3500 on my board, CPU and two Vega FE's. One card alone is $3k....Dayum. Still a sexy card and would love to play with it if I had the scratch.
 

Lmah2x

Gawd
Joined
Apr 3, 2014
Messages
929
Waiting for benchmarks before I make any stupid decisions. Honestly if it's not double the Titan Xp I probably will not buy it, I have a feeling it will be around 60% faster. This will probably be the top card for 2-3 years though since Nvidia hasn't updated their roadmap in awhile.
 

Krenum

[H]F Junkie
Joined
Apr 29, 2005
Messages
15,972
Hopefully this will run all my games at 60 fps on my 1080p monitor. Wish me luck!
 

Sprayingmango

[H]ard|Gawd
Joined
Jan 20, 2012
Messages
1,259
Hilarious to see that some people in this thread do not understand the purpose of this card. It's also hilarious that AMD is somehow still in business....their Vega cards can't even compete with Pascal and NVidia is already on to their next gen cards lol
 

Nightfire

2[H]4U
Joined
Sep 7, 2017
Messages
2,411
Any reason why nVidia is using HBM? Seems that AMD got a lot of crap here for doing the same.
 

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
56,392

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
Any reason why nVidia is using HBM? Seems that AMD got a lot of crap here for doing the same.
Because its a V100 card targeted for something else than gamers and not sold for 500$. Its the poor AI researchers card with 110Tflops for 3000$.

The hilarious part is that it got one broken HBM2 stack/controller while the rest of the chip is intact. Perhaps why HBM3 seems to be ditching the stupid parallel bus and try to copy HMC with SerDes whenever it comes in a few years.
 

razor1

[H]F Junkie
Joined
Jul 14, 2005
Messages
10,155
Only broken part is a stack of HBM2/1024bit controller. Quite funny considering the hype of that.

Well yeah that is what I'm waiting for, what will that 100 gb/s do over the Titan Xp, I'm sure Volta has other things to compensate for the small increase in bandwidth but still.

Got over the buying itch last night lol, I'll wait and confirm, I just really want that chip......
 
Top