Nvidia Announces The GTX Titan X; Features Maxwell, 8-Billion Transistors

[turkey3_scratch]

People, this is not a consumer gaming card. It's for extreme rendering and modeling. Rendering companies will make good use of this.

 

[TNT27]

People, this is not a consumer gaming card. It's for extreme rendering and modeling. Rendering companies will make good use of this.

Uhm, no, The Titan series are marketed towards gamers, they have a whole different line of similar gpus for rendering/modeling applications.

Titans are usually used for multi monitor setups or high resolutions/Extreme Post Processing

 

[Grognak]

Will this do like the original Titan and lose to a 980 Ti, too?

 

[090909090909]

for people who are taking about HBM. Memory bandwidth isn't that important. The GTX 980 beats R9 290X at 4K though the R9 290X has more bandwidth. and you should know that with 384-bit memory bus interface and 12 GB, the Titan X bandwidth should hit the roof and maybe more tham 4 GB HBM ( memory bandwidth increases with the increase of both of memory capacity and memory bus interface. GTX 970 has 3.5 GB with a 196 GB/s bandwidth and 0.5 GB with a 28 GB/s bandwidth of despite the two parts has a 384-bit memory bus interface and notice that it should a 4 GB with a 228 GB/s bandwidth),but a 12 GB GDDR5 memory will increase power consumption.

HBM doesn't just increase bandwidth significantly, it also increases speed in a huge way. (Same kind of thing as comparing gddr3 to gddr5) All while using less power

bandwidth is speed. when I say that my car speed is 60 Km/h it means that my car goes 60 km at the time of one hour. bandwidth is (how "fast" a memory can transfer data to or from the GPU) or (how much data a memory can transfer to or from the GPU at the time of a second(like my car)). what I said before won't work all the time. because In the future, when I want to increase bandwidth, I won't increase capacity because that will increase the cost and power consumption and I will simply increase the memory bus interface by upgrading the memory type and that is why HBM is the future,but what I said before will work in current situation and a 12 GB Gddr5 with a 384-bit will be very fast and "maybe" beat 4 GB HBM

 

[closetwhisperer1]

It is definately Nvidias best looking card. I mean it is mouthwatering.

 

[none12345]

So, that means maxwell with 12 blocks instead of 8 or 4. Which would be 7.8 billion based off the 5.2 number, but i guess we can round that up, even tho you didn't round the other number down.

Id make the 11G fast +1GB slow joke(it wouldnt be 11.5 becasue 12 gig divides to 1 gb chips on 12 channels) but these will probably only be fully functioning 12 block dies. Unless they add on an extra one so they have something to disable in the case of a bad one. In that case there actually could be an 11+1 issue. Man wouldnt that suck!

Tho that does beg the question. What will they do with defective dies. I can see them cutting some down to 980 specs. And if that is the case, depending what they cut, they might end up with some kinda wierd slow memory split in die harvested 980 conversions.....even more possible fun!

 

[An AFK User]

Nvidia 's counter to AMD's R9 295X2 looks great. I sure hope the Titan X's TDP isn't too high. I wonder what Nvidia's counter to the rumored "Hydra Cooler" would be.

 

[everygamer]

All I can say is take the Kid and the Kites out of the demo, and drop in my Skyrim character and lets have some fun

 

[Bob Austin]

I'll wait till the 390x with hbm to come out, then nvidia will slash the price of the titan. Then I'll buy the titan x, and 2 weeks later gtx 980 ti comes out and then I will kick myself!

 

[090909090909]

So, that means maxwell with 12 blocks instead of 8 or 4. Which would be 7.8 billion based off the 5.2 number, but i guess we can round that up, even tho you didn't round the other number down.

Id make the 11G fast +1GB slow joke(it wouldnt be 11.5 becasue 12 gig divides to 1 gb chips on 12 channels) but these will probably only be fully functioning 12 block dies. Unless they add on an extra one so they have something to disable in the case of a bad one. In that case there actually could be an 11+1 issue. Man wouldnt that suck!

Tho that does beg the question. What will they do with defective dies. I can see them cutting some down to 980 specs. And if that is the case, depending what they cut, they might end up with some kinda wierd slow memory split in die harvested 980 conversions.....even more possible fun!

this is in the GTX 970 not the GTX 980. the GTX 980 uses normal 4 GB.

 

[photonboy]

Blah, blah "memory is faster.. better.. " blah, blah.

Here's a TIP for people who over-analyze how different types of memory work:

Benchmarks, benchmarks, benchmarks.

 

[090909090909]

Blah, blah "memory is faster.. better.. " blah, blah.

Here's a TIP for people who over-analyze how different types of memory work:

Benchmarks, benchmarks, benchmarks.

that is the right thing.

 

[chicofehr]

I am sure most of the extra transistors are for the cuda cores and wont improve gaming performance much. I hope that AMD can release a 8GB 380x/390x card as they already have a 8GB 290x out now so it would be a shame if a newer card would have less then the current model. Also with 4k/1440p becoming more common, 4GB could become a bottle neck. I got my 3x1600 setup and 4GB with some of the new games would be a problem if I run with max settings.

 

[dragonsqrrl]

I am sure most of the extra transistors are for the cuda cores and wont improve gaming performance much.

I'm not sure what you mean by this, Cuda cores contribute a great deal to game performance. They're also commonly referred to as shader cores.

 

[dragonsqrrl]

People, this is not a consumer gaming card. It's for extreme rendering and modeling. Rendering companies will make good use of this.

Uhm, no, The Titan series are marketed towards gamers, they have a whole different line of similar gpus for rendering/modeling applications.

Titans are usually used for multi monitor setups or high resolutions/Extreme Post Processing

The Titan's target a prosumer market between the high-end GeForce lineup and Quadro. It's primary strengths target developers and researchers that depend on double precision performance, but don't need ECC or Quadro driver specific optimizations and validations. If you just need FP64 and a large volume of fast memory they offer a much better price/performance ratio than Quadro or Tesla at the time of launch.

I honestly don't think they make much sense as strictly gaming cards, with perhaps the exceptions of the initial launch of the original Titan, and the upcoming launch of the Titan X, when they're also the fastest gaming cards on the market (for those with lots of money and live on the bleeding edge). But in the case of the Titan X most people must realize that there's probably going to be a 6GB GM200 card to top off Nvidia's regular consumer gaming lineup sometime down the road.

 

[chaosmassive]

3 or 4 SMM will be disabled on Titan X with enabled FP64

then in few months popped out GTX 980 Ti with full SMM enabled and FP64 disabled

next few months they will sell Titan Black X with full SMM and FP64 enabled

well done Nvidia..!

 

[alextheblue]

12 GB Gddr5 with a 384-bit will be very fast and "maybe" beat 4 GB HBM

Not unless they're running that GDDR5 at some really insane speeds. In the 390X, the HBM memory is expected to have a bandwidth of 640GB/s. Even the base 390 is expected to push 480. Not to mention that all the new cards are going to be at least GCN 1.2 so they get the lossless color compression (ala Tonga) across the board.

Now, memory bandwidth alone doesn't decide a victor. So that doesn't mean that these cards are guaranteed a victory. But as far as memory speed goes? There's no way HBM is going to lose to GDDR5 in such a wide configuration.

 

[Memnarchon]

1st gen HBM is limited to 4GB VRAM. So since it has 12GB VRAM, I doubt its gonna have HBM...

That and it's already known that Nvidia will be sticking with GDDR5 for Maxwell.

And HBM is currently limited to 4GB only on the rumored 4096-bit interface on the 390X. Of course there are other ways to achieve greater capacities, but it'll cost you.

I was talking about 1st gen HBM that will be available soon.
At the moment there are no known ways to achieve greater capacities unless you are talking about 2nd gen of HBM (or you know something more that the media don't know ).
HBM overall speccs

 

[fuzzion]

If you have to ask about the price then you cant afford one.

 

[dragonsqrrl]

1st gen HBM is limited to 4GB VRAM. So since it has 12GB VRAM, I doubt its gonna have HBM...

That and it's already known that Nvidia will be sticking with GDDR5 for Maxwell.

And HBM is currently limited to 4GB only on the rumored 4096-bit interface on the 390X. Of course there are other ways to achieve greater capacities, but it'll cost you.

I was talking about 1st gen HBM that will be available soon.
At the moment there are no known ways to achieve greater capacities unless you are talking about 2nd gen of HBM (or you know something more that the media don't know ).
HBM overall speccs

You could widen the interface, which is what I was referring to.

 

[iamrayth]

Assuming 384 bit bus, the memory would have to be clocked at 12.5 ghz minimum to equal the rumored bandwidth of 390x. And memory capacity has no effect on gpu bandwidth, it is determined by the number of memory chips and clock speed.

 

[LordConrad]

I can't wait to play Solitaire on this. :-D

 

[VIPChristian]

That was awesome, but how will I afford one, Great Great Depression continues.

 

[siman0]

The Titan line is aimed at enthusiast grade gamers. They are not made for graphical rendering but do have the power to do so. The Titan line is made for gamers that much is evident of the printing on the box. Nvidia just thought it good to make another line for the enthusiast market. Its not made for everyone to have.

 

[Blueberries]

Yay, a new entry-level workstation card!