NVIDIA Unveils The Titan X At GDC 2015

[trick0502]

http://videocardz.com/55013/nvidia-geforce-gtx-titan-x-3dmark-performance

only 30% fast than a 980 (stock for stock). is that a bust?

 

[MrMike]

3DMark is pretty much irrelevant. Wait for an actual review from [H], Anand, etc for whether it's a bust or not.

 

[TekRok]

I didn't see this till someone else brought it to my attention but the NVIDIA CEO quoted saying that the Titan X will be faster than their previous dual GPU card the Titan Z. If so that's insane power of 1 card.

Source

Also rumored is that the stock clock speed of the Titan X is 1390Mhz which may explain which it will be faster than the Titan Z.

Is it just me or its the same length as the GTX980? The cooler size looks the same other than the 6+8 connectors.

 

[damstr]

It does look pretty much the same size at be 980.

 

[Iklos]

while i agree that 3dmark is irrelevant, nV normally have great drivers for it..

But these are early days.. but still under 40% more performance than 980 gtx will be a fail in my book. (for the price, which i think will be close to 999$)

If i am wrong and this goes for something like 800, or performs like a champ.. well.. *droll*

 

[trick0502]

while i agree that 3dmark is irrelevant, nV normally have great drivers for it..

But these are early days.. but still under 40% more performance than 980 gtx will be a fail in my book. (for the price, which i think will be close to 999$)

If i am wrong and this goes for something like 800, or performs like a champ.. well.. *droll*

http://videocardz.com/55013/nvidia-geforce-gtx-titan-x-3dmark-performance

stock for stock it shows 26%

 

[atreidesgoldenpath]

So how many of these cards will they sell? I don't see the market for a graphics card this expensive?

 

[TekRok]

http://videocardz.com/55013/nvidia-geforce-gtx-titan-x-3dmark-performance

only 30% fast than a 980 (stock for stock). is that a bust?

Called it. I knew it was 30% faster at most. What is more hilarious is Nvidia fanboys calling for AMD's destruction with Titan card release...If you look at leaked 390x specs its 30% faster than a 980 on average which would make it a direct Titan X competitor which will likely cost far less. Big win for AMD.

FYI, I use a GTX980 before people start calling me an AMD cyborg.

 

[5150Joker]

Those aren't Titan X benchmarks and it's videocardz. They're the same type of b.s. site as WCCFmakeupshittech.com. Wait for the 17th before you start spouting off nonsense.

 

[TekRok]

Those aren't Titan X benchmarks and it's videocardz. They're the same type of b.s. site as WCCFmakeupshittech.com. Wait for the 17th before you start spouting off nonsense.

So you have a better source or information in general? Youre way too optimistic.

 

[5150Joker]

So you have a better source or information in general? Youre way too optimistic.

I apply common sense unlike you. They don't have a Titan X on hand, I could go make shit up and post pretty graphs too. Nobody aside from NVIDIA and the credible reviewers (e.g. Hard|OCP) they gave samples to know the final clocks or performance of this card in games. Besides, worst case scenario if the base Titan X is 30% that puts it in line w/the previous Titan. People are expecting at best 40% increase for the R390X vs 290X which still puts it behind this thing. So again, how would this be a failure?

 

[TekRok]

I apply common sense unlike you. They don't have a Titan X on hand, I could go make shit up and post pretty graphs too. Nobody aside from NVIDIA and the credible reviewers (e.g. Hard|OCP) they gave samples to know the final clocks or performance of this card in games. Besides, worst case scenario if the base Titan X is 30% that puts it in line w/the previous Titan. People are expecting at best 40% increase for the R390X vs 290X which still puts it behind this thing. So again, how would this be a failure?

Considering what you just said, you dont know the final specs of the 390x cards, so the same applies to your statement.
I dont believe I said anything related to failure, so you might want to recheck my posts.

 

[trick0502]

Those aren't Titan X benchmarks and it's videocardz. They're the same type of b.s. site as WCCFmakeupshittech.com. Wait for the 17th before you start spouting off nonsense.

do you really think nvidia will give benchmarks? they might give specs, but i dont think they will give clocks or benchs. they will hold out to see what the 390 does. if they say our card does x, amd will say our card does x+1. its all marketing.

 

[limitedaccess]

The data from the videocardz numbers are being interpreted incorrectly.

If you compare the lower result for the GTX 980 (stock) and Titan X (stock? 1002 listed speed) the Titan X is actually ~35% faster for performance and ~40% for extreme. The previous values being listed may be mistaking absolute and relative differences. 100%/74% and 100%/71% is the relative difference not 100%-74% or 100%-71%.

 

[5150Joker]

Considering what you just said, you dont know the final specs of the 390x cards, so the same applies to your statement.
I dont believe I said anything related to failure, so you might want to recheck my posts.

I read what you wrote and you brought up "a big win for AMD" when assuming a 30% increase for the R390x vs 980. So that means 30% increase vs 980 + 24% deficit of 290x vs 980 = 54% gain over the 290X and you say I'm being optimistic?

do you really think nvidia will give benchmarks? they might give specs, but i dont think they will give clocks or benchs. they will hold out to see what the 390 does. if they say our card does x, amd will say our card does x+1. its all marketing.

What are you talking about? Reviewers will have all the information, they're under NDA as usual.

 

[-Dragon-]

Yeah 35+% improvement at likely 1080p and likely larger boosts at higher rez, definitely in for one especially if it comes it around a grand...

 

[xorbe]

Called it. I knew it was 30% faster at most.

I agree with you that it will be hard to nail that +50% linear scaling (especially if the vram speed is dropping to 6GHz like TPU data suggests), but you should wait for a better source than that. Things like high resolution should squeeze out better scaling.

 

[delita]

Yeah 35+% improvement at likely 1080p and likely larger boosts at higher rez, definitely in for one especially if it comes it around a grand...

I guess we will see with 4k benchmarks what kind of difference 12GB of RAM makes

Honestly if you ask me, if you are planning on getting this card for anything less than 1600p....you are definitely wasting your money.

 

[Dayaks]

I guess we will see with 4k benchmarks what kind of difference 12GB of RAM makes

Honestly if you ask me, if you are planning on getting this card for anything less than 1600p....you are definitely wasting your money.

DSR has made resolution statements meaningless.

The 50% increase in memory bandwidth should most certainly help higher/AA/resolutions.

Those "leak" sites have logical fallacies. I wouldn't trust those a bit.

 

[CrazyRob]

The data from the videocardz numbers are being interpreted incorrectly.

If you compare the lower result for the GTX 980 (stock) and Titan X (stock? 1002 listed speed) the Titan X is actually ~35% faster for performance and ~40% for extreme. The previous values being listed may be mistaking absolute and relative differences. 100%/74% and 100%/71% is the relative difference not 100%-74% or 100%-71%.

I was going to point this out as well. A 40% performance improvement on a card with a 50% increase in shader processors at a slightly lower clock speed seems reasonable to me, since rarely do we see linear scaling. I'd like to see that improved over time, but if these numbers are accurate, they are about what we should expect, IMO.

 

[DeadSkull]

Uh oh top right corner is missing a screw:

Sorry couldn't resist.

No backplate for a $1000 ++ card. nVidia so cheap, first screwing us out of 0.5gb ram, then screws and now a backplate.

 

[bigyellocup]

Hmmm... ill take 2

 

[TekRok]

No backplate for a $1000 ++ card. nVidia so cheap, first screwing us out of 0.5gb ram, then screws and now a backplate.

I highly doubt they would ship the card without one seeing how the 970/980 cards have the backplate included with blower style cooler. This was most likely a press piece, so id wait for the official ship model.

 

[MorgothPl]

A bit disappointed with a revealed performance graph. In 3D Mark it's only 26% faster than 980 and 40% than 970. Was hoping for 40%, so the 980 Ti would be 20% faster. RIght now we will propably have 980 Ti in like 10% diff, so OCed 980 will take it.

http://www.ocaholic.it/modules/news/article.php?storyid=12014

 

[Dayaks]

A bit disappointed with a revealed performance graph. In 3D Mark it's only 26% faster than 980 and 40% than 970. Was hoping for 40%, so the 980 Ti would be 20% faster. RIght now we will propably have 980 Ti in like 10% diff, so OCed 980 will take it.

http://www.ocaholic.it/modules/news/article.php?storyid=12014

From the article itself: "According to early details, GTX Titan X should cost US $999, which is the same price as earlier released Nvidia Titan graphics cards, but you should take all these informations with a grain of salt, as after all, these are just still rumors."

Videocardz was also off on the 980 by 20-30% when that was "leaked".

 

[Araxie]

did you check the source?.. its the same videocardz performance charts mentioned one page back.. fail charts to gain some clicks..

 

[CrazyRob]

did you check the source?.. its the same videocardz performance charts mentioned one page back.. fail charts to gain some clicks..

not only are they the same graphs, looks like MorgothPI also misinterpreted them. The graphs are using the Titan X as the baseline, so you have to recalculate the percentages if you want to know how much FASTER it is than a 980. The graphs show how much SLOWER the 980 is. If you assume the 980's scores are the 100% baseline, the stock Titan x is ~35% faster in Performance and ~40% faster in extreme.

Again, that's assuming these numbers are actually accurate.

 

[Fixall]

No backplate for a $1000 ++ card. nVidia so cheap, first screwing us out of 0.5gb ram, then screws and now a backplate.

the 12gb of ram on the back of the card is bound to get toasty so I'd be pretty surprised if the Titan X didn't ship with a backplate... I assume they just wanted to show off the PCB at GDC and left the backplate off.

 

[tajoh111]

not only are they the same graphs, looks like MorgothPI also misinterpreted them. The graphs are using the Titan X as the baseline, so you have to recalculate the percentages if you want to know how much FASTER it is than a 980. The graphs show how much SLOWER the 980 is. If you assume the 980's scores are the 100% baseline, the stock Titan x is ~35% faster in Performance and ~40% faster in extreme.

Again, that's assuming these numbers are actually accurate.

Add in these cards are conservatively clocked and you have a card that is basically 50% faster than a gtx 980. Titan X at these clocks has a 11% clock disadvantage to the gtx 980.
So add 10% these scores and you have something that is about 50% faster. If someone makes a bios mod that allows overvoltage and a waterblock, these cards are gonna fly. Ever without water, these cards should be challenging a r9 295x2 once overclocked.

 

[trick0502]

Add in these cards are conservatively clocked and you have a card that is basically 50% faster than a gtx 980. Titan X at these clocks has a 11% clock disadvantage to the gtx 980.
So add 10% these scores and you have something that is about 50% faster. If someone makes a bios mod that allows overvoltage and a waterblock, these cards are gonna fly. Ever without water, these cards should be challenging a r9 295x2 once overclocked.

And when you overclocking the 980 that advantage goes back to 30%. You have to compare stock to stock.

 

[LordEC911]

I think he's thinking of the Titan Black.

Nope. SLI doesn't scale as well as people seem to think.
I did say "many situations" when I meant "some situations." Other than that, my statement is accurate.


Cause yeah... I have no idea what I'm talking about...

The ~50% increase in transistors means next to nothing about performance, it just means that they have a 600-620mm2 GPU on their hands.

The 50% increase in actual specs is what can be used to determine performance.
That means peformance will be ~40% better than the GTX980.

Now clockspeed, the larger you make a GPU, the more variables there are with voltage and speed. You typically don't clock as high as the smaller GPUs in the family. Obviously Maxwell 2 has made some significant strides in the clockspeed department but I would expect a ~10% decrease in clocks compared to GTX980.

So 30-35% over GTX980 seems about right.
That leaves room for a GTX980Ti to be ~15% better than GTX980.

It could definitely reach up to 40-50% better in some situations but IMO, on average, it will be a bit lower than that since scaling is never 100%.
Unless there was some glaring bottleneck in GM204 that they resolved in GM200/210... But I think most of us believe that GM204 is a very well balanced GPU.

 

[Dayaks]

I like OC vs OC better. If this card is voltage unlocked (which I really doubt) or hard moddable I am going to be pretty excited and it could destroy most 980s (since most are voltage locked).

I would imagine the current maxwell bios editor would work. Should only take 5 minutes to mod the bios. I'll rush my chilled water system ahead of schedule and buy a 5960x if it is truely unleashed....

 

[Brackle]

hmmm 3dmark score only 30% faster? I expected a lil better.....but it isn't in-game benchmark so ill take it with a grain of salt.

Also who knows if a 1222mhz overclock is good or not, or even final clocks. Or if its cut down etc.

too many unknown variables to take this as accurate

Now the $629 295x2 looks like a fuckin steal when looking at that benchmark

 

[xorbe]

+30% with 6GB would have been fine.

 

[DooKey]

snip

Now the $629 295x2 looks like a fuckin steal when looking at that benchmark

Not really considering the state of CF support in games.

 

[razor1]

A bit disappointed with a revealed performance graph. In 3D Mark it's only 26% faster than 980 and 40% than 970. Was hoping for 40%, so the 980 Ti would be 20% faster. RIght now we will propably have 980 Ti in like 10% diff, so OCed 980 will take it.

http://www.ocaholic.it/modules/news/article.php?storyid=12014

err your math is wrong ...... its 34% faster then the 980 and 50% faster than the 970 in 3dmark. The 980 is 26% slower then then titan x and the 970 is 40% slower than the titan x in 3dmark.

 

[Brackle]

Not really considering the state of CF support in games.

As of right now its about as good as SLI.

 

[Knil]

If the Titan X is 35-40% faster than the 980 I'll be in one (maybe two), it will be an nice upgrade for my 780 Ti especially since I'm gaming at 4k now.

 

[n=1]

I like OC vs OC better. If this card is voltage unlocked (which I really doubt) or hard moddable I am going to be pretty excited and it could destroy most 980s (since most are voltage locked).

I would imagine the current maxwell bios editor would work. Should only take 5 minutes to mod the bios. I'll rush my chilled water system ahead of schedule and buy a 5960x if it is truely unleashed....

The vanilla Titan's voltage could be unlocked although it was supposedly unintentional, and nVidia made sure they hard locked the voltage on the Titan Black.

I really doubt they'll leave voltage unlocked for the Titan X. Which reminds me: how many of you would still be interested if the Titan X could only be unlocked through a hard mod?

 

[Liger88]

I'll wait 2-3 months and get a 6GB 980 Ti for $350 less. No way in hell it sells at the retail price of $999 even if that's true. It'll stay around $1,199 or more unless it doesn't sell fast, which it will because Nvidia.

EDIT: That brings up, why are they using so many memory chips these days anyways? They do have higher densities I would think.