Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [WCCF]NVIDIA GeForce Titan Is Not GTX 780 – Performance Surpasses GTX 690
New Posts  All Forums:Forum Nav:

[WCCF]NVIDIA GeForce Titan Is Not GTX 780 – Performance Surpasses GTX 690 - Page 19  

post #181 of 360
I don't remember seeing 1400+MHz GTX 460's...
post #182 of 360
And kepler isn't fermi, there weren't 1400mhz 460s because the 460 just could not clock that high. Just like GK110 will (can be said with almost certainty seeing that GK110 teslas run at the same frequencies as 2x GK104 teslas) launch at higher frequencies and will reach higher frequencies than the 580 did.
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
post #183 of 360
Do you just not get the point I'm making here or are you being intentionally obtuse? People know the performance of a 460. It was a lower mid-range sku during the first run of Fermi. To say that there is no difference between that and the GTX680 just because they share a similar die size is just wrong. I don't care what any of you say. You can't directly compare the performance of a previous generation lower midrange card to a current high end card and then make assumptions about the performance of a future card based on that. That's just silly and inaccurate...
post #184 of 360
Quote:
Originally Posted by Majin SSJ Eric View Post

Do you just not get the point I'm making here or are you being intentionally obtuse? People know the performance of a 460. It was a lower mid-range sku during the first run of Fermi. To say that there is no difference between that and the GTX680 just because they share a similar die size is just wrong. I don't care what any of you say. You can't directly compare the performance of a previous generation lower midrange card to a current high end card and then make assumptions about the performance of a future card based on that. That's just silly and inaccurate...
The point he is trying to make is that the gtx680 was marketed as a high end card while it really only is a mid end card.
And die size and power consumption wise those 2 are really alike
Quote:
Originally Posted by Majin SSJ Eric View Post

To say that there is no difference between that and the GTX680 just because they share a similar die size is just wrong. I don't care what any of you say.
We weren't comparing performance just saying that the 680 is a mid end card marketed as a high end card.
But if you don't care then we can't really reason with you can we tongue.gif
Edited by maarten12100 - 1/31/13 at 1:53pm
post #185 of 360
@ majin

But I don't get what you're basing your claims on. I'm trying to compare the situation here from a purely technical pov to the difference between GF104 and GF110. And as you've yourself said, price isn't a factor here so neither is all this mid range/high end SKU stuff.

The facts are:

GF104 and GK104 are both first gen fermi/kepler
GF104 and GK104 are both missing GPGPU features
GF104 and GK104 are of a similar size (GF104 a bit bigger but GK104 clocked more aggressively, evens out)
GF104 and GK104 share the same placement in NV's internal code naming scheme
GF104 and GK104 both have a similar memory bus usually found on mid range NV cards
GF104 and GK104 both launched about a year before the high end GPU of the 2nd generation based on the same architecture (GF110 and GK110)
GF104 and GK104 were both mainly designed as gaming GPUs

GF110 and GK110 are both 2nd gen fermi/kepler
GF110 and GK110 have extensive GPGPU features meant for tesla products mostly
GF110 and GK110 are of a similar size (no exact info on clocks yet)
GF110 and GK110 share a similar place in NV's naming scheme
GF110 and GK110 both have a similar memory bus usually found on high end NV GPUs
GF110 and GK110 both launched a bit less than a year after the introduction of fermi/kepler
GF110 and GK110 are both designed mainly for GPGPU

I just can't see how the comparison isn't valid as long as you forget about the pricing and what the cards are actually called. We all know those mean nothing when talking about specs.

The comparison essentially really does boil down to die size, which mostly regulates the manufacturing costs of a GPU. Comparing die sizes will give you a good idea of what the card should be compared to. And since fermi and kepler are very similar architecturally you can compare them in this fashion.
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
post #186 of 360
Quote:
Originally Posted by maarten12100 View Post

The point he is trying to make is that the gtx680 was marketed as a high end card while it really only is a mid end card.
And die size and power consumption wise those 2 are really alike
We weren't comparing performance just saying that the 680 is a mid end card marketed as a high end card.
But if you don't care then we can't really reason with you can we tongue.gif

And that is where you are wrong. It wasn't just marketed as a high end card, it also performs as a high end card (something the GTX 460 never did). If you want to compare the 460 to a Kepler chip you would have to compare to the GK104 in the 660Ti which is the Kepler range's midrange chip. And before you break your fingers typing back that the 660Ti and 680 are both GK104, remember that not all GK104's can be made into 680's (due to binning for clock speeds) so there is a performance difference between the two. That is why the comparison is flawed. We don't know how high Nvidia could have binned GF104 if they had tried to make it into the GTX 580 but if they had then the comparison would be valid...
post #187 of 360
Quote:
Originally Posted by Alatar View Post

It already has cuda cores locked due to yield issues. The ones getting geforce treatment will most likely just be Tesla K20X/K20 cores that didn't make the TDP requirements (read: high leakage) and can be clocked higher while pulling lots of power.

So yes I actually do think they would sell a locked chip like this. It's exactly the same situation as the GTX 480 except that this thing is already the revised version.

And according to your logic we should never have had stuff like the 8800GTX, GTX 280 etc. They will sell it if it makes them money, directly or indirectly. In this case you get the marketing boost of having a monster card, breaking all the records etc. and also get to profit from your bad dies that can't go into enterprise products. win-win.
My logic was that they do a jump every second generation, because architecture is expensive to build.
Bender
(18 items)
 
  
Bender
(18 items)
 
  
post #188 of 360
Quote:
Originally Posted by th3illusiveman View Post

My logic was that they do a jump every second generation, because architecture is expensive to build.

Yeah I got that, however while that might generally have been true, in order to make more accurate estimations you need more info to back it up, which is where all this stuff about rejected GPUs for teslas, GK104 -> GK104 etc. stuff comes in.
Quote:
Originally Posted by Majin SSJ Eric View Post

And that is where you are wrong. It wasn't just marketed as a high end card, it also performs as a high end card (something the GTX 460 never did). If you want to compare the 460 to a Kepler chip you would have to compare to the GK104 in the 660Ti which is the Kepler range's midrange chip. And before you break your fingers typing back that the 660Ti and 680 are both GK104, remember that not all GK104's can be made into 680's (due to binning for clock speeds) so there is a performance difference between the two. That is why the comparison is flawed. We don't know how high Nvidia could have binned GF104 if they had tried to make it into the GTX 580 but if they had then the comparison would be valid...

Performs as a high end card relative to what? If you're going to say AMD, it makes no difference since the discussion was strictly about Nvidia. And even if it did, that would mean that lower end 8800GTS 320MB cards which cost around 300 bucks were high end in NV's lineup just because AMD's 2900XT sucked, bad.

When trying to compare GK104 -> GK110 transition to GF104 -> GF110 there is absolutely 0 reason take into account AMD's GPUs unless prices are a factor.

You also have to remember that there were more than one 460, and when they moved to 560s, there were even more. Comparing a fully fledged mid range chip (460) to a gimped mid range chip (660Ti) makes no sense. That's not how the chips are positioned when supposed costs are taken into account.

I don't know how many times I have to spew this in GK110 threads but the "mid range chip" stuff has nothing to do with:
  • Price
  • Performance
  • Product names
  • Competing products

What it is all about is: cost of manufacturing, die sizes (cost), the relative place of the GPU in NV's line of GPUs (internal), features (such as proper dp perf) and so on. Please do not confuse the two.
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
 
Benching
(17 items)
 
 
CPUMotherboardGraphicsRAM
[i7 5960X @ 4.8GHz] [Rampage V Extreme] [Titan 1400MHz (1500MHz bench)] [Various] 
Hard DriveCoolingCoolingCooling
[250GB 840EVO +2x SpinpointF3 1TB RAID0] [LD PC-V2 SS Phase Change] [XSPC X2O 750 pump/res] [Monsta 360 full copper + EK XT 360 + XT 240] 
MonitorPowerCaseAudio
[Crossover 27Q LED-P 1440p+ASUS 1200p+LG 1080p] [Corsair AX1200] [Dimastech Easy v3.0] [Sennheiser HD558s] 
CPUCPUMotherboardGraphics
FX 8320, FX 8350, Phenom II x2 555BE i7 3930K, i7 860, i7 4770K, 68x Celeron D CVF, commando, 2x RIVE, Z87X-OC Asus 4870x2, Sapphire 4870 
GraphicsGraphicsGraphicsGraphics
2x 5870, 5850, 5830, 5770 2x 3870x2, 3870 GTX Titan, GTX 480, GTX 590 GTX 285, GTX 260, 4x 9800GT, 8800GTX 
RAMHard DriveCoolingCooling
4x4GB vengeance, 2x4GB predatorX, 2x1GB OCZ DDR2 Intel X25-M 80GB LD PC-V2 SS Phase Change OCN Marksman 
CoolingCoolingOSPower
2x old tek slims (GPU) Various watercooling stuff win7, winxp AX1200 
Case
test bench / cardboard box 
  hide details  
post #189 of 360
I am specifically talking about performance so everything you just typed is irrelevant. You said that the jump in performance would be the same going from a 680 to a Titan as it was going from a 460 to a 580. I'm saying that the 680 performs better in the Kepler architecture than the 460 did in the fermi architecture (due to binning for higher clock speeds) so the correlation is flawed. The performance of a 680 =/= the performance of a 460. What do you not understand about that? I am specifically talking about performance which is what YOU brought up.
Quote:
Originally Posted by Alatar View Post

Performs as a high end card relative to what?

Relative to itself. The GTX 680 IS the highest performing GTX 6XX is it not????
Edited by Majin SSJ Eric - 1/31/13 at 3:10pm
post #190 of 360
Quote:
Originally Posted by Majin SSJ Eric View Post

And that is where you are wrong. It wasn't just marketed as a high end card, it also performs as a high end card (something the GTX 460 never did). If you want to compare the 460 to a Kepler chip you would have to compare to the GK104 in the 660Ti which is the Kepler range's midrange chip. And before you break your fingers typing back that the 660Ti and 680 are both GK104, remember that not all GK104's can be made into 680's (due to binning for clock speeds) so there is a performance difference between the two. That is why the comparison is flawed. We don't know how high Nvidia could have binned GF104 if they had tried to make it into the GTX 580 but if they had then the comparison would be valid...
That al rather depends on what man considers to be performing as a high end card the 7970 performaces equal as a gtx680 but those cards should be pushed on their tdp due to a large die at decent clocks.
So that they can really be called high end as in pushing the limit for a die.

Gtx 480 > 529 mm^2
Gtx 580 > 520 mm^2
Gtx 680 > 294 mm^2

So the reason why we compare it to mid end die's is
Gtx 460 > 332 mm^2

They actually just took half the waffer they would normally and clocked it up and they could do this because it was performing just as well and used just as much clocked high, as Amd's counterparts.

ps we can quit the discussion now as there is no way one side is gonna agree or not let's get back to the Titan biggrin.gif
Edited by maarten12100 - 1/31/13 at 3:18pm
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Rumors and Unconfirmed Articles
This thread is locked  
Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [WCCF]NVIDIA GeForce Titan Is Not GTX 780 – Performance Surpasses GTX 690