Wednesday, September 11th 2013

NVIDIA GeForce GTX 750 Ti Detailed?

How far can you strip down a GK104? Very far. Reports from the Chinese press and the whirring rumor mill there speaks of a new mid-range graphics card SKU taking shape at NVIDIA, named GeForce GTX 750 Ti, which is being designed to succeed the GeForce GTX 650 Ti Boost, outperform the GeForce GTX 660, and become NVIDIA's posterboy for this year's AAA shooter releases - Call of Duty: Ghosts, and Battlefield 4, when the company expects a tiny spike in GPU sales.

Based on the same G92 GK104 silicon as several other 600- and 700-series GeForce GTX products, the GTX 750 Ti will be configured a notch above the GK106-based GeForce GTX 660. To begin with, it's expected to feature 960 CUDA cores, a notch below the 1152 cores on the GeForce GTX 760. It has the same 80 TMUs as the GTX 660, but 32 ROPs and a 256-bit wide memory interface (compared to 24 ROPs and 192-bit on the GTX 660). Clock speeds are expected to be higher, too, at 1033 MHz core, 1098 MHz GPU Boost, and 6.00 GHz (GDDR5-effective) memory. There's no clarity on when exactly NVIDIA plans to launch the SKU, but we expect that to happen before AMD's late-September high-end extravaganza.

Source: MyDrivers
Add your own comment

18 Comments on NVIDIA GeForce GTX 750 Ti Detailed?

#1
jigar2speed
Based on the same G92 GK104 silicon
I see what you did there :toast:
Posted on Reply
#2
NC37
by: jigar2speed
I see what you did there :toast:
I approve of this completely :toast:
Posted on Reply
#3
Fiery
FinalWire / AIDA64 Developer
In case the chip would really have a 11C7 device ID, then it is based on GK106, and not on GK104.
Posted on Reply
#4
marsey99
by: Fiery
In case the chip would really have a 11C7 device ID, then it is based on GK106, and not on GK104.
but where is the fun in that?

i mean its not like amd ever reused old cores and renamed them....:roll:
Posted on Reply
#5
birdie
A ridiculous fake - see the driver version string - it lists two different drivers versions.

I don't know what this news is doing on TPU - it's shameful really.
Posted on Reply
#6
Easy Rhino
Linux Advocate
by: birdie
A ridiculous fake - see the driver version string - it lists two different drivers versions.

I don't know what this news is doing on TPU - it's shameful really.
even if it is fake it is still fun to talk about.
Posted on Reply
#7
Casecutter
Kepler – fused to the enth-degree…
AMD, like they ever imagined deriving 6 models from chips over a year and half old. I have said this is the slight-of-hand Nvidia has perfected in their first foray in buying complete wafers... Incremental differences for those who can be confounded by something new.

The early word was this was a GK106 part, but that makes no sense unless Nvidia has been concealing the fact that the GK106 part has had 4 memory controller in it all this time...
Posted on Reply
#8
JDG1980
What kind of TDP are we looking at here? Any chance it's low enough that a good passive cooler like the Accelero S1 Plus might be able to cope?
Also, when is Nvidia going to get with the program and include two DisplayPort outputs on their cards?
Posted on Reply
#9
arterius2
by: birdie
A ridiculous fake - see the driver version string - it lists two different drivers versions.

I don't know what this news is doing on TPU - it's shameful really.
what the heck are you talking about?
looks fine to me.
Posted on Reply
#10
Fiery
FinalWire / AIDA64 Developer
by: arterius2
what the heck are you talking about?
looks fine to me.
It says "9.18.13.2680 (ForceWare 326.41)". In the version number "9.18.13.2680", the last 5 digits (ignore the dots) show the ForceWare version. So from the version number it should be ForceWare 326.80, while in the brackets there's a different version number. That's quite suspicious.
Posted on Reply
#11
zsolt_93
Fiery u are always suspecting things... Its ok just its a dev driver based on 326.80 that works with the new GPU, there was a 326.40 beta out, but not a .41 from what i remember. Why would anyone fake the driver version though?
Posted on Reply
#12
birdie
by: zsolt_93
Fiery u are always suspecting things... Its ok just its a dev driver based on 326.80 that works with the new GPU, there was a 326.40 beta out, but not a .41 from what i remember. Why would anyone fake the driver version though?
Are you dumb? The driver string always lists two identical drivers versions, like:



9.18.13.2018 Forceware 320.18

The screenshot in the news is a total fake.
Posted on Reply
#13
zsolt_93
Have you ever used an unreleased card/driver? I am sure you haven't so you can't say this is fake, only people who do review can confirm or infirm this rumor, they know how these drivers work, usually fakes present visible photoshopping or just a bios name change but never seen someone faking the driver version.
Posted on Reply
#14
atikkur
we want only 800 series...
we want only 800 series...
we want only 800 series...
we want only 800 series...
we want only 800 series...
we want only 800 series...
Posted on Reply
#15
HumanSmoke
by: Casecutter
Kepler – fused to the enth-degree…
AMD, like they ever imagined deriving 6 models from chips over a year and half old. I have said this is the slight-of-hand Nvidia has perfected in their first foray in buying complete wafers... Incremental differences for those who can be confounded by something new.
It's actually seven ( GTX 660 OEM, 660 Ti, 670, 680, 690, 760, 770)...or eight if this card is a reality- the same number of models as Tahiti ( 7870XT, 7950/7950 Boost, 7970, 7970 GE, 7990, 8950 OEM, 8970 OEM, 8990 OEM)

Marketing sleight (yes, that's how it's spelled) of hand is far from an Nvidia-only affliction
Posted on Reply
#16
kn00tcn
by: zsolt_93
Have you ever used an unreleased card/driver? I am sure you haven't so you can't say this is fake, only people who do review can confirm or infirm this rumor, they know how these drivers work, usually fakes present visible photoshopping or just a bios name change but never seen someone faking the driver version.
:wtf:





look how stupidly easy it is to change your driver version, how can you trust any leak so much in the first place?
Posted on Reply
#17
Casecutter
by: HumanSmoke
7950/7950 Boost, 7970, 7970 GE
Although, they are the same chips, nothing fused of or gleded... I should've said derivatives not models.
Posted on Reply
#18
xorbe
Any chance this card will actually happen in the near future? One of my machines needs a new card. The lowest 700 series card is $250 and 170W. The "next card down" is the 660 non-Ti for $200 and 140W, which is more like what I'm looking for. But if it's about to be replaced ...
Posted on Reply
Add your own comment