Upcoming GTX 750 Ti tested

[Enigmoid]

Sigh. That is an entirely different benchmark.

Fire Strike test runs in 1080p. It use more VRAM. It have more visual graphics running, like tessellation and shadows.
Whatever. Believe what you guys want. I`m not being part of this discussion anymore.

There is still a small difference when you realize that the 640 is clocked ~6% higher.

 

[ams23]

Understand: less complex cores don't mean weaker cores.

There is no evidence to suggest "less complex" CUDA cores for GTX 750 Ti. Even the ultra low power Kepler.M SMX is not less complex than other Kepler SMX (other than the fact that TMU and ROP throughput is halved, which is not the case with this rumored GTX 750 Ti).

 

[ams23]

GTX 650 Ti Boost scores 22% more than GTX 650 Ti. Why?

A) GTX 650 Ti Boost is clocked 12% higher than GTX 650 Ti.
B) GTX 650 Ti Boost have 24 ROPs while GTX 650 Ti have 16 ROPs.
C) Memory bandwidth is moot on 3DMark Performance preset. Sure you can add a couple of GPU points but it does not impact the 3DMark score because the memory bus is not being capped out.

At the end of the day, the GTX 750 Ti perf. improvement in 3dmark (relative to GTX 650 Ti) is limited by the memory bus interface. The available mem. bandwidth is tied to the memory bus interface, and the ROP throughput is tied to the memory bus interface.

 

[Cloudfire777]

Here is the first Maxwell chip which will be released in February!

GM107

Comparison with GK107

Notice the core count increase. Notice the increase in die size.

 

[iMacmatician]

Here is the first Maxwell chip which will be released in February!

GM107

I've always had a feeling that GM107 would be NVIDIA's "Bonaire" in terms of similar die size (and performance).

 

[Arkadrel]

I've always had a feeling that GM107 would be NVIDIA's "Bonaire" in terms of similar die size (and performance).

Their both around 160mm^2 chips, and both on 128bit busses.
Going by this thread, so that compairison seems reasonable enough.

When does the NDA lift? anyone know?

 

[Cloudfire777]

Can anyone guess the power consumption difference between Xeon E3 and i7-3960X running Furmark?
If anyone knows, I can give an approx answer on how much less power a GTX 750 Ti use

 

[Mondozei]

Via Sweclockers(who have been doing great on this story):

Maxwell 750 leaked via Chinese etailer Tmall.

http://www.sweclockers.com/nyhet/18249-geforce-gtx-750-pa-bild-med-specifikationer

I followed their link and it seems the picture was removed? Anyway, SweClockers saved the promotional picture. Looks sleek!

 

[JDG1980]

Via Sweclockers(who have been doing great on this story):

Maxwell 750 leaked via Chinese etailer Tmall.

http://www.sweclockers.com/nyhet/18249-geforce-gtx-750-pa-bild-med-specifikationer

I followed their link and it seems the picture was removed? Anyway, SweClockers saved the promotional picture. Looks sleek!

If the TDP is low enough not to need an auxiliary power connector, then fanless cards from aftermarket vendors should be feasible.

 

[f1sherman]

You have every single server/workstation card in fanless edition.
And that includes AMD's dual Tahiti S10000.

OTOH I don't think there is >=128bit MC card on the market that is truly fanless, i.e. can sustain 100% GPU workload without additional case airflow.

 

[OCGuy]

What a cool look ahead to Maxwell. However, it will be hard to extrapolate this info to the 8 series, because the 20nm real Maxwell allows a 1.9x density increase, and a 30% speed bump per TSMC.

FYI this "leak" was a play out of AMDs recent trick book. Anyone who thinks this wasnt intentional to dampen the Mantle spotlight is fooling themselves.

 

[lagokc]

Compared to Kepler, these new NVIDIA Maxwell chips would feature a slightly enhanced SMX unit structure and made some optimizations so that the double-precision algorithm logic unit can now be used for the Single-precision too supporting all the new SP instructions.

Read more: http://wccftech.com/alleged-nvidia-...project-denver-maxwell-refresh/#ixzz2s4wpt1a7
​

Any word on whether nVidia has added BIT_ALIGN_INT to Maxwell's instructions so the miners can start using nVidia cards instead of just buying up all the AMD cards? To not sell to this market seems negligent on nVidia's part.

 

[iMacmatician]

From UK Gaming Computers: "Nvidia GTX 750 & 750 Ti Benchmarks."

Another interesting feature is neither card has a power connector drawing all the required power from the PCI-E Slot.
[…]
Another note is that both these new cards are factory overclocked.

So, how do they actually do? The numbers speak for themselves;

GPU			 3DMark 11 Score
Nvidia GTX 750		 5250
Nvidia GTX 750 Ti	 5963
Nvidia GTX 650 Ti	 4819
Nvidia GTX 660 2GB	 6727

OC to non-OC isn't the fairest comparison, so I'd really like to know what the stock clocks are.

 

[Cloudfire777]

From UK Gaming Computers: "Nvidia GTX 750 & 750 Ti Benchmarks."

Those cards run really high clocks. 1163MHz with a small card with just a tiny fan? Is this because Maxwell output less heat I wonder?

Anyhow looking at 3DMark11 Graphic score:

GTX 650: 2759
GTX 750: 4870 (+75%)
GTX 650Ti: 4344
GTX 750Ti: 5611 (+29%)

This could be pretty interesting

 

[iMacmatician]

Those cards run really high clocks. 1163MHz with a small card with just a tiny fan? Is this because Maxwell output less heat I wonder?

Maxwell, at least the GM107 chip, appears to use very little power. Since an OCed version is still ≤ 75 W I suspect a stock clock version could be considerably below 75 W.

 

[Cloudfire777]

Videocardz have just listed a bunch of GTX 750 and GTX 750 Ti models

http://videocardz.com/49544/msi-geforce-gtx-750-ti-maxwell-series-detailed

 

[Gloomy]

I haven't been keeping up with this. Would this make a good PhysX card?

 

[Pottuvoi]

I haven't been keeping up with this. Would this make a good PhysX card?

If rumors are true about the ARM core, it would be exceptional.
Otherwise it's just a good/great.

 

[coffeejunkee]

Seems the rumor about no pci-e connector was wrong: http://www.techpowerup.com/197807/gtx-750-taken-apart-sips-power-from-a-single-6-pin-connector.html

Looking at that pcb this better be cheap.

 

[Arkaign]

So the 750 (non-ti) is faster than the 650ti (non-boost)? That's actually pretty sweet (for cheap gaming boxes, no PCIe power, etc).

The 650ti Boost is kind of irrelevant in the discussion because it was a one-off product that I can't find ANYWHERE except used. Me and a friend were trying to help a guy build a midrange gaming PC, and the stupid GPU prices right now are insane. He wanted to buy a 650 (non-Ti!) for $145, and I was like NOOOOOOOOOOOOPE. But the best I can find is a $189 660 vanilla OC. Where the heck is the 650 ti boost for $139? Where is the 660 for $149? What's going on?

 

[Arkadrel]

So the 750 (non-ti) is faster than the 650ti (non-boost)? That's actually pretty sweet (for cheap gaming boxes, no PCIe power, etc).

http://www.techpowerup.com/197807/gtx-750-taken-apart-sips-power-from-a-single-6-pin-connector.html

Picture of a 750 (non ti) with a power connector pin:
http://www.techpowerup.com/img/14-02-12/28c.jpg


Even the 750 (non TI) uses a 6-pin power connector.
Which likely means TDP over 75w.

I mean the 260 (90w? TDP) is abit faster than the 750ti (by rumors).

Now we know the 750 (non ti) has a power connector as well.

I dont think Maxwell judgeing from this card will be all that much more power effecient than the 260 card is.

 

[f1sherman]

Only non-reference, OC-ed partner boards will come with 6-pin power connector for the sake of overclocking.

THE CHIP IS CRAAAZY:

Core: 1163 MHz
VRAM: 1253 MHz
GPU Temperature: 35 °C
Fan: 30%
GPU load: 99%
Memory Controller Load: 74%
PowerConsumption: 51% TDP

 

[Mondozei]

Core: 1163 MHz
VRAM: 1253 MHz
GPU Temperature: 35 °C
Fan: 30%
GPU load: 99%
Memory Controller Load: 74%
PowerConsumption: 51% TDP

Looks pretty sweet if true.

 

[coffeejunkee]

Only non-reference, OC-ed partner boards will come with 6-pin power connector for the sake of overclocking.

GPU Temperature: 35 °C
GPU load: 99%

That TPU picture looks about as reference as reference can be.

And I'd like to see a bit more sustained load before I believe those temperatures.

 

[Arkaign]

http://www.techpowerup.com/197807/gtx-750-taken-apart-sips-power-from-a-single-6-pin-connector.html

Picture of a 750 (non ti) with a power connector pin:
http://www.techpowerup.com/img/14-02-12/28c.jpg


Even the 750 (non TI) uses a 6-pin power connector.
Which likely means TDP over 75w.

I mean the 260 (90w? TDP) is abit faster than the 750ti (by rumors).

Now we know the 750 (non ti) has a power connector as well.

I dont think Maxwell judgeing from this card will be all that much more power effecient than the 260 card is.

Also from what I've been able to tell from everywhere, this is sort of Maxwell Beta. Some new design, but on an outgoing process.

The real next-node Maxwell (and whatever AMD comes up with of course) will be a gigantic leap in perf/watt. We've been stuck on 28nm for ageeeeeeeeeees.