Conclusion

Perhaps most importantly of all for us, mainstream Maxwell charts what's possible, and probable, from Nvidia's desktop GPU stable in 2014.

The GeForce GTX 750 Ti is the unlikely GPU candidate for the all-new Maxwell architecture that is primed to power notebook, desktop and workstation graphics cards for the next two years. Maxwell is a thorough reworking of the Kepler architecture widely in evidence today, with specific focus on improving energy efficiency and per-core performance.

We mentioned two years ago that 'Kepler is Fermi done right. It is designed to excel in one metric: performance per watt'. That thinking continues to permeate through Nvidia as Maxwell arguably constitutes a still-bigger performance-per-watt improvement over Kepler.

By introducing Maxwell on a mainstream GPU Nvidia is showing where its priorities lie. Energy-efficient processing is a must in TDP-constrained environments such as laptops and high-end tablets... and it wouldn't surprise us in the slightest if Maxwell is slow to infiltrate the high-end enthusiast space currently plied by Kepler-based 700-series GPUs.

Perhaps most importantly of all for us, mainstream Maxwell charts what's possible, and probable, from Nvidia's desktop GPU stable in 2014. With the rather large assumption of Maxwell scaling nicely to multiple-GPC GPUs used in premium cards, it is reasonable to envisage a 150W, smaller-die GPU producing GTX 780-like numbers and a 250W part increasing top-end perf by 30-50 per cent.

But vapid conjecture doesn't affect the bottom line. The Maxwell architecture is able to do more with less silicon real estate than ever before, thus likely helping Nvidia's gross profit margins compared to AMD's current price- and performance-comparable GPUs that requires a 43 per cent larger die. The slower-performing Radeon R7 260X uses a slightly larger die than Maxwell, as well. Over time, preferably this year, AMD will need to match Nvidia's improved performance-per-watt and performance-per-mm² metric if it is to continue competing successfully at every price point.

Yet there remains reservation to our praise. Nvidia is launching the GTX 750 Ti GPU at $149, or £115, and that pricing is (un)comfortably higher than AMD's Radeon R7 260X and in line with the recently-released R7 265 GPUs. AMD's used the pricing hammer to good effect, and the company can rightfully claim to have the fastest gaming GPU south of $150.

Nvidia's GeForce GTX 750 Ti certainly won't set any pure performance records, but, sensibly equipped with a 2GB framebuffer, if you're after a mainstream graphics card that can play some of the latest games at a full-HD resolution allied to decent-quality settings, all with the minimum noise and power consumption, the first Maxwell-powered GPU makes a convincing case for your money.

The Good

Excellent power consumption
Cool and very quiet
Significant increase in perf-per-watt
Primed for small-form-factor systems

The Bad

No SLI support
No bundled game(s)

HEXUS.awards

Nvidia GeForce GTX 750 Ti 2GB

HEXUS.where2buy

A wide range of GeForce GTX 750 and GeForce GTX 750 Ti cards are available to buy from Scan.co.uk.*

HEXUS.right2reply

At HEXUS, we invite the companies whose products we test to comment on our articles. If any company representatives for the products reviewed choose to respond, we'll publish their commentary here verbatim.

*UK-based HEXUS community members are eligible for free delivery and priority customer service through the SCAN.care@HEXUS forum.

Login with Forum Account

Don't have an account? Register today!

Looks a decent bump in efficiency but I am not sure about the pricing TBH.

The Maxwell architecture is able to do more with less silicon real estate than ever before, thus likely helping Nvidia's gross profit margins compared to AMD's current performance-comparable GPUs that require almost twice the die-size.

Emm,what??

The GM107 is 148MM2 and the Pitcairn GPU is 212MM2. That is 43% bigger and the R7 265 uses a die salvaged chip too. Please can you change that part of the review as it is massively inaccurate.

If it had the bigger bus cards in that price bracket have, maybe. But that tiny little 128-bit bus is a hell of a bottleneck on this, and at that price it's no sale.

The cooler looks quite, well unpleasant. Surely a larger single slot design would have been just as effective, more like that of an old single slot 6670, Then at-least it may be more appealing to users of Small Form Factor PC with highly limited space

I know I am being a bit of a git about this but the fully enabled Pitcairn die(212MM2) is being used in the R9 270X,which is another level of performance above the GTX750TI. For instance,the R9 270X is around 35% faster in Crysis3 and 45% faster in BF4:

http://hexus.net/tech/reviews/graphics/66145-nvidia-geforce-gtx-750-ti-28nm-maxwell/?page=7
http://hexus.net/tech/reviews/graphics/66145-nvidia-geforce-gtx-750-ti-28nm-maxwell/?page=5

So in terms of die area,they are quite comparable for performance/mm2 IMHO,although Pitcairn has to include a 256 bit memory controller in the die which takes up more space.

I think the better comparison might be to the Bonaire XT GPU in the R7 260X which is around 160MM2 including a 128 bit memory controller,and is anything from similar speed to upto 15% to 20% slower. However,part of the die area is taken up by the Tensilica audio DSP too which Pitcairn lacks. OTH,even compared to a die salvaged Pitcairn graphics card,Bonaire does poorly regarding power consumption,especially so when compared to the GM107. It does make me wonder whether that DSP is not helping for some reason.

Ultimately,the GM107 gives Bonaire XT a bit of a bloody nose and performance/watt looks very good as a whole.

What CAT-THE-FIFTH said :)