Final thoughts and rating

AMD rolled out the Radeon HD 7970 GPU in December 2011. Primed for the high-end of the graphics card market and now available from just over £400, our initial recommendation of the card has been tempered by the recent arrival of the excellent NVIDIA GeForce GTX 680. Given a direct choice between the two, and it's a close one, we'd veer towards the GTX 680.

Sapphire Technologies is looking to change enthusiasts' thinking with the release of the HD 7970 OC Dual-X card, priced at around £470, or £40 above a reference card on which this model is based. Equipped with a quality heatsink-and-fan unit and shipping with reference-beating frequencies, available via two BIOS positions, there's plenty of good here.

Overclocking to the highest frequencies we've seen thus far from the underlying Tahiti GPU, Sapphire's card is about as good as it gets from an AMD front right now, we feel. But the green spectre of the GeForce GTX 680 continues to haunt AMD's finest GPU, and even though Sapphire has pulled out many of the expected stops, at a hiked price, our recommendation would still go to NVIDIA's Kepler. Perhaps Sapphire needs to release the bleeding-edge TOXIC version pronto.

The Good

Decent clocks
Very quiet
Easy three-screen setup
Overclocks well

The Bad

Still not as good as GeForce GTX 680
More expensive than GTX 680

HEXUS Rating

Sapphire Radeon HD 7970 OC Dual-X

HEXUS Where2Buy

TBC.

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.

Login with Forum Account

Don't have an account? Register today!

Poor AMD. Yes it may overclock well but the price you pay is ridiculous. £40 more, a lot more power and heat usage, and still can't beat a stock clocked 680.

Only thing they can do now is reduce the price and hope people will rather save a couple of quid against gaining a few FPS

remember though the Nvidia card isnt what youd call stock clocked… its probably hitting 1100mhz iirc? Nvidias boosting method means that stability and product life could be severely reduced, the fact that you can get a 7970 to 1200mhz shows its possible, i think that the energy effciency is being mishaped by the game used for that benchmark, i think i mentioned it last time but you guys should really switch back to whatever game you used to stress it for power, or atleast switch to a more neutral game like BF3.

Id love to shove one of these cards in my water loop, im sure it would kick the 680s ass then :D. Ideally AMD need to improve yield as im thinking they are having to widen their testing to let other batches in, they need to manage a stock clocked chip to come out to retail at 1200mhz with about 5% better power efficiency, it can be done but i think it will take the guys at the labs a bit longer.

We need a similar 4890, or at the very least they need to turn the work put into the 7870 chip and expand it, that is the truely epic beast… overclocking to 1.2ghz gives between 7950/70 levels (mostly at 70 levels) and still only uses the power of less than a gtx460 448!. If they could expand it a touch more with a bit of bandwidth and some more ROPS or just in general a less severe cut back in units then it would be the best gaming chip ever.

Core clock now adays brings more performance than more cores or anything unfortunately, this is pretty much like what we were vs single/dual/triple/quad… it was hard to say more cores is better due to games performing better with a single chip running at 4ghz vs a dualcore running at 2 :P.

At this point my next card would be the 680, I dont run the resolutions that these things are benched at and I run with vsync turned on, nvidia's new features that adjust the clock of the card on the fly based on required output win it for me simple as..

Not had an Nvidia card in my machine for as long as I can remember, infact the last nvidia card that I wanted was the original Geforce DDR card back, erm 12/13 years ago (!?!?) think I've been ATi/AMD since then…

Actually although I originally thought that the Nvidia Turbo technology had a preset upper limit,it increasing does not seem that way from what I gather.

Some people are reporting 8 bins of 13MHZ as the maximum boost level but others are reporting higher boost levels.

From HardOCP:

“This means the GPU clock speed could increase from 1006MHz to 1.1GHz or 1.2GHz or potentially even higher. (Kyle saw a GTX 680 sample card reach over 1300MHz running live demos but it could not sustain this clock.) The actual limit of the GPU clock is unknown.”

1.)It means the maximum percentage is determined by things such as heat(not just a simple case of whether the boost hits the maximum preset value or not).

It will be interesting to see dependent on the conditions,how high and how high many times the boost is activated.

2.)The maximum boost number will also be determined by the quality of your GPU,ie,like if you were to overclock it manually. It would mean even under the same conditions it would vary from card to card and batch to batch.

3.)If the 28NM process improves newer cards might end up faster than old ones,unless binning is less stringent later on.

Unless you disable the boost you cannot get an accurate baseline speed for the card,as dependent on the sample it might boost a small amount or a larger amount. It seems you cannot disable the boost.

It will be interesting to see how this pans out TBH,as I suspect the GTX670 and GTX670TI will have similar technology.

The GTX670 against the HD7870 battle is going to be interesting. Before my current HD5850 1GB my previous fastest card was a 8800GTS 512MB. Nvidia has far better support for games under Linux,so it I will be excited to see how it goes.

Hicks12
…think that the energy effciency is being mishaped by the game used for that benchmark, i think i mentioned it last time but you guys should really switch back to whatever game you used to stress it for power, or atleast switch to a more neutral game like BF3.

GTX 680 power figures are lower in each of our games. We use Batman because it, for now, gives the most consistent results.