For our final bit of AMD news today, in a bit of an unusual move AMD has informally announced a new Radeon HD 7750.

The new 7750, which will not be replacing the old 7750, is a higher clocked, higher power consuming Cape Verde card, which takes advantage of the fact that AMD had to underclock the 7750 for power consumption purposes. As you may recall the original 7750 was designed to be the most powerful 75W card on the market – the limit for power from a PCIe slot – and to get there AMD had to clock it at just 800MHz. In reality Cape Verde is such a small GPU that it can easily hit higher clocks, but since 7750 was already approaching the limits of a PCIe slot, there wasn’t much headroom to work with.

AMD GPU Specification Comparison
  AMD Radeon HD 7770 AMD Radeon HD 7750 (New) AMD Radeon HD 7750 (Old)
Stream Processors 640 512 512
Texture Units 40 32 32
ROPs 16 16 16
Core Clock 1000MHz 900MHz 800MHz
Memory Clock 4.5GHz GDDR5 4.5GHz GDDR5 4.5GHz GDDR5
Memory Bus Width 128-bit 128-bit 128-bit
Frame Buffer 1GB 1GB 1GB
FP64 1/16 1/16 1/16
Transistor Count 1.5B 1.5B 1.5B
PowerTune Limit 100W 83W 75W
Manufacturing Process TSMC 28nm TSMC 28nm TSMC 28nm
Architecture GCN GCN GCN
Price Point $159 $109+ $109

As such AMD has minted a new 7750 reference design for their partners to work with. The new 7750 (which we’re dubbing the 7750-900) utilizes the exact same Cape Verde GPUs, but is clocked at 900MHz rather than 800MHz, which given AMD’s near-linear performance scaling should put performance at around 10-12% higher than the original 7750-800. On the hardware side of things the real change is that in order to achieve this AMD had to increase the amount of power the card could be supplied with, which means they had to give up on only drawing power from the PCIe slot and to start using external power too. The new 7750 reference board – which we’re told is very similar to the 7770 reference board – is a longer board that features a PCIe power socket in order to provide that additional power. Accordingly the PowerTune limit/TDP has risen from 75W to 83W.

And to be clear here, the new 7750 is as official as any other AMD defined product; this is not a factory overclocked part where partners are technically running a part out of spec. AMD will be qualifying all 7750 GPUs to run at both 800MHz and 900Mhz, and partners will have the option of using their 7750 GPUs to either build the 7750-800 or the 7750-900 based on their market needs. For that reason both models will continue to exist on the market.

Unfortunately this also means that they will be sharing the same name. AMD has told us that they don’t want to make their product stack more confusing by adding to it, which is understandable. At the same time however this is the textbook case for a new SKU; 7751, 7755, or 7760 would all be appropriate choices and far less confusing since the 7750-800 is a unique card due to its power requirements. As it stands the only way to tell apart the 7750-800 and 7770-900 is going to be based on clockspeed.


XFX Radeon HD 7750 BEDD. Image Courtesy Newegg

As for pricing and availability, the 7750-900 is available now, while pricing is still in flux. XFX’s Radeon HD 7750 Black Edition Double Dissipation is the first 7750-900 card on the market, and other partners will be putting out their own 7750-900 cards soon. We don’t have a clear shot of the front of the PCB, but Newegg has posted a shot of the far edge, where we can clearly see the PCIe power socket and part of the PCB, which also makes it clear that this isn’t a 7770 PCB. AMD’s official MSRP for the 7750-900 is the same as the 7750-800 at $109, while XFX is charging $115 after rebate for their card. Given the higher performance of the 7750-900 we don’t have a great deal of confidence that AMD’s partners will actually roll it out for $109, but given the fact that the 7770 is down to $129 they may not have a lot of choice.

Finally, the fact that AMD has chosen to launch a higher clocked 7750 at this point in time is by no means a coincidence. With the impending release of desktop GK107 cards – the OEM-only GK107 based GT 640 was announced back in April – this is an obvious counter-launch. AMD would appear to be expecting GT 640 pricing to be near the 7750, which would explain the need for a higher clocked card at the same price. We’ll be looking at 7750-900 performance next week when our reference sample arrives, so we should be better able to quantify performance and power then, along with what NVIDIA’s performance would need to be if indeed they’re targeting 7750 pricing.

POST A COMMENT

31 Comments

View All Comments

  • Rookierookie - Friday, June 01, 2012 - link

    I'm pretty sure that this represents a new low in GPU naming. Reply
  • Assimilator87 - Friday, June 01, 2012 - link

    At least it's an increase in performance. The first thing that pops into my head is when nVidia cut the 9600 GSO from 96 shaders to 48, although I'm sure there are many examples of crippled rebadges from both parties. Reply
  • Oxford Guy - Saturday, June 02, 2012 - link

    Tu quoque fallacy.

    Nvidia could be worse than AMD and that doesn't exonerate AMD's shady business practices at all.
    Reply
  • CeriseCogburn - Monday, June 11, 2012 - link

    Our brilliant fellow posters won't notice the : " Now at 900mhz ! " in big yellow letters on the box, and of course they will fail to have a 250 watt PS and extra molex to power the 6 pin, and won't notice it has a 6 pin either.
    They're the best gamerz in the world too, and know it all.
    Reply
  • Taft12 - Friday, June 01, 2012 - link

    I think it merely equals the low of the GTX 560 Ti - 448 cores edition ;( Reply
  • MonkeyPaw - Friday, June 01, 2012 - link

    Where were you when the 8800gt became the 9800gt?

    Maybe it should be called the 7750+.
    Reply
  • Oxford Guy - Saturday, June 02, 2012 - link

    "Where were you when the 8800gt became the 9800gt?"

    That's not the same thing as selling two different products with the same name, which is the problem here, unless you're talking about the die-shrunk version of the 9800gt in comparison with the 65nm version. Yes, Nvidia has been very bad in the naming department, too. I doubt anyone will dispute that. That doesn't make AMD less culpable. There's a fallacy called tu quoque.
    Reply
  • CeriseCogburn - Monday, June 11, 2012 - link

    Maybe they should just laugh at the nerds and geeks and self builders who they know, for a fact, constantly scrounge for every last penny they can get in discount off an amd card - and scrutinize frame rates down to the 1fps at 1920X1200, but just can't read a cards major specs or power requirements...

    I bet they laugh at the immense hilarity of the suddenly incapable and incompetent.

    Maybe we should force testing on all upgraders and require them to have a license and be certified so we don't have idiots complaining all the time.
    Reply
  • piroroadkill - Saturday, June 02, 2012 - link

    Haha, no. Free extra performance..
    NVIDIA still rules the roost by a long way when it comes to fucked up naming.
    How about a GT 630, a GT 630 or a GT 630? (Yes, there are 3 of those)
    ... and so on, and so on. You could go through and list all the rebrands too.
    Believe me, this is nowhere NEAR a new low in naming.
    Reply
  • MrSpadge - Saturday, June 02, 2012 - link

    Just f*ckin call it an HD7760 already. They've got 4 digits and still can't figure out how to make good use of them..

    I'm already waiting for the "HD7970 GHz Hyper Ultra Mega Super XXX Edition", which in another universe might just have been called HD7980. OK, I was quoting the nVidia naming scheme here.. but I guess you get the point ;)
    Reply

Log in

Don't have an account? Sign up now