ATI's Radeon HD 4770 slides leaked

Being a 40nm part it must run cooler and be able to O-clock quite a bit.

Cna't wait to see the first benchies for this, looks nice if in fact these slides are factual.

This is shaping up to be an excellent card, very overclockable, fast at stock anyway, and for about £80? Spot on ATi :)

Looks like a direct competitor for the 4830. It makes sense for ATI to replace the the 4830 with a higher margin part, but why not mention of it INQ?

9800 GT
Old card, new sticker

GTdash - 9800 GT
Same card, new Sticker !?

LMFAO

I hate how people can not plot data correctly in order to make things look visually better!

The relative performance graphs start at 0.8? Either start it at 0 (so you can actually see the 'relative' difference), or start it normalized at 1 and simply plot the % better. The graphs make a 30-50% change look (visually) 2-3X better.

This is what happens when marketing does the graphs.

and most probably the pci-e power connector as well. f...k. isn't the double slot space heater line up is perfect as it is? does it really need another one? and with the die shrink it was the opportunity to lower the heat output.
hopefully the lower parts of 47xx will go that way...

waiting for ok preformance vid card that is cool and quiet (4670 is almost there).

They forgot the Innovation Vs Rebranding slide. OMG RFLOL

did DAAMIT just pull an nVidia with those performance slides? XD

I'm an AMD fan and all, but "Important Inflection Point in the Semi Conductor Industry"...?!
What a waste of slide space, guys!
Powerpointlessness at it's best.

Forget SaraLee, If Pat Gelsingers Photo is acccurate, it spawned Intel session on Type here: http://en.hardspell.com/doc/enshowcont.asp?id=5630 Tukwilia Part, Enough cores To: iA 'Bee Graphics Family Tree. Todays World Demands Billions of Trannies & Intel got 'Em. TS Drashek

I'm just waiting for fan boys to jump on this one... Both from the Reds and the Greens!

I'm always sitting on the fence.. My bum is hurting on this story (I hope you get that joke!) but I think I'm falling to the red side.

I'm just a bit fed up with the re-branding, paper releases and promo pushing of the green team.

To anyone that is seriously comparing CPU's to GPU's.. Just notice the major difference in their design in terms of computing data!

Wouldn't this be the whole renaming thing if this card was called...say a 4930...when I saw 4770 I immediately thought of a card that sat between the 4670 and 4830 ...but it more sits on the 4840 from what I'm seeing.

By using a 128bit memory bus and not the 256bit it can not be called a 48xx
But by having gddr5 in cant be called a 46xx either so it sit right in between.

achiles: the 4770 with 128-bit GDDR5 seems to have similar memory bandwidth as 256-bit GDDR3, making the 4830 redundant (and mine is only 6 weeks old :( ). BUT if cost is a design concern ($100 target), wouldn't GDDR5 be a be a bit pricey?

what a great IT website——PCPOP.com

Any info from them as to what the leakage is like on 40nm, and also what were the problems gating this getting released to the public?

I reckon the wattage could be down to this....

OK, this officially sucks. It has a power connector and is listed to use 80W. As such, it's likely not going to be well-suited for truly passive cooling.

Sucks. Really does.

GDDR3 256bit bus v GDDR5 128bit bus. one takes up more die space, the other has more expensive memory. there are OTHER advantages with using a smaller bus and faster memory (heat, power etc). overall it is probably slightly cheaper to have faster memory than a wider bus, at the cost of a little performance in this case 51Gb/s bandwidth to the 4830's 55.

overall expect a part inbetween the 4830 and 4850 (obviously) but closer to the 4850 than the former. sounds rather attractive for crossfire - 180 pounds, maybe less, for something that would surely beat down a GTX285? deal. over to you, nvidia

What the hell are you talking about??

All Nvidia results are nominalized to 1 in the graph and ATI are ranging from 1.1 to 1.5, which means 10 to 50% increase in performance. 0.9 and 0.8 in the graph is absolutely irrelevant to the results. it's obviously only cosmetic...

What part of this utterly simple procedure don't you understand, ''scientist''?

Ramon

Scientist has a very good point. The whole point of bar charts is that you can compare different things just by looking at the length of the bar, rather than having to compare the values. By starting at 0.8, it makes the 30% odd advantage that the 4770 has look like a 200% advantage, which is ridiculous. So once again you have to look at values to get a real idea of how the performance compares.

-1 for the red team, stop acting like the greens!

Running COH at a resolution of 1080 isn't demanding, hell my 9800MGTS can do that. Running the game with DX10 on a 1080 resolution puts any video card to a test hands down. I had a GTX280, I got around 43fps in DX10. I'm using a 4870X2 and it can't do DX10 in 1080 period, I can do it 1650 x 1050 but just not 1080. The game starts glitching up and crashes. Putting up a COH 1920 x 1080 slide without telling me if they ran it in DX10 is misleading.

How does this compare to the 4750 that was linked to a few weeks ago by the inq? That was labelled as the replacement for the 4830, but now there appears to be the same product with a different 3rd digit. What gives?

well it _seems_ like a good deal for the cash, bet it runs the optimized folding code just fine - very nice using (8800/9800)gt's to show baseline in this - if i were buying a card this is the range i look at so yey? btw that side by side slide is BADASS! WAY TO GO AMD/ATI

I might be thinking of a different article - but are you thinking of the RV740 review by Guru3D, in which they called the card the '4750'?

If that's the case, then that was just a speculated name for the card. There will apparently be a '4750' with clock speeds around 650MHz on the core (like the RV740 sample in the Guru3D review), but it will feature GDDR3 memory rather than GDDR5, crippling it.