ATI 4770 info

[JAYDEEJOHN]

http://translate.google.com/translate?sourceid=navclient&hl=en&u=http%3a%2f%2fpublish.it168.com%2f2009%2f0331%2f20090331020601.shtml

 

[hannibal]

Well well, the 40nm low middle end card. Nice to see wattage numbers compared to 4830!

 

[salem80]

it will be closer to HD 4850 when overclock it

 

[hannibal]

The problem with earlier 40nm seems to be that they are not good overclockers. Too much leaking current. This is the biggest reason why there is not any higher end 40nm products coming at this moment.

 

[Kari]

hmm that's funny, they made a direct comparison with nvidia card... that's really unusual... makes me wonder if those slides are authentic or not...

atleast the cooler looks to be better suited for the chip than the one on the 4850, ie quieter and cooler

 

[JAYDEEJOHN]

These slides are just like the 4850 slides, which was also to compete with the 9800GT. Of course it was better, now, will this be better just as much?

 

[mactronix]

Cant see them releasing a card that dosent hand the 9800gt its arse in a sling. Problem is as i see it is this.
Ok so the 4830 is going and this will replace and slighlty better it. So what happens to the 4850 ? If this card performs a bit better than a 4830 then its bang on the heels of the 4850 and both cant survive in the same market.
So what ? This card in its two forms GDDR3 and GDDR5 will cover both cards and the 4870 will be fased out by the 4890 ? And do they plan that card in GDDR3 and GDDR5 ?
If thats a fair summary we can expect some price drops then ?

Mactronix

 

[cyber_jockey]

Who would have guessed the 98gt will get a fight from a 128bit graphics card.

 

[boudy]

I too, wonder what will happen to the 4850.
Maybe we had this all wrong. Maybe the 4770 is phasing out the 4850, and the 4750 is phasing out the 4830. It wouldnt make sense to leave a GDDR3 card right in the middle of 2 GDDR5 cards, would it?

 

[mactronix]

The way i heard it they already anounced the demise of the 4830, It was just a way of using up the defective 4850/4870 chips anyway. This is standard procedure in the GPU world. It just shows how good the yeilds are on the process. Time was you could get as many as 5 differant variations of the same chip.

Mactronix

 

[Dekasav]

The problem with earlier 40nm seems to be that they are not good overclockers. Too much leaking current. This is the biggest reason why there is not any higher end 40nm products coming at this moment.

Might you have some evidence to back this up, I heard there were/are problems with TSMC's 40nm process, but I really haven't seen much. And... I haven't seen any info on overclocking of R740.

 

[mactronix]

http://www.theinquirer.net/inquirer/news/878/1050878/tsmc-40nm-revealed

Mactronix

 

[Dekasav]

Hmmmm. Could be that they've had issues, but something from the inquirer doesn't really seem too much like proof, to me. I'll wait and see how much (if any) improvement (power usage and clocks) R740 has in an actual review.

Unless there's another source, of course.

 

[L1qu1d]

http://72.14.205.132/translate_c?hl=en&sl=de&tl=en&u=http://www.computerbase.de/news/hardware/grafikkarten/ati/2009/maerz/infos_bilder_ati_radeon_hd_4770/&usg=ALkJrhgJBepAWxB88GzX7DXFdTHgGC08cA

To me it looks like a 4830 with a die shrink and a massive overclock, should be really good. If the 4830 beat the 9800 GT in alot of tests (by very little). This card might be getting to 4850 standards.

I wonder what they are targeting.

 

[Dekasav]

What I really wonder is what the GDDR5 memory on it is rated as. If it's only rated for 3200Mhz (Effective), too bad, but if it's rated higher, like the HD 4870 (which runs its memory at 900 instead of its rated 1,000, I believe), it could put a good match against the HD 4850.

 

[4745454b]

Who cares? The power to game on a 22" LCD just hit $99. Thank you AMD. (it would be better if it didn't need the PCIe plug, but again, who cares?)

 

[boudy]

Who cares? The power to game on a 22" LCD just hit $99. Thank you AMD. (it would be better if it didn't need the PCIe plug, but again, who cares?)

The plug doesnt mean much though: http://www.fudzilla.com/index.php?option=com_content&task=view&id=12866&Itemid=34

I like the article that Liquid put up, it makes me feel that the 4770 needs only a slight overclock to reach 4850 speeds. Overclocking potential on this card is absolutely massive. Cant wait for it to come out.

 

[Dekasav]

With its core speed advantage, the HD 4770 is VERY close to the HD 4850.

Shader power is ~equivalent ( .96 TFLOPs vs. 1TFLOP)
ROP power is stronger for the HD 4770 by 20% (assuming perfect scaling with clockspeed)
Bandwidth is ~20% lower for the 4770, but I expect the GDDR5 memory to overclock another 100Mhz or so, granting a 12%+ increase in bandwidth.
Texturing power is also very close, (32 TMUs @ 750 = 24k; 40 TMUs @ 625 = 25k, difference is ~4%, same as shader power)

So the only thing the HD 4850 has a significant advantage in is bandwidth and the fact that it will probably overclock better (can't 4850's hit 700Mhz+?), but since we don't have a concrete 40nm card, it's difficult to determine how far the 4770 will OC.

Overall, it seems a very close race, as the initial Guru 3D benchmark showed, and I _think_ the HD 4770 will make up ground on the HD 4850 once overclocked (mainly due to extra bandwidth).

 

[JAYDEEJOHN]

Im thinking, being the 4850 is already BW bottlenecking somewhat, having the BW on these cards may be critical

 

[cyber_jockey]

If it does beat the 4850 doesn't that mean its a big threat to the 98gtx ? I sense huge droppings from nvidia. Now if they could only reach a high end card based on that nano meters.

 

[Dekasav]

I don't see any reason to believe they can't reach a high-end card based on 40nm. The 4770 is stronger than the HD 4830 in every regard but bandwidth (which can be solved by using faster GDDR5), yet has power consumption down 30% (80watts from 110watts).

Seems to me like ATI is just testing the waters, and seeing how the 40nm process works for them, especially since this card is very highly clocked (750Mhz).

 

[mactronix]

Hmmmm. Could be that they've had issues, but something from the inquirer doesn't really seem too much like proof, to me. I'll wait and see how much (if any) improvement (power usage and clocks) R740 has in an actual review.

Unless there's another source, of course.

Yes i know its the inquirer but its been reported other places as well and its been known about as far back as last december. Its not just been discovered it is a real issue.
Dont take my word for it, i assume you have a web browser

Mactronix

 

[mactronix]

I really want this card to be good as well but i really think that some of the assumptions/guestimations as far as its Overclocking potential are concerned may be wishfull thinking. personally if its slightly faster than a 4830 but cheaper(which is by far the most likely outcome at this stage) i will be happy. I think the revision card will be where we see all this potential you guys are talking about.
I really hope im wrong but all the noises are that the leakage is a real big issue for them.

Mactronix

 

[Dekasav]

Only thing I'm concerned about overclocking is the memory, which shouldn't care in the least about TSMC's 40nm process.

And I'm looking for something to beat an HD 4830, too, actually. I just think it might be closer to the HD 4850, personally. But we'll wait and see (which is much harder than speculating, doing simple mathematics, and hoping it does well).