Monday, February 13th 2012

Radeon HD 7770 Specifications Confirmed in GPU-Z Screenshot

Much like a previous exposé with Radeon HD 7950, the specifications of Radeon HD 7770 that were rumored in our previous article have been confirmed by users. It confirms several specifications, starting from the stream processor count of 640, to the 128-bit wide GDDR5 memory interface. The GPU has an out of the box core clock speed of 1.00 GHz, it could be possible that this is a factory-overclocked card, if not, the core clock speed rumor sparked off by the Verdetrol marketing campaign are true, after all.

Source: DonanimHaber
Add your own comment

45 Comments on Radeon HD 7770 Specifications Confirmed in GPU-Z Screenshot

#1
Crap Daddy
While we don't know the performance yet, the specs don't look too promising. I would have expected this to replace performance wise the 6870 at a lower launch price point but everything seems to be getting some weird prices under Rory's lead. I'd say wait for the benchmarks but be prepared to grab a 6870 just in case at 150$ while they last. On a second thought, make that two.
Posted on Reply
#2
sergionography
NdMk2o1o said:
It needs to be beating it by a good margin if it is to remain at the same price of a 6770' as they pulled the same bullshit with 6770>5770 and kept the same performance at the same cost over 2 generations, so all they have done is given us 5770 performance, no more now less over 3 gens whilst keeping the cost the same over the generations.
the 5770 was priced well over 200 as it was selling and was considered med-high at the time as the 5800 series were the high end, the 5770 was technicaly replaced by the 6800 series as amd moved up a number digit
the 6770 was a rebranded 5770 but was well cheaper


meirb111 said:
cheaper die size yet price will be a higher power counsumption isnt a big plus for most people for home pc's they look at price /preformance its not a "win win" its win some lose some.
price may be higher than current 6770's(since they got cheaper now) but eventualy they will become similar or even cheaper as amd has room to be competitive in price but that wont really happen untill nvidia releases their next gen
as for die size it will allow for higher clocks which is what allowed the 1ghz factory clock when 6770 was 800-850(thats 15-20%)
Posted on Reply
#3
sergionography
another thing that has to be noted aswell is that according to this article the 640 cores are NOT vliw4
they are GCN cores(same architecture as 7900 series)
Posted on Reply
#4
brandonwh64
Addicted to Bacon and StarCrunches!!!
_JP_ said:
Your card is overclocked. That is not a real world scenario, and not a good comparison because the 7770 is at stock.
Not my card :)
Posted on Reply
#5
jpierce55
sergionography said:
the 5770 was priced well over 200 as it was selling and was considered med-high at the time as the 5800 series were the high end, the 5770 was technicaly replaced by the 6800 series as amd moved up a number digit
the 6770 was a rebranded 5770 but was well cheaper




price may be higher than current 6770's(since they got cheaper now) but eventualy they will become similar or even cheaper as amd has room to be competitive in price but that wont really happen untill nvidia releases their next gen
as for die size it will allow for higher clocks which is what allowed the 1ghz factory clock when 6770 was 800-850(thats 15-20%)
The 5770 was never over $200. I bought it around release for $130. The 6770 was never $200 either. The nice thing about 5770/6770 is they x-fire together, I doubt the 7770 will.
Posted on Reply
#6
cadaveca
My name is Dave
I paid $109 for my 6770. If this card is similarily priced, cool, but i expect maybe a bit more higher price, because 7970 is higher-priced than previous gens. Will have to wait for reviews, I guess.

Would be nice if these cards supported more than two in Crossfire too...might be interesting to see 4x7770 vs single 7970.
Posted on Reply
#7
Cheeseball
btarunr said:
Cape Verde uses Graphics CoreNext, not VLIW4.
Oh okay, so it's using RISC functionality. If that's the case, then this should be able to trounce the HD 5770/6770 by... 10% more performance perhaps?

If it had 800 shaders, then it would be between a HD 6850 and HD 6870.

Just a guess as VLIW4 is single instruction-based only.
Posted on Reply
#8
_JP_
brandonwh64 said:
Not my card :)
Didn't know. "Your example" then. :)
Posted on Reply
#9
R_1
sergionography said:
the 5770 was priced well over 200 as it was selling
Since when the initial $160 price is " well over 200". Maybe that 200 was in currency of People's Bank of China. :roll:
Posted on Reply
#10
brandonwh64
Addicted to Bacon and StarCrunches!!!
_JP_ said:
Didn't know. "Your example" then. :)
HAHA LOL, yea I google imaged that pic. My 5770 is installed in my main ATM while I wait on my new 560TI 2GB Card
Posted on Reply
#11
Casecutter
It's pretty well established, Cape Verde with Graphics CoreNext will replace the X7XX line, and with close to a 6870 performance, while lower power. Pricing will yes be slightly wacky, but it's not Rory... It's no rivals from Nvidia presently or coming soon (GK106), along with the new reality of TSMC 28Nn pricing, that basically negated the normal shrink cost improvement.

MSRP of 5770 back in November 2009 was set at $160, now some E-tailer have "jumped the gun" and showed a $185 price for the 7770. If you read the historical comments, folks like yourselves wrote after the 5770 review (almost 2½ years ago) you can probably reuse them in the coming days/week for this. Except this time we know AMD won't leave the big gap, that's what the 78XX is there to cover and counter the GK104. AMD will be able to drop price of the Cape Verde and adjust if a GK106 makes the grade, something as of late Nvidia had chosen not to significantly challenged with the like of GTS450 or GTX550ti.

But will this (?) give Nvidia a perfect picture of what Graphics CoreNext will provide, and will know how to direct AIB’s the clock those GK104 cards. The problem may well be AMD is... sand-bagging on both the high/low end? Knowing Nvidia will stick their neck out to take the high-end crown, while holding off on what Cape Verde they don't expose the middle ground. I say they realize/betting they know what Nvidia can bring at this level with GK106 offering, not much different than the GTS450 or GTX550ti did previously against their rivals. They know the fight is really going to be with the mainstream-volume GK104/Pitcairn, and they all are playing those cards close to their chest. :rolleyes:
Posted on Reply
#12
Dj-ElectriC
Everybody thinks that the HD7770 is worse then the HD5770
Its not, chillax already :shadedshu
About the price... oh boy *grabs popcorn and waits for people's reaction soon*
Posted on Reply
#13
ViperXTR
if this is using GCN instead of the VLIW5 of the 5770/6770 it could prolly match the performance but the pixel and texel fillrate along with memory bandwidth might hamper it, then again this is a card with (supposed to be) different architechture, would have to wait for test results.
Posted on Reply
#14
sergionography
R_1 said:
Since when the initial $160 price is " well over 200". Maybe that 200 was in currency of People's Bank of China. :roll:
thats wierd, I could swear back in 2009-2010 when shopping for graphics cards every other 5770 was over 180 dollars on newegg and i ended up buying a gts 250 which was 120 dollars at the time while the 5750 was around 140 dollars
but that mightve been just newegg raising the prices sinse nvidia was still using the older 200 series
Posted on Reply
#15
micropage7
look at the fabrication, it goes from 40nm to 28nm that means the chip would be colder than previous chip
Posted on Reply
#19
vagxtr
natr0n said:
It looks to be worse than a 5770 or I am wrong?
You're wrong. Dont look with such narrow minded on raw specs

jpierce55 said:
I show 80gb bandwidth on my 5770 .... yet I do wonder on the downgrade. I DOUBT the higher clock and improved architecture will permit much increase over the old cards considering the shader loss. [...] This cards sounds like a fail :mad:.
It's not all in raw memory bandwidth. You compute with a CPU/GPu not the memory buffers ;) Dont you?

You cant compare kiwis vs. mangos color and shape if you're doing quality comparison. These are different architectures, and theres huge improvement betwen HD4870 and HD5700 series when the latter achieved almost same results on half-width memory bus. Do you remember?

I agree it's a fail when you look at it's HIGH PRICE but there should be huge room for price drops on this cards. Even on 75USD for HD7750 and 90USD for HD7700 long before EOL DAMN would make a considerable profit. (And they're MSRPed 110USD and 160USD respectively)

It's nice to know that two-an-half month later HD 7770 are priced at more reasonable 140USD but still at least 30USD too much. I hope that in next two-and-half month they'll cut them to normal price of 110USD :D



sergionography said:
this will have 640 GCN cores while 6770 and 5770 used the old vliw5 which had 10cu's aswell (except vliw5 has 80 stream processors per cu totaling 800 while vliw4 and gcn has 64 yet perform better than the vliw5's 80)
Different computing approach has nothing to do with raw memory bandwidth during gameplay, but improved texture compression does :toast:

GCN should proof itself better only in easier code optimizations than pretty WaferIncognita VLIW5 even for ATi which developed it and most of their engineers are now exported with benefits from new DAMNs conglomerate. So ATis VLIW5 and code optimizations doesnt play well with each other in the same universe. I still believe VLIW5 is far better approach just it nede far more time for properly implement it. They had crapstart with its firs R600 implementation and ever since they were doing bugfixes ... sucessfully i might add in HD3800/HD4800/HD5800 series

And then they skip back to HD6900series and VLIW4, so it might be another reason why they return back to X900 branding for their GPUs based on high-end chips .... preparing for new NON-BUG-BOTHERED GPU implementation ... GCN (fancy name for 15yrs old SIMD)
Posted on Reply
#20
vagxtr
Cheeseball said:
Yup, remember that the HD 5850 has 1440 shaders, but a HD 6870 that has 1120 shaders performs just about the same due to newer VLIW4 architecture.
HD6870 uses same old VLIW5 just they squeeze up maximal throughput to available shader clusters by rejecting unnecessary cluster would remain idle because of limited dual-arbiter implemented in HD5800/HD6800 series but now reducing waste of useful silicon. HD6900 used quad-arbiter for every 384SPs arranged in VLIW4 clusters featuring 64SPs per VLIW4 cluster. So one arbiter for every 6 VLIW4 cluster in HD6900 series or one for every 10 VLIW5 cluster ion HD5800 make significant throughput advantage for the first one.
Posted on Reply
Add your own comment