• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

5770 & 5750 are out on newegg now.

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Die shrink, intrstruction sets, etc are different. Its not just a die shrink. Its not major like DX9 to 10 (7900GTX to 8800GTX for example), but its not just a die shrink.
 
This is an awesome card , did not redefine price performance ?. 4870 launched at a cool price tag of 279 $ , this is basically same performance for 160 $ + eyefinity dx11 and cooler operation .It cost more but only until 4870 old stock clears out , this card has a potential of being a 100 $ card , 4870 is 256 mm2 this is 180 mm2 :) Very much impressed by this card , like many others would like it cheaper tho .

Edit not because it is not worth the dough , just because am broke atm :(

/facepalm


5770 is not a high end part it is mainstream .4870 is a high end part and so is 5870 .Also ATI cards been overclocking rather solidly and most reviews site large improvements once the card is overclocked.

I have to respectfully disagree with you here. You're comparing launch prices from over a year ago to a card from the next generation so launch prices are meaningless. Current street prices or maybe even current MSRP, not that it matters nearly as much as street prices, are the way to compare.

People have come to expect ATi to deliver better price/performance than the current market prices, or at least the same price/performance. Now I'm willing to dismiss 2560x1920 performance because that's a high-end resolution and the 5770 is a mid-range card but lower resolutions don't change the performance comparison. Even with less tangible 'value add' features like DX11and Eyefinity some people like me are mildly disappointed because the cards only roughly fit in to the price/performance curve rather than having superior price/performance.

I do believe performance will be better as drivers mature though. xbitlabs does driver comparisons every once in a while so I'll be looking out for that review.
 
It doesn't scale linearly with memory speed increases so no the bus width doesn't hold it back. Dang that's a myth that just won't die :bang head

What I meant was, it has the same sp's as a HD4870, higher core and memory clock speed, and is basically a HD4870 with DX11. I would have thought with the same amount of cores + higher clock speed would mean a faster card, no?

The architecture is a little different, but I still think that based on the specs the 5770 and 4870 should be equal in performance or the 5770 should be better...except for the difference in the bus-width.

That's why I think that the bandwidth (or bus width) is important here. I think that if the 5770 had a 256-bit bus it would beat the 4870 due to it's almost equal specs and higher clock speeds.

If you compare the 4870-512 to the 4850-512 and set the core clocks equal the 4870 will win. Why? Has to be the bandwidth IMO.

GDDR3-256bit or GDDR5-128bit: A bit limiting at high rez.

GDDR3-448/512bit or GDDR5-256bit: Bandwidth to spare at high-rez.
 
Everyone loves to pontificate about it and be certain about their answer because of logical deduction. But ya know what, there's an easy way to test it using the scientific method it's not like this is far out physics theory, it's just a graphcis card. Take one of them, oc the memory, and see if it scales linearly with memory increase. If so it's bandwidth constained, if not it's not. Unfortunately all the folks around the net would rather discuss it than try it. But wait, here's a place that did! Only one game tested but...

http://www.techpowerup.com/reviews/HIS/HD_5770/33.html

11.3% benefit from a 9% core and 25% (!) memory oc. Bandwidth constrained?
 
I have to respectfully disagree with you here. You're comparing launch prices from over a year ago to a card from the next generation so launch prices are meaningless. Current street prices or maybe even current MSRP, not that it matters nearly as much as street prices, are the way to compare.

People have come to expect ATi to deliver better price/performance than the current market prices, or at least the same price/performance. Now I'm willing to dismiss 2560x1920 performance because that's a high-end resolution and the 5770 is a mid-range card but lower resolutions don't change the performance comparison. Even with less tangible 'value add' features like DX11and Eyefinity some people like me are mildly disappointed because the cards only roughly fit in to the price/performance curve rather than having superior price/performance.

I do believe performance will be better as drivers mature though. xbitlabs does driver comparisons every once in a while so I'll be looking out for that review.


People say that ATI brings price performance and they do ,but keep in mind last round ATI was launching against GTX280 and they had to compete, this season they have the whole dx11 and windows7 launch party to themselves ,all we need is NV back in the game and this will likely be a 100 dollar card like i sad , but right now this card is worth 160 bucks simply because it is the only mainstream dx11 card , it a selling point for OEM systems and a big one.

Right now nothing stops anyone from getting a 4870 instead of 5770 its a bit better card performance wise but then again you do want 5770 more :D so now you see .:soda: 5770 priced right where it suppose to be in AMD lineup just a bit over 4870 and cheapest 1gb 4870 is 150$ and cooler nowhere near as good as new 5xxx series reference cooler .


Right now AMD will enjoy high margins and the consumer didnt lost anything , we getting same thing for same amount of money , thats why we need NV but that wont happen for a while .I dont really disagree with you , just makings observations mostly , i do myself want cheaper cards too .
 
Everyone loves to pontificate about it and be certain about their answer because of logical deduction. But ya know what, there's an easy way to test it using the scientific method it's not like this is far out physics theory, it's just a graphcis card. Take one of them, oc the memory, and see if it scales linearly with memory increase. If so it's bandwidth constained, if not it's not. Unfortunately all the folks around the net would rather discuss it than try it. But wait, here's a place that did! Only one game tested but...

http://www.techpowerup.com/reviews/HIS/HD_5770/33.html

11.3% benefit from a 9% core and 25% (!) memory oc. Bandwidth constrained?

I wouldn't call it scientific if more than one variable is changed and only one game is tested. See below.

And I'm not certain about my answer...just stating my opinion based on my understanding of the technology. I don't feel that the link you provided disproves my theory.

I just can't see ATI making their SP's less efficient. The only other thing that's different is the bus-width. What is your theory for lower performance despite equal SP's and higher clocks?

You need to do it with one or the other, not both imo. Cos you don't know how much of that increase is from memory or core :S
 
Well yeah they increased both so it's not really quite right but 11% on a 25% memory increase, even with the core increae there too? What would be best would be: stock, mem oc, core oc, mem+core oc and see how they compare. Even then 'overall oc' comparisons are more valid that just assuming and theorizing :p
 
So, anybody have one yet? We need to run some tests! :)

I agree. I am really curious about HDMI sound output with these cards for a HTPC. I am very interested in a 7.1 sound solution that is all together. I am also curious about how loud they are and possibly quieter aftermarket cooling options again for a HTPC.
 
We can speculate all we want.

The performance is underwhelming because everyone is seemingly forgetting that these are mainstream cards and not high-range smashing machines.

People thought for years that a mainstream card had to beat a high-range card from the previous generation, but people are forgetting that that's not ATI's modus operandi anymore.

I feel the memory bus is the reason for the hampering. GDDR5 with a 128-bit bus? I think even GDDR3 cranks a 128-bit bus. The 8800GTX launched with 192-bit GDDR3 if I'm not mistaken and even the 2900XT had a 256-bit if I recall correctly.

The weird thing is, if the architecture is so superior, why does the 285 GTX manage to perform so well without using GDDR5? I never understood that.
 
I can't remember what the 8800 gtx had but the gts had a 320-bit bus. 256-bit buses have been in the game since the Parhelia and 9700 pro. It would be weird to go back.
 
GDDR or bus is not an issue , bandwidth is , you need sufficient bandwidth to effectively feed the core .Higher speed GDDR5 is implemented thus eliminating the need for bigger bus . 4870 had to much bandwidth as stated by AMD . And seeing how 5770 performs almost as good as 4870 with twice smaller bus seems they been right .
 
8800GTX had a 384-bit bus and 768MB of GDDR3 RAM.

Bandwidth is just the bus-width multiplied by the effective memory speed.



GDDR3@2GHz/512-bit = GDDR5@4GHz/256-bit
-------GTX285--------------------4890--------

GDDR3@2GHz/256-bit = GDDR5@4GHz/128-bit
--------4850----------------------5770----------



It's a whole lot cheaper to keep the bus-width down, and use faster memory to keep bandwidth similar or better.

I agree that the 4870 has excessive bandwidth, but the 4850 is limited w/ only half the bandwidth. Otherwise, you could OC the 4850 far enough and match the 4870's performance. I think the bandwidth is a little limiting on the 5770 as well for similar reasons.
 
Back