First review of HD5770 up

[v8envy]

ATI may not think so, but there's still a need for a "mainstream enthusiast" part. This ain't it. Like the 1600XT and 8600GT before it, taking a high end part and chopping it exactly in half produces a "mainstream" part only at the MSRP level. Not from a performance viewpoint. Judging from history, games will want at least 2/3 of high end part for satisfactory mainstream gaming. And just like SM3.0 and DX10 were checkpoint features for the low-end-sold-as-performance-mainstream 1600XT and 8600GT, DX11 is unlikely to be valuable on what looks to be the current generation's lower end hardware.

Definitely lots of room here for improvement. The 4770 will make a nice low mainstream/upper budget card once the prices drop below $100. Leaving a huge pricing and performance hole for the future 4830. That's the next card to look forward to.

 

[MODEL3]

the 5770 performs (i checked 5 reviews) around -5-10% lower than what i expected.

If the price falls with M.I.R. to $149 (or better to 139$ with $20 M.I.R.), i think it is a better value than 4870 1GB (lower power consumption ($$$) DX11 and all the other advantages like new features, lower temperatures. etc...)

Plus i think that with future titles (H1 2010 games) the graphics engines will be more GPU engine limited than memory limited, so i guess for those titles the 5770 will be close to 4870 1GB perf. (1920X1200 4AA 16AF)

Also i saw some strange results, possibly the drivers need a little bit work.

I have a feeling that when we test again with H1 2010 games the performance difference with 4870 1GB will be somewhat lower than what we are seeing now.

 

[RussianSensation]

Originally posted by: Fox5

Err, I don't remember the 6600gt wiping the floor with a 9800XT. They were about equal, and traded off in games.

6600GT vs. 9800XT
Doom 3 1024x768
6600GT =79 (70% faster)
9800XT = 46.5

Far Cry 1600x1200 4AA/8AF
6600GT = 22.4
9800XT = 20.5
*tie

Halo 1280x1024 noAA/8AF
6600GT = 49.2 (36% faster)
9800XT = 36

Jedi Knight 1600x1200 noAA/AF
6600GT = 66.4 (24% faster)
9800XT = 53.7

Jedi Knight 1280x1024 4AA/8AF
6600GT = 37.6 (loses but close)
9800XT = 41

Source Graphics 1600x1200 4AA/8AF
6600GT = 42.8 (29% faster)
9800XT = 33.1

Unreal Tournament 2004 1600x1200 noAA/AF
6600GT = 42.8 (wins but close)
9800XT = 38

Ok you are right that I am wrong to say it "wiped" the floor with 9800XT. However with latter games, 6600 only pulled away even more in games like Chronicles of Riddick, Lock On Modern Air combat. But 5770 just doesn't have random situations where it's beating a 4890 by 70%! or even 30%. Maybe DX11 will give it a 20% boost in performance?

 

[v8envy]

Well, perhaps comparing against a currently shipping parts from the same manufacturer is the wrong thing to do. The 8800GT/8800GTX/8800GTS/9800GT/9800GTX+/250GT (and any other rebadged G80 models I missed) would be a better historical analog. When compared to the G80/G92 parts you can definitely see a dramatic improvement in current gen midrange vs. previous generation high end.

 

[Vertibird]

Originally posted by: v8envy
ATI may not think so, but there's still a need for a "mainstream enthusiast" part. This ain't it. Like the 1600XT and 8600GT before it, taking a high end part and chopping it exactly in half produces a "mainstream" part only at the MSRP level. Not from a performance viewpoint. Judging from history, games will want at least 2/3 of high end part for satisfactory mainstream gaming. And just like SM3.0 and DX10 were checkpoint features for the low-end-sold-as-performance-mainstream 1600XT and 8600GT, DX11 is unlikely to be valuable on what looks to be the current generation's lower end hardware.

Definitely lots of room here for improvement. The 4770 will make a nice low mainstream/upper budget card once the prices drop below $100. Leaving a huge pricing and performance hole for the future 4830. That's the next card to look forward to.

All ATI needed to do was give us a 256 bit/800 stream processor part for midrange. But then this would resulted in a die shrink of 4890 and DX10.1.

Engineering a midrange part is no doubt easier and faster when all that needs to be done is chop Cypress in half. Instant DX11 part.

P.S. Both HD5770 and HD5870 could probably use a boost in bandwidth.

 

[AzN]

These should be priced much lower. It's basically overclocked 4850.

 

[n7]

Wow, the 5770/5750 get a big fat meh from me.

Very disappointing performance.

Low power consumption, Eyefinity, & DX11 are great & all, but these cards really don't perform as they need to, or they sure as hell need to be a lot cheaper.

Definitely unimpressive :thumbsdown:

 

[nOOky]

It solidifies the thinking that many enthusiasts agree with. Whether it's nvidia or ATI, any extraneous selling points usually take a backseat to pure horsepower. We want fps, enough to run our favorite games at our chosen resolution with the appropriate eye candy turned on. Right now Eyefinity, DX11, PhysX etc. are mere fluff for the majority of gamers out there. Judging by the popularity of upcoming games (Modern Warfare 2 especially) none of these things are a factor yet. That's why methinks the 5770 is sort of a flop, I was expecting 4870 performance with a lower power draw.
Just my opinion.

 

[v8envy]

Originally posted by: Vertibird

All ATI needed to do was give us a 256 bit/800 stream processor part for midrange. But then this would resulted in a die shrink of 4890 and DX10.1.

No, what they needed to do is release these parts as 960 and 1120 stream processor models. Yes, they would have lost in a few cases because of lower memory bandwidth, but they'd also have won a few. It wouldn't be a clear cut loss to the 4870, never mind 4890 across the board. They'd have still sold out the 4-series to less informed buyers.

I expect there will be a 5790 refresh and that may be a card worth getting. The same memory cost for the 5750 as the 5870 guarantees it.

 

[Rezist]

After looking at the pricing, and all the data something is made obvious. There's many people that were waiting for this level of card trying to decide what to purchase. ATi has two agendas to fulfill sell out the 4800 series and make profit on the 5700 series. There price is artifically high right now to force the 4800 series out the door and it will likely work.

In 2 months I expect to see these cards all drop roughly 30-40$ each. Possibly up to 50$.

 

[cusideabelincoln]

Originally posted by: bunnyfubbles
Has anyone taken a closer look at the CrossFire results? According the AT review, crossfired 5770s are almost always just as fast if not faster than a 5870. Just about every game except Dawn of War II (and to a lesser extent, Crysis) dual 5770s actually seem like a better deal than a 5870, especially when the idle and load power numbers are also so very close. I find this strange considering the 5770 is virtually half of a 5870, and would have assumed scaling inefficiencies to keep dual 5770s consistently below that of a 5870. That being said, I think that might be indicative of room for improvement via drivers...that or a potential 5770 X2 would be a very enticing proposition (assuming such a product would be priced around $300)...

I think, and this is just a loose theory, that Crossfire can mitigate the fact each card is restricted by memory bandwidth or even memory amount.

 

[Vertibird]

Originally posted by: cusideabelincoln

Originally posted by: bunnyfubbles
Has anyone taken a closer look at the CrossFire results? According the AT review, crossfired 5770s are almost always just as fast if not faster than a 5870. Just about every game except Dawn of War II (and to a lesser extent, Crysis) dual 5770s actually seem like a better deal than a 5870, especially when the idle and load power numbers are also so very close. I find this strange considering the 5770 is virtually half of a 5870, and would have assumed scaling inefficiencies to keep dual 5770s consistently below that of a 5870. That being said, I think that might be indicative of room for improvement via drivers...that or a potential 5770 X2 would be a very enticing proposition (assuming such a product would be priced around $300)...

I think, and this is just a loose theory, that Crossfire can mitigate the fact each card is restricted by memory bandwidth or even memory amount.

Or maybe it just means HD5870 is restricted by bandwidth too.

(Someone at Hardforum did a comparison of Crossfired 4890s vs a single 5870 @1920x1200 resolution in Crysis. Crossfired 4890s gave ~10% better frames)

 

[Kraeoss]

hrmm this look like a great buy... i think i'll get a 5770 by black friday... anyone know when that is ?....

 

[Ika]

Originally posted by: Kraeoss
hrmm this look like a great buy... i think i'll get a 5770 by black friday... anyone know when that is ?....

Thanksgiving? really?

 

[cusideabelincoln]

Originally posted by: Vertibird

Originally posted by: cusideabelincoln

Originally posted by: bunnyfubbles
Has anyone taken a closer look at the CrossFire results? According the AT review, crossfired 5770s are almost always just as fast if not faster than a 5870. Just about every game except Dawn of War II (and to a lesser extent, Crysis) dual 5770s actually seem like a better deal than a 5870, especially when the idle and load power numbers are also so very close. I find this strange considering the 5770 is virtually half of a 5870, and would have assumed scaling inefficiencies to keep dual 5770s consistently below that of a 5870. That being said, I think that might be indicative of room for improvement via drivers...that or a potential 5770 X2 would be a very enticing proposition (assuming such a product would be priced around $300)...

I think, and this is just a loose theory, that Crossfire can mitigate the fact each card is restricted by memory bandwidth or even memory amount.

Or maybe it just means HD5870 is restricted by bandwidth too.

(Someone at Hardforum did a comparison of Crossfired 4890s vs a single 5870 @1920x1200 resolution in Crysis. Crossfired 4890s gave ~10% better frames)

If or if not the 5870 is restricted by bandwidth (I personally the 5870 is slightly restricted, but not majorly) doesn't mean Crossfire doesn't mitigate bandwidth issues.

The 5770 has less memory bandwidth than the 5870; in fact it has half, right? Putting two of them together in Xfire mitigates this problem, and thus its relative performance to the 5870 is increased. This is my theory, although there could easily be some other explanation for Xfire scaling.

The 5770 is basically a 5870 cut in half, with a 50 MHz decrease in core clock speed. It has half the bandwidth and half the shaders. It doesn't particularly make sense, on the superficial level, for two of them to outperform the a single 5870. So something else is going on.

 

[toyota]

Originally posted by: cusideabelincoln

Originally posted by: Vertibird

Originally posted by: cusideabelincoln

Originally posted by: bunnyfubbles
Has anyone taken a closer look at the CrossFire results? According the AT review, crossfired 5770s are almost always just as fast if not faster than a 5870. Just about every game except Dawn of War II (and to a lesser extent, Crysis) dual 5770s actually seem like a better deal than a 5870, especially when the idle and load power numbers are also so very close. I find this strange considering the 5770 is virtually half of a 5870, and would have assumed scaling inefficiencies to keep dual 5770s consistently below that of a 5870. That being said, I think that might be indicative of room for improvement via drivers...that or a potential 5770 X2 would be a very enticing proposition (assuming such a product would be priced around $300)...

I think, and this is just a loose theory, that Crossfire can mitigate the fact each card is restricted by memory bandwidth or even memory amount.

Or maybe it just means HD5870 is restricted by bandwidth too.

(Someone at Hardforum did a comparison of Crossfired 4890s vs a single 5870 @1920x1200 resolution in Crysis. Crossfired 4890s gave ~10% better frames)

If or if not the 5870 is restricted by bandwidth (I personally the 5870 is slightly restricted, but not majorly) doesn't mean Crossfire doesn't mitigate bandwidth issues.

The 5770 has less memory bandwidth than the 5870; in fact it has half, right? Putting two of them together in Xfire mitigates this problem, and thus its relative performance to the 5870 is increased. This is my theory, although there could easily be some other explanation for Xfire scaling.

The 5770 is basically a 5870 cut in half, with a 50 MHz decrease in core clock speed. It has half the bandwidth and half the shaders. It doesn't particularly make sense, on the superficial level, for two of them to outperform the a single 5870. So something else is going on.

the 5870 and 5770 have the same 850mhz clockspeed.

 

[woolfe9999]

I don't see why the 5770 needs to go below $100 to be a worthwhile part. Knock about $25 off the MSRP and it's about at parity price/performance wise with the 4870. The features then make it the better buy. IMO, it's a mistake to compare MSRP of a product just released with ARP of products that have been on the market for a long while. Ordinarily you'd see a card at $159 MSRP go below $140 within a week of launch, though in the current situation all the new AMD boards are staying at MSRP longer than usual because of limited availability. That will certainly change though, likely in a matter of weeks. I'd say judge the product for what the likely street price will be a little later on rather than putting too much stock in the snapshot of release day pricing.

- woolfe

 

[LW07]

I honestly feel "meh" about the 5770. It just seems like this generation's 8600GT/X1600XT.

Given what has happened in the past I believe that sooner or later this generation will get its own 8800GTS 320/X1900GT card which performs between the midrange and high end cards.

I just wish that all midrange cards would be like the 6600GT was and compete successfully with last generation's high end stuff.

Basically, when there's a good series of cards out, the performance usually ends up to be roughly equivalent to the last generation's midrange for the lower-end parts, equal to the last generation's high end parts for the new midrange parts, and roughly equal to an SLI/Crossfire configuration of last year's high end parts with the new generation's high end parts.


ATI has managed to do well with the 5870 in getting it to compete with the GTX 295, but its midrange 5770/5750 need to come down in price since they consistently lose to the 4870/GTX 260 when they need to be slightly faster. Maybe the drivers haven't matured yet. The 128-bit memory bus is definitely the bottleneck in this case.


On another note, even though I don't have one, its amazing to me how in 3 year's time that 8800GTX-level performance(erg 8800GTX/8800GT/4850) is still good enough to play todays games in good resolution and good details. Kinda reminds me of the Radeon 9700/9800 days when it took a long time for those things to finally no longer be good enough.

 

[bryanW1995]

Originally posted by: toyota

Originally posted by: Patrickz0rs

Originally posted by: netxzero64

Originally posted by: Patrickz0rs

Originally posted by: SSChevy2001
1 more reviews up
http://www.driverheaven.net/re...?reviewid=855&pageid=1

Wow the 5770 is on par with the 4870 even when the 5770 has DX 11. Looks like the 5770 will be no match for the 4890 even though it only has DX10.1

that's the bad news for me coz I thought that it would match the 4890 or just a notch lower than the 4890... sigh...

Everyone told me to wait and see how the 5770 and from the two reviews it seems nothing like the 4890. I will wait a few more days, otherwise I will just get this 4890 for $179.99 after rebate.

thats 20 bucks more when and if you get the rebate. the 5770 is pretty close to the 4890 at just 1440x900 and consumes way less power. it also has DX11 which will matter in a couple months. either go 5770 and just accept it as a good value or go 5850 and know that you are wasting a lot of it a just 1440x900. I know its a rough decision but I would probably go 5770 since 100 bucks more for the 5850 is a huge price jump. if you were at 1680 with your current rig I would say 5850 all the way.

dude, what's up with your 5770 love fest? 4890 is a clearly superior card, and since it's dx10.1 it will take advantage of some of the dx11 features that 5770 has. If we use every single DX rollout as a comparison, by the time dx11 becomes a big deal the 5770 will be slow as a turd. $260 is way too much to spend on his res, $179 isn't too shabby.

@op, the only thing that you might consider is that if you can wait the 5830 should be in your sweet spot. Unfortunately, they might not roll that one out for a while.

 

[toyota]

Originally posted by: bryanW1995

Originally posted by: toyota

Originally posted by: Patrickz0rs

Originally posted by: netxzero64

Originally posted by: Patrickz0rs

Originally posted by: SSChevy2001
1 more reviews up
http://www.driverheaven.net/re...?reviewid=855&pageid=1

Wow the 5770 is on par with the 4870 even when the 5770 has DX 11. Looks like the 5770 will be no match for the 4890 even though it only has DX10.1

that's the bad news for me coz I thought that it would match the 4890 or just a notch lower than the 4890... sigh...

Everyone told me to wait and see how the 5770 and from the two reviews it seems nothing like the 4890. I will wait a few more days, otherwise I will just get this 4890 for $179.99 after rebate.

thats 20 bucks more when and if you get the rebate. the 5770 is pretty close to the 4890 at just 1440x900 and consumes way less power. it also has DX11 which will matter in a couple months. either go 5770 and just accept it as a good value or go 5850 and know that you are wasting a lot of it a just 1440x900. I know its a rough decision but I would probably go 5770 since 100 bucks more for the 5850 is a huge price jump. if you were at 1680 with your current rig I would say 5850 all the way.

dude, what's up with your 5770 love fest? 4890 is a clearly superior card, and since it's dx10.1 it will take advantage of some of the dx11 features that 5770 has. If we use every single DX rollout as a comparison, by the time dx11 becomes a big deal the 5770 will be slow as a turd. $260 is way too much to spend on his res, $179 isn't too shabby.

@op, the only thing that you might consider is that if you can wait the 5830 should be in your sweet spot. Unfortunately, they might not roll that one out for a while.

love feast? what are you talking about? I was simply presenting both sides along with my opinion on what I would do. yes the 4890 is the better card but at just 1440x900 a 5770 is pretty decent too. nothing really wrong with going the 4890 route and then upgrading later I guess.


also we dont know what DX11 will bring. in the only one game that has it now the 5770 in DX11 is dead even with the 4890 in DX10.1. http://firingsquad.com/hardwar..._performance/page6.asp


As you saw in our BattleForge benchmarks though, the tables turn in DX11 apps. Here the 5770 is able to leverage DirectX 11 to gain a performance edge over the Radeon 4870, even though the 5770 is giving up significant amount of memory bandwidth to the 4870. Future DX11 titles like Aliens vs Predator and DiRT 2 will be using DX11 even more extensively than BattleForge, so it may be able to pull away even further from the 4870.

 

[bryanW1995]

why would techpowerup use cat 9.6 for the older ati cards? that probably had a lot to do with the tie between 4870 512 and 5770. their review also showed 4890 about 15% slower than gtx 285...

 

[LCD123]

I finish reading this thread and wow you guys have unrealistic expectations. Ive pointed out in another thread that 128 bit ram will create a bandwith bottleneck. It beats the hd4850 only because of higher clocks but can't equal the hd4870 because the hd5770 has 2/3 the memory bandwith. You are expecting too much from a $159 card. It's not highend nor will it run todays games maxed out at very high resolutions, otherwise no one would buy a $300+ card. Just buy the best card at your budget level and run games at whatever setting your card can handle.

 

[cusideabelincoln]

Originally posted by: bryanW1995
why would techpowerup use cat 9.6 for the older ati cards? that probably had a lot to do with the tie between 4870 512 and 5770. their review also showed 4890 about 15% slower than gtx 285...

I think obviously they did not re-run their tests for each review they do. They must use the same numbers through various different articles, and just add numbers as more reviews are done. Of course there does need to be a time to redo them all. Hopefully that comes soon.

Originally posted by: toyota

Originally posted by: cusideabelincoln

Originally posted by: Vertibird

Originally posted by: cusideabelincoln

Originally posted by: bunnyfubbles
Has anyone taken a closer look at the CrossFire results? According the AT review, crossfired 5770s are almost always just as fast if not faster than a 5870. Just about every game except Dawn of War II (and to a lesser extent, Crysis) dual 5770s actually seem like a better deal than a 5870, especially when the idle and load power numbers are also so very close. I find this strange considering the 5770 is virtually half of a 5870, and would have assumed scaling inefficiencies to keep dual 5770s consistently below that of a 5870. That being said, I think that might be indicative of room for improvement via drivers...that or a potential 5770 X2 would be a very enticing proposition (assuming such a product would be priced around $300)...

I think, and this is just a loose theory, that Crossfire can mitigate the fact each card is restricted by memory bandwidth or even memory amount.

Or maybe it just means HD5870 is restricted by bandwidth too.

(Someone at Hardforum did a comparison of Crossfired 4890s vs a single 5870 @1920x1200 resolution in Crysis. Crossfired 4890s gave ~10% better frames)

If or if not the 5870 is restricted by bandwidth (I personally the 5870 is slightly restricted, but not majorly) doesn't mean Crossfire doesn't mitigate bandwidth issues.

The 5770 has less memory bandwidth than the 5870; in fact it has half, right? Putting two of them together in Xfire mitigates this problem, and thus its relative performance to the 5870 is increased. This is my theory, although there could easily be some other explanation for Xfire scaling.

The 5770 is basically a 5870 cut in half, with a 50 MHz decrease in core clock speed. It has half the bandwidth and half the shaders. It doesn't particularly make sense, on the superficial level, for two of them to outperform the a single 5870. So something else is going on.

the 5870 and 5770 have the same 850mhz clockspeed.

I misread it; I was in a hurry. But whether it's the same or lower, my point stands. If the clockspeed on the 5770 was higher, then that might explain the Xfire scaling. But it's not, so it doesn't.