View Full Version : AMD Radeon HD 5750/5770
phoronix
10-13-2009, 12:10 AM
Phoronix: AMD Radeon HD 5750/5770
In late September AMD had introduced the Radeon HD 5850 and Radeon HD 5870 graphics cards as the successors to the Radeon HD 4850 and Radeon HD 4870, respectively. These graphics cards, which are part of the Evergreen GPU family, have been performing quite nicely according to reports, but we have yet to test either of these Cypress graphics cards under Linux. Today though AMD is introducing the first midrange graphics cards in the Evergreen family. Under the Juniper codename, the Radeon HD 5750 and HD 5770 are being launched with both graphics cards being quite similar except for the ATI Radeon HD 5770 shipping with slightly higher core and memory clocks along with a different heatsink. In this review we have the first Linux-based benchmarks of these two new graphics cards, which are also the first publicized Linux benchmarks from any AMD Evergreen graphics processor.
http://www.phoronix.com/vr.php?view=14264
The fan on this heatsink is moderately sized, but during our testing we found it to operate quietly.
The fan on the 5750 looks really nice. Remember, bigger fans can spin at a lower RPM, but move still move more air than smaller fans (as those of us who owned nFurnace4 boards before the advent of heatpipes can attest to).
Darkfire Fox
10-13-2009, 12:39 AM
Wow. The 5800 series was released, what -- three weeks ago? And now the 5700 series is being launched? Impressive. Does this mean we'll have 5600 series cards launching by early November? ;)
Of course, I'm still waiting for RadeonHD 2000/3000/4000 series GPUs to have proper OpenGL support via Mesa. Arch Linux should have the 2.6.32 kernel -- and OpenGL -- by late December. We hope. :D
monraaf
10-13-2009, 12:49 AM
The claimed idle power consumption looks really good, I wonder if this also holds true in Linux... Also it would be nice if you benchmark some games running under wine (e.g. Half-Life 2) in a normal resolution, I don't think many people have a 2560x1600 display...
iseggev
10-13-2009, 01:41 AM
Will these guys be able to use the r600/r700 mesa drivers, or will drivers need to be written from scratch to support them?
agd5f
10-13-2009, 02:22 AM
Will these guys be able to use the r600/r700 mesa drivers, or will drivers need to be written from scratch to support them?
We'll be adding support to the r600 mesa driver.
Drago
10-13-2009, 02:51 AM
We'll be adding support to the r600 mesa driver.
This is just COOL!
haplo602
10-13-2009, 03:24 AM
hmm ... can you include opengl 3 version string also ? I always miss that from any ATI/Nvidia driver or card review and have to go hunting for it.
anyway these look nice ... maybe as prices go down a 5770 will fit into my build instead of the current GTS 250.
EarthMind
10-13-2009, 03:40 AM
Too bad I bought a Radeon HD 4890 card a bit longer than a month ago :(
JeanPaul145
10-13-2009, 04:35 AM
Does anybody have any idea whether or not the FOSS radeon drivers support the 3-head display feature? I am aware that dualscreen is working nicely nowadays, but I don't want to extrapolate from that. And if the drivers don't support that feature, in what timeframe might it be able to do so?
GreekGeek
10-13-2009, 05:21 AM
Hi Yall & Michael,
um, like dude, di you just say ....
"The Unified Video Decoder 2 (UVD2) support with ATI Avivo HD may also be ignored by Linux users at this time, as we have been talking about the X-Video Bitstream Acceleration (XvBA) for a year now and it still has yet to be usable by ATI Radeon customers, but it finally will be in the very near future. UVD2 in ATI Radeon hardware will finally mean something on Linux, but we will have to wait to share more until permitted."
Holly Smokes, that should so have been the title of this article!
That is FANTASTIC NEWS!!!!
We can organize, UVD2 / X-Video Bitstream Acceleration (XvBA) Parties, accross the www, to celebrate! :-)
Just as well I went for a 5850! :-)
*GeekNirvana*
*OMG* *OMG* *OMG*
Right, that is enough for now....
Waits for further updates, in a mild state of shock.
GreekGeek :-)
rohcQaH
10-13-2009, 05:34 AM
Does anybody have any idea whether or not the FOSS radeon drivers support the 3-head display feature? I am aware that dualscreen is working nicely nowadays, but I don't want to extrapolate from that. And if the drivers don't support that feature, in what timeframe might it be able to do so?
as I understand it, the OS drivers currently support zero displays on 5xxx ;)
the third display isn't more difficult to program than the first two, so I'd expect all three outputs to start working at the same time.
Edit: hmm, you can use DVI+DVI+DisplayPort or DVI+HDMI+DisplayPort. Looks like DisplayPort isn't possible yet (http://wiki.x.org/wiki/RadeonFeature). Then again, this page (http://wiki.x.org/wiki/DisplayPort) claims that DVI over DisplayPort works anyway. So after all, I don't know more than you do. ;)
As Nvidia GTX 260 is about as expensive as the HD 5750 I would add it as comparision. Maybe GTX 275 too, even if it currently more expensive. I really dislike that no test mentions the extremely bad xv performance with fglrx. The tearing is so extreme what it is basically useless.
Also 8.67 must be 9-11 not 9-10 as 8.65 is the current 9-9 driver.
Pedric
10-13-2009, 06:25 AM
anyway these look nice ... maybe as prices go down a 5770 will fit into my build instead of the current GTS 250.
I am also looking for a comparison of the Nvidia GTS 250 and the HD5750, as the 1 GB GTS250s are priced similarly to the HD5750s...
GTX 260 has got the same price too, would be much smarter to buy GT200 instead of G92.
rohcQaH
10-13-2009, 07:46 AM
windows benchmarks suggest that HD5770 = HD4870 = GTX 260 (with only a few percent difference in either direction, depending on the game)
Don't just calculate with the retail price though, power consumption is a factor as well. Here's some values:
Scenario #1: a life-time of 3 years with an average of 2 hours of load + 5 hours idle per day. (my scenario, since I often work at home and like gaming on weekends)
Scenario #2: a life-time of 3 years, 24/7 on, an average of 2 hours of load (for those warez kiddies ;))
Calculating with a price of 0.20¢ for 1kW/h (I pay slightly more).
Costs for Power for different cards
Scenario #1 Scenario #2
card idle load kWh price kWh price
HD 5770 17W 108W 330 66€ 646 129€
HD 4870 55W 160W 652 130€ 1675 335€
GTX 260 38W 182W 607 121€ 1314 263€
So if you plan to actually use the card, the retail price is often secondary. And you'll quickly see that the 5770 will turn out to be the cheapest one - unless your power is paid by mom :p
V!NCENT
10-13-2009, 08:11 AM
Finally. ATI is back on track with a solid card. This won't be nessecairilly be a killer in Linux, but I'm glad that ATI spends it's time making better products than that nVidia trash nowadays.
May ATI take the lead again. I don't really care that much that my ATI card doesn't work that well on Linux, but at least they put developpers on FLOSS drivers and release spec. AMD is a good company. They try to earn money buy making a better product.
chrisr
10-13-2009, 09:41 AM
How much better is the HD 4670's 2D performance compared with the HD 4650's? Because I've got a 4650, and its current 2D performance is bad. So if these 5750/5770 cards are even worse, then they aren't going to be much use on the desktop.
XAA acceleration only, still?
L33F3R
10-13-2009, 11:13 AM
heh, just in time for L33F3R to buy a new card. :p
So it's public now: XvBA is coming soon ;)
cynyr
10-13-2009, 01:15 PM
Now I'm sure i could take some scale measurements from the photos, that's a bit more work than I would like to go through. I'm wondering what the dimensions of these cards are? mainly distance from the PCI cover plate to the end of the card. Not all of us are using full tower cases. Also I would be interested to see how these number compare to say a GTS250 or the new 220/210 cards from Nvidia.
Now I'm sure i could take some scale measurements from the photos, that's a bit more work than I would like to go through. I'm wondering what the dimensions of these cards are? mainly distance from the PCI cover plate to the end of the card. Not all of us are using full tower cases. Also I would be interested to see how these number compare to say a GTS250 or the new 220/210 cards from Nvidia.
The 220/210 cards aren't anything special judging from what I've read. semiaccurate.com has some notes on them, and anandtech has some benchmarks here (http://www.anandtech.com/video/showdoc.aspx?i=3657&p=15).
But you're right, some size measurements would be nice - I did find them somewhere, but can't remember where now (sorry).
Michael
10-13-2009, 01:44 PM
GeForce 220 benchmarks coming on Linux next week.
cynyr
10-13-2009, 02:37 PM
But you're right, some size measurements would be nice - I did find them somewhere, but can't remember where now (sorry).
ohh something else that would be nice to have measured is the amp draw of these cards under idle and at load. That would make ensuring enough power supply capacity much easier (under the assumption that you can get that info from your PSU manufacturer.)
i found dimensions. http://www.anandtech.com/video/showdoc.aspx?i=3658&p=2
cynyr
10-13-2009, 02:42 PM
But you're right, some size measurements would be nice - I did find them somewhere, but can't remember where now (sorry).
Would be nice to loads discussed in terms of X amps on the 12V rail, and Y amps on the 5V. or watts if you prefer, but separated for each voltage.
The Anandtech article has dimensions. The 5770 is just shy of 9" with the shroud which is expected to be dropped by manufactures due to the low TDP of the card. The 5750 is 7.15" no shroud to extend that.
Apopas
10-13-2009, 06:58 PM
I really dislike that no test mentions the extremely bad xv performance with fglrx. The tearing is so extreme what it is basically useless.
Isn't opengl performance good enough?
DeepDayze
10-13-2009, 07:52 PM
heh, just in time for L33F3R to buy a new card. :p
So now you deciding on a Radeon rather than a GeForce?
L33F3R
10-13-2009, 08:25 PM
So now you deciding on a Radeon rather than a GeForce?
S3 so far actually.
DeepDayze
10-13-2009, 09:08 PM
S3 so far actually.
I am not really keen on S3 cards...they are kinda low end and the Openchrome driver is rather lacking
BillBroadley
10-13-2009, 10:44 PM
Page 2, quiet, not quite.
BillBroadley
10-13-2009, 10:58 PM
Any users of any of the 57xx or 58xx cards under linux out there?
Obviously these cards are cheap, plenty fast, a good deal... especially if you care about noise/power/heat.
But does HD playback work with compiz? Google earth? Games?
I'm just hoping for something that works better than my old cheap 8600 gt.
jstokes
10-14-2009, 12:43 AM
I'm real happy with the AMD/ATI hardware under linux.
I plan on getting a 5XXX card in the near future, Probably after the first of the year. It's not just the hardware, I find CC works well under Linux. and the Driver release cycle has gotten better.. definitely better then nVidia. As of late I find the nVidia has been slipping, there cards are hot, over priced and their Linux drivers release schedule is poor.
Windows will have Eyefinity first, but they should be released in the November update. I am really looking forward to that. main reason to upgrade.
BillBroadley
10-14-2009, 12:49 AM
I'm real happy with the AMD/ATI hardware under linux.
I plan on getting a 5XXX card in the near future, Probably after the first of the year. It's not just the hardware, I find CC works well under Linux. and the Driver release cycle has gotten better.. definitely better then nVidia. As of late I find the nVidia has been slipping, there cards are hot, over priced and their Linux drivers release schedule is poor.
Windows will have Eyefinity first, but they should be released in the November update. I am really looking forward to that. main reason to upgrade.
So does HD playback, games, google earth and the like work with your setup and compiz?
Hasenpfote
10-14-2009, 05:08 AM
Besides offering competitive performance -- and superior in some cases -- to its predecessors, the Radeon HD 5700 series also has the advantages of moving to the 40nm fabrication process, dual-link DVI with DisplayPort and HDMI with Eyefinity support, reduced power consumption, each graphics card is only dependent upon a single 6-pin PCI-E power connector, TeraSCALE 2 Unified Processing Architecture, PCI Express 2.1, and support for DirectX 11 (if it matters to you).
Nice summary! Let me comment it:
1. What is the advantage for me of 40nm? I dont care for 40nm! They can produce in 130nm, if card is cheap, fast, silent and has a low power consumption. So overall: no advantage
2. Dual Link DVI: Does it work under Linux without problems with Display Port and HDMI? Did you make any tests?
3. Eyefinity: As you already mentioned: Not useable under Linux -> no advantage
4. Reduced Power consumption: Does the card reduce clocks and speeds when being idle work under Linux? How is the consumption under Linux? Where are the TESTS(!!!)?
5. Only 6-pin power connector! WOOT!!!! FANTASTATIC! I always have problems finding a power connector :confused:. Rest see point 1.
6. TeraSCALE 2 Unified Processing Architecture: yeah right... hummm :confused: Can it be used under Linux?
7. PCI express 2.1: What an adavantage... Unfortunatly the cards are so slow under Linux due to bad drivers, that they cant take adavantage of a bigger bandwidth. Correct me with TESTS(!!!) if I am wrong.
8. DirectX11: Yeah! What an big advantage under Linux!
So the summary is in reality:
Nice card, nothing new besides the name and some internal stuff no real world linux user would see. If you are a Windows user, check reviews of other sides!
Congratulations for such a nice and really usefull review!
tball
10-14-2009, 05:12 AM
So does HD playback, games, google earth and the like work with your setup and compiz?
Well, if you patch xorg-server, you will get wonderfull performance.
I got hd3650 on a laptop.
- I can play hd movies with xv, without problems.
- Wine works fine (there is alot of trouble with wine, when using fglrx < 9.7).
- kwin (kde4) works very good.
The only thing I can point my fingers on, is the 2d performance compared to the oss drivers. I use the oss driver right now with experimental 3d support. It works very stable, and compiz with 2d is like, YEAH!
BlackStar
10-14-2009, 05:28 AM
Nice summary! Let me comment it:
1. What is the advantage for me of 40nm? I dont care for 40nm! They can produce in 130nm, if card is cheap, fast, silent and has a low power consumption. So overall: no advantage
2. Dual Link DVI: Does it work under Linux without problems with Display Port and HDMI? Did you make any tests?
3. Eyefinity: As you already mentioned: Not useable under Linux -> no advantage
4. Reduced Power consumption: Does the card reduce clocks and speeds when being idle work under Linux? How is the consumption under Linux? Where are the TESTS(!!!)?
5. Only 6-pin power connector! WOOT!!!! FANTASTATIC! I always have problems finding a power connector :confused:. Rest see point 1.
6. TeraSCALE 2 Unified Processing Architecture: yeah right... hummm :confused: Can it be used under Linux?
7. PCI express 2.1: What an adavantage... Unfortunatly the cards are so slow under Linux due to bad drivers, that they cant take adavantage of a bigger bandwidth. Correct me with TESTS(!!!) if I am wrong.
8. DirectX11: Yeah! What an big advantage under Linux!
So the summary is in reality:
Nice card, nothing new besides the name and some internal stuff no real world linux user would see. If you are a Windows user, check reviews of other sides!
Congratulations for such a nice and really usefull review!
1. This implies point 4.
2. Never had any problems with DVI or HDMI on Linux (be it fglrx, nvidia or the OSS drivers). I doubt 5xx0 will change this.
3. FUD. AMD has already showcased X-Plane running on Linux with Eyefinity.
4. Obviously yes. My X1950 and 4850 show *identical* idle power consumption on Linux and Windows.
5. Yes, the single 6pin connector is significant. Other cards in the same performance bracket need more than that.
6. This is marketing speak.
7. OpenGL performance is identical on Windows and Linux (they use the same driver, duh). 2D performance is worse on Linux, but quite usable now (unlike, say, 6 months ago).
8. Shader Model 5.0 is available through OpenGL, so yes this is a big advantage on Linux.
You have no idea what you are talking about and it shows.
Nice summary! Let me comment it:
1. What is the advantage for me of 40nm? I dont care for 40nm! They can produce in 130nm, if card is cheap, fast, silent and has a low power consumption. So overall: no advantage
The lower nm production comes generally hand-in-hand with lower power, or higher clock speeds (better performance). They may also produce more chips off the same wafer, allowing lower costs. So overall, advantage.
2. Dual Link DVI: Does it work under Linux without problems with Display Port and HDMI? Did you make any tests?
3. Eyefinity: As you already mentioned: Not useable under Linux -> no advantage
Dunno about the dual link dvi stuff, but eyefinity doesn't work under linux yet - at least not for the consumer. AMD have it running internally I believe (someone correct me if I'm wrong there). Either way, it will likely run on linux eventually, so that's an advantage.
4. Reduced Power consumption: Does the card reduce clocks and speeds when being idle work under Linux? How is the consumption under Linux? Where are the TESTS(!!!)?
5. Only 6-pin power connector! WOOT!!!! FANTASTATIC! I always have problems finding a power connector :confused:. Rest see point 1.
6. TeraSCALE 2 Unified Processing Architecture: yeah right... hummm :confused: Can it be used under Linux?
Unified processing architecture will make it better for shaders and OpenCL. So yes, it can be used under Linux.
7. PCI express 2.1: What an adavantage... Unfortunatly the cards are so slow under Linux due to bad drivers, that they cant take adavantage of a bigger bandwidth. Correct me with TESTS(!!!) if I am wrong.
8. DirectX11: Yeah! What an big advantage under Linux!
So the summary is in reality:
Nice card, nothing new besides the name and some internal stuff no real world linux user would see. If you are a Windows user, check reviews of other sides!
Congratulations for such a nice and really usefull review!
Some of the DX11 features are also available as OpenGL extensions, and you can bet it will support all of OpenGL 3.2 and associated GLSL version.
rohcQaH
10-14-2009, 06:08 AM
3. Eyefinity: As you already mentioned: Not useable under Linux -> no advantage
What is usable on linux is the amount of connectors on the back. The ability to connect three monitors is the reason I'd choose this over the 4770. No more crawling under the desk to attach the projector.
As others said, the full eyefinity stuff will work too, eventually. But since the eyefinity version of the hardware isn't out yet, that'll have to wait (on both windows and linux).
4. Reduced Power consumption: Does the card reduce clocks and speeds when being idle work under Linux?
The review didn't say anything about power management, just power consumption. That's a hardware feature and a valid comparison to the other tested cards, which utilize the same driver and thus the same power management code.
Congratulations for such a nice and really usefull review!
Congratulations for such a nice and really useful review of the nice and useful review.
For the record, the performance comparison on linux is something you don't see on any other site, and the results are useful to anyone who considers buying a new ATI card in the near future.
Hasenpfote
10-14-2009, 06:54 AM
1. This implies point 4.
No! As I wrote, I dont care for the fabrication process as long as the card is fast and consumes less power! I can remember where the new fabrication process did not bring any advantages in power supply. Or did you see any AMD AthlonX2 with 35W in 65nm? I only own a 90nm with 35W! And you only can buy 45W CPUs!
2. Never had any problems with DVI or HDMI on Linux (be it fglrx, nvidia or the OSS drivers). I doubt 5xx0 will change this.Nice. Me neither. But I dont have a Display Port monitor. But I can assume a __REVIEW__ site has one and may do some TESTS with all kind of combinations. Did they do it in the review?
3. FUD. AMD has already showcased X-Plane running on Linux with Eyefinity.Yeah. And for consumers?
4. Obviously yes. My X1950 and 4850 show *identical* idle power consumption on Linux and Windows.Obviously? Well, obviously a new card is faster than an old card. So why do a review overall?
5. Yes, the single 6pin connector is significant. Other cards in the same performance bracket need more than that.And other cards dont need it but only use it for safety or to reduce load on a line for the power supply. But yeah, you are right! This is an HUGE advantage over other cards.
6. This is marketing speak.So I count this as an advantage and a big point FOR the 57x0 series.
7. OpenGL performance is identical on Windows and Linux (they use the same driver, duh). 2D performance is worse on Linux, but quite usable now (unlike, say, 6 months ago).I dont believe this, but as far as I dont have a proof for this, you are right.
8. Shader Model 5.0 is available through OpenGL, so yes this is a big advantage on Linux.Which games under Linux user Shader Model 5? Which games under Windows use DirectX11?
You have no idea what you are talking about and it shows.Well.. You may be right. But then dont call this a review but a benchmark-marathon without deeper investigations!
The lower nm production comes generally hand-in-hand with lower power, or higher clock speeds (better performance). They may also produce more chips off the same wafer, allowing lower costs. So overall, advantage.Well, lower power is wrong! (Athlon X2 EE SFF 35W in 90nm, 45W in 65nm!). The fact, that ATi can cut more out of a wafer is not an advtange of the costumer.
Unified processing architecture will make it better for shaders and OpenCL. So yes, it can be used under Linux. ... Some of the DX11 features are also available as OpenGL extensions, and you can bet it will support all of OpenGL 3.2 and associated GLSL version.Is it usable NOW? If not, dont name it as advatange but a future feature or something like that. I bought a 3870 2,5 years ago with the advantage of having open source drivers. WEll, when the drivers are there, the card is old and I can buy a new one. So IMO future feature dont count!
What is usable on linux is the amount of connectors on the back. The ability to connect three monitors is the reason I'd choose this over the 4770. No more crawling under the desk to attach the projector.Does it work? Normally this should be __tested__ in a so called review! Not only copying from marketing slides!
The review didn't say anything about power management, just power consumption. That's a hardware feature and a valid comparison to the other tested cards, which utilize the same driver and thus the same power management code.Sorry, but it seems I missed the table where the power comsumption is compared of all tested cards. Yeah, I think its on page 9 or 10, which are missing btw...
Congratulations for such a nice and really useful review of the nice and useful review.You're all welcome :)
For the record, the performance comparison on linux is something you don't see on any other site, and the results are useful to anyone who considers buying a new ATI card in the near future.For the record: It is only useful if he already owned an ATi card. I missed some cards from Nvidia. But fortunalty, Nvidia play no significant role on the graphic market.
Shader model 5.0, and dx11 are new - who seriously expects a great range of games to use them just yet? Battleforge, btw, does use something from dx11 (dxcl I think), and yes, the new radeon cards shine when that's used (only used in high detail settings if I remember rightly).
Low power generally comes from reduced nm fabrication - but they can also jam more features in (which in turn, may increase power usage). You really can't expect them to just make the same thing at lower nm - there's no point, it doesn't move the tech forward.
OpenGL 3.2 is probably already supported with the latest catalyst drivers - I don't know, I don't use them yet. All extensions are in place already, however, so you can program with it now.
Unified stuff as far as I understood it means the same calculation unit can be used for vertex or pixel (or geometry) shaders. That means it should be able to balance loads better, or re-use calc units for later pipeline stages - but that's more a hardware thing and is already being used by actually using the card.
rohcQaH
10-14-2009, 09:15 AM
Nice. Me neither. But I dont have a Display Port monitor.
So it just happens that Michael doesn't have a DisplayPort-Monitor, either. Do you expect him to spend big money just to verify an advertised feature?
I've read quite some windows reviews of the card, not a single one actually mentioned if DisplayPort works or not. Why are your standards for a small site like phoronix significantly higher?
Which games under Linux user Shader Model 5? Which games under Windows use DirectX11?
So what if you personally don't care about SM5 right now? Does that mean that nobody else is allowed to care, either? Does that mean that a review mustn't mention the card's capabilities?
Developers care about SM5, they want cards that can run their newest code. It's a real advantage to a graphics developer, right now.
If you really need to complain about little semantic details, DX11 support is not a "future" thing, the card actually supports DX11 today. It doesn't matter to you today, but this isn't an article about DX11, but about the GPU.
In this case the feature was even qualified with the phrase "if it matters to you". Duh.
The fact, that ATi can cut more out of a wafer is not an advtange of the costumer.
40nm is not a direct advantage for the customer. But the results of 40nm are, that is low power consumtion (and less cooling problems), smaller cards and cheaper production. You get more GPU power for less money.
Sure, you can say "But a GPU in 55nm with equal features is exactly as good". Right. But you're missing the point that an equal GPU in 55nm simply isn't possible.
For the record: It is only useful if he already owned an ATi card. I missed some cards from Nvidia. But fortunalty, Nvidia play no significant role on the graphic market.
Fun fact: I own a nvidia card, and I still think the review was useful.
From previous reviews I know how different cards compare (including nVidia cards). I have a basic understanding about the speed of a HD4870. With this comparison, I now have a basic understanding about the speed of the HD5770. That's all I wanted to know.
If you want to know more, you could try to educate yourself by reading more reviews. You could also try asking nicely, that usually gets you an answer. Or you can just make some unrealistic demands and then complain that they weren't met. Which is obviously way more fun.
Hasenpfote
10-14-2009, 09:39 AM
If you want to know more, you could try to educate yourself by reading more reviews. You could also try asking nicely, that usually gets you an answer. Or you can just make some unrealistic demands and then complain that they weren't met. Which is obviously way more fun.I was only critizing (one can say in a unfriendly way), that the advantages from the summary are mostly not tested in the review.
The whole review was just benchmarking. When I read the review, I know the cards is as fast as a 4870. Nothing more. For the rest, I have to trust the advertisement or other reviews, which are mostly done on Windows and have to trust, that the behaviour in Windows and Linux is the same.
So a company gets a Phoronix Award for bringing out new midrange hardware which is as fast as last generation performance hardware. Ok. So please give AMD and Nvida awards for their previous generations of hardware, too (just kidding ;) )
PS: And of course its fun to make AMD-fanboys upset (Kano is becoming too friendly... :D )
Apopas
10-14-2009, 09:48 AM
As of late I find the nVidia has been slipping, there cards are hot, over priced and their Linux drivers release schedule is poor.
Come on, they released a new version every week.
Svartalf
10-14-2009, 10:00 AM
Isn't opengl performance good enough?
Actually...it's important to have good Xv support because many video playback apps use Xv instead of alternatives because it works right (and better than the other options...) on other devices/drivers. Kano's right in his gripe- it DOES need to be fixed however it's arrived at on that one.
Svartalf
10-14-2009, 10:02 AM
The biggest question I would have would be of whether or not fglrx happens to support the OpenGL 2.1 ARB extensions for supporting the bulk of DirectX 10/11 functionality under OpenGL... I know that there were already proposed extensions by the ARB at the time that DirectX 10 came out...
BlackStar
10-14-2009, 10:33 AM
The biggest question I would have would be of whether or not fglrx happens to support the OpenGL 2.1 ARB extensions for supporting the bulk of DirectX 10/11 functionality under OpenGL... I know that there were already proposed extensions by the ARB at the time that DirectX 10 came out...
Fglrx supports 3.1 + most of the 3.2 spec (full support is bound to come soon). This covers ~99% of the DX10 functionality.
Some DX11 features are already supported via extensions (and have been supported for a long time - tesselation comes to mind). However, there's a lot of new stuff in DX11 that will take time to become formulized into OpenGL extensions or core features. My guess is that we will see a new OpenGL 3.3 release around Q1 or Q2 2010 that brings OpenGL into rough feature parity with DX11.
Most DX11-level features should become available as extensions before that - the issue is that Nvidia does not have DX11 hardware yet, so they don't have the incentive to create new OpenGL extensions for those features right now. The actual 3.3 spec will likely be delayed until Nvidia has hardware ready for release (sad but true).
I was only critizing (one can say in a unfriendly way), that the advantages from the summary are mostly not tested in the review.
The whole review was just benchmarking. When I read the review, I know the cards is as fast as a 4870. Nothing more. For the rest, I have to trust the advertisement or other reviews, which are mostly done on Windows and have to trust, that the behaviour in Windows and Linux is the same.
Well, starting with Catalyst 9.x, the behavior between Windows and Linux pretty much *is* the same (2d and video performance excepted, but Michael hinted that the latter may finally be resolved soon). How do I know? I have tested Ati cards on both operating systems: 3d performance, power consumption and OpenGL driver bugs are pretty identical.
4. Obviously yes. My X1950 and 4850 show *identical* idle power consumption on Linux and Windows.
Obviously? Well, obviously a new card is faster than an old card. So why do a review overall?
Do you have any indication that power consumption may be different between OSes on 5xx0? Because all the indications *I* have is that it will be identical.
5. Yes, the single 6pin connector is significant. Other cards in the same performance bracket need more than that.
And other cards dont need it but only use it for safety or to reduce load on a line for the power supply. But yeah, you are right! This is an HUGE advantage over other cards.
Are you really implying that te 4870 has 2x 6-pin connectors just for fun? The GT260 also has 2x 6-pin connectors. Is that just for fun too?
The fact that 5770 gets you the same or better performance with a single connector indicates a HUGE advantage in power consumption over both cards - which most reviews confirm.
You can twist the facts all you like, but the numbers speak for themselves. The 5770 *is* more power efficient and the single 6-pin connector is the direct consequence of that.
Chad Page
10-14-2009, 12:29 PM
Direct3D 11 brings compute shading support. While the 48xx can already work with OpenCL, the 5xxx will do better probably.
In the Gallium3D era the extra shader flexibility will probably be even *more* useful under Linux. I think we'll see very interesting applications in a year or two...
mattmatteh
10-14-2009, 02:18 PM
So is there any documentation released ? Is there any start on the opensource driver ? Sounds like nice hardware ( minus the fan and extra power connector), but seems like it will be 2 or 3 years before it has a stable (open source) driver ? I am still waiting for r600 and r700 to become stable enough that i would buy.
matt
smitty3268
10-14-2009, 02:18 PM
Direct3D 11 brings compute shading support. While the 48xx can already work with OpenCL, the 5xxx will do better probably.
In the Gallium3D era the extra shader flexibility will probably be even *more* useful under Linux. I think we'll see very interesting applications in a year or two...
I believe the r700 cards support DC4.1 and OpenCL1.0, while the r800 cards will give full DC5 and a to-be-created OpenCL 1.1 version.
Chad Page
10-14-2009, 02:20 PM
The 5xxx is close enough to the 4xxx that it won't be too far behind, esp since AMD people are working on the open source drivers.
I figure it'll be a few weeks/months? until the documentation's released though. Legal can be odd. ;)
bridgman
10-14-2009, 03:01 PM
I don't think the HD5xxx parts will need much in the way of new documentation, at least not until the Mesa devs start to work on supporting DX11 features.
There'll probably be a new ISA doc to cover the shader differences -- we'll know more over the next few weeks.
mattmatteh
10-14-2009, 03:02 PM
Thats good to hear it should be supported soon. I thought AMD/ATI was going for same day release with linux ? ( fglrx, all documentation, open source would nice with at least something started)
Chad Page
10-14-2009, 03:04 PM
I don't think the HD5xxx parts will need much in the way of new documentation, at least not until the Mesa devs start to work on supporting DX11 features.
There'll probably be a new ISA doc to cover the shader differences -- we'll know more over the next few weeks.
Ah, very nice indeed...
(btw, on the non-open side, I'm looking forward to playing with OpenCL on my 4830. :) )
And I think fglrx already works with the 5k's... or will in 9.10. The support infrastructure is there to get free 2D working quickly, but not instantly ;) It's a far cry from the 1xxx and 2xxx series where there was no support for a very long time.
bridgman
10-14-2009, 03:08 PM
Thats good to hear it should be supported soon. I thought AMD/ATI was going for same day release with linux ? ( fglrx, all documentation, open source would nice with at least something started)
Nope, that was never the plan. We *are* going for same day fglrx support, since fglrx can leverage common development effort (the main advantage of a proprietary driver), but the open source work starts shortly before launch when everything has been locked down and we start coding/testing on commercially available hardware.
kernelOfTruth
10-14-2009, 06:52 PM
Nope, that was never the plan. We *are* going for same day fglrx support, since fglrx can leverage common development effort (the main advantage of a proprietary driver), but the open source work starts shortly before launch when everything has been locked down and we start coding/testing on commercially available hardware.
I hope you guys will add support for 5750 and 5770 to the opensource drivers soon enough - at least when their general availability gets better here (meaning during the next 3-8 weeks) since I'm planning to get me one of those & thanking you guys via $$$ :)
tmpdir
10-15-2009, 03:07 PM
Hi Yall & Michael,
um, like dude, di you just say ....
"The Unified Video Decoder 2 (UVD2) support with ATI Avivo HD may also be ignored by Linux users at this time, as we have been talking about the X-Video Bitstream Acceleration (XvBA) for a year now and it still has yet to be usable by ATI Radeon customers, but it finally will be in the very near future. UVD2 in ATI Radeon hardware will finally mean something on Linux, but we will have to wait to share more until permitted."
I second that...
I just convinced myself to buy a nvidia card and I read this... but this time i'm not waiting a whole year ;)
I don't think that it will take 1 year, but with more drawbacks than you would like ;)
GreekGeek
10-15-2009, 07:14 PM
Hi Yall & Kano,
"Drawbacks?" Sounds like you have heard something from someone who has used this as yet unreleased software.
Um, care to share?
Greekgeek :-)
Well you can try the mplayer needed, thats mplayer vaapi. It learnt OSD with nvidia now, but h264 which does play only play h264 profile 4.1 and not 5.1. that is no problem using vdpau with the same mplayer but differnet vo/vc.
GreekGeek
10-15-2009, 07:57 PM
Hi Yall & Kano,
I was thinking more about drawbacks & FGLRX-XvBA-UVD2.
I gather you are pointing out issues with the nVidia implimentation, which is a whole different bundle of joys....
One hope I have, is that the playback issues with video (the popped up again, watching Pink Panter cartons, just now), in particular video tearing.
But I maybe confussing "video accelleration," with "Vsynch" or hoping for a silver bullet.
Still, my new 5850 does play Civ-IV, the way it is meant to be played! ;-)
nVidia lost my vote, with the way the screwed with the code on "BATMAN: ARKHAM ASYLUM" game.
*BFN*
Greekgeek.
For standard content just use opengl as output and amdcccle to force quality mode/vsync on. Interesting is only 1080p.
GreekGeek
10-15-2009, 11:12 PM
Hi Yall & Kano,
openGL does get rid of the video-tearing and on my new build, works fine. I think I tried it with my "old build," and the hardware was too slow.
VLC barfs. Xine and SMP are good. But with .mkv, SMP will sometimes shrink the video playback size, in 16.9 aspect-which is odd.
Again, thanx for the heads up. :-)
Greekgeek. :-)
Well i often use VLC 1.0.2. You can set it to opengl output as well. My favorite player is mplayer however as i don't know how to enable something similar to -af volnorm with vlc (Xine has got a way to enable that too, just differently). Xine works with opengl with an additional override in .xine/config:
video.output.opengl_renderer:2D_Tex
it took a while till i found that out - otherwise i had no gui for vdr. But for 1090p even fast systems get out of sync sometimes when rendered to opengl. Saidly mplayer has got that problem too when the input is not 100% optimal, then a/v sync is lost and it does not resync. Win player do that, i really want that feature on linux too, that would fix lots of problems. It is useless to test 1080p docus when you can not see the speaker to find those issues however.
Hasenpfote
10-16-2009, 02:00 PM
The ability to connect three monitors is the reason I'd choose this over the 4770. No more crawling under the desk to attach the projector.And you know what REALLY funny is:
If you have more then 1 monitor connected, the idle power consumption goes from 18W to 50W!
Source (in German):
http://ht4u.net/news/21031_multi-monitor-betrieb_verwandelt_desktop-grafikkarten_in_stromfresser/
Unfortunalty this is not ATi specific. Nvidia has the same problem (in the article 275 and 295 are named).
Thank you graphic vendors for lieing to your advanced customers! Your idle power is NOT the defined idle power!
And thank you to all websites doing so called reviews! You ALL failed!
mtippett
10-16-2009, 02:30 PM
And you know what REALLY funny is:
If you have more then 1 monitor connected, the idle power consumption goes from 18W to 50W!
Source (in German):
http://ht4u.net/news/21031_multi-monitor-betrieb_verwandelt_desktop-grafikkarten_in_stromfresser/
Unfortunalty this is not ATi specific. Nvidia has the same problem (in the article 275 and 295 are named).
Thank you graphic vendors for lieing to your advanced customers! Your idle power is NOT the defined idle power!
And thank you to all websites doing so called reviews! You ALL failed!
For driving displays you need a particular bandwidth of memory to allow the display controllers to read out memory and still leave headroom for 3D and video clients to do initial rendering (until the clocks ramp up). This is immutable.
For the single monitor case, you can use very low clocks quite safely. The more monitors you add the higher the memory clock (and possibly the engine clock) will need to be raised. Nothing sinister, but fundamental resource constraints.
Regards,
Matthew
rohcQaH
10-16-2009, 04:33 PM
And you know what REALLY funny is:
If you have more then 1 monitor connected, the idle power consumption goes from 18W to 50W!
I was expecting it to increase, but that's more than I imagined. Thanks for the link! And thanks for the clarification, mtippett!
I'd hope the drivers are smart enough to to power down when the monitors are off or not connected though. (for example when the computer is doing some simulations while I'm afk)
As that's a technical problem that won't change, I'd still pick the 5770. It's not 13W with 2 monitors, but still less than other comparable cards. And as I just bumped my head against the desk when switching connectors, I really want that third output ;)
As soon as phoronix says "drivers are ready", I'm buying.
Hasenpfote
10-16-2009, 05:33 PM
For the single monitor case, you can use very low clocks quite safely. The more monitors you add the higher the memory clock (and possibly the engine clock) will need to be raised.But why not being honest to the customer and saying "Our cards need idle 18W, when you only use one monitor! Otherwhise its 50W (or 60W or whatever)!" All reviews I read about the 5700/5800 series where saying, how great the idle power consumption is (which is true!). Especially with a feature like Eyefinity and prices for a 22" starting at 125Euros one can assume that people are using more than one monitor (if the girlfriend/wife does not have a different opinion). And suddenly it isnt so great anymore and people are asking, why their cards are so loud and being silent in the reviews. (In the link, they say, that the fans are running with 1400RPM.) The reviews a bought by AMD, AMD sucks, never again AMD, buy Nvidia, etc. pp. On heise.de in the comments people are still complaining about driver issues with AMD (on my side I had never any problems since my 9500pro). So some things can become kind of a longliving urban legend which can cost customers.
So to quote rohcQaH: Are the drivers (now or in the future) intelligent enough to determine, if a monitor is active or not and adjust the clocks as needed (with steps like 50% for 2 monitors, 75% for three, etc.)? Are there any plans to let the user decide, if he wants to have lower clocks and problems (on 3dcenter.org they say something about flicker) or higher clocks and no problems?
Thank you in advance for an answer.
Ant P.
10-16-2009, 05:57 PM
The more monitors you add the higher the memory clock (and possibly the engine clock) will need to be raised. Nothing sinister, but fundamental resource constraints.
What if it were possible to lower the refresh rate when it's not needed? I assume there's at least some LCDs out there that can run at 24Hz for 1080p24 and the like. It could be an extra timeout in the same way the DPMS stuff works.
rohcQaH
10-16-2009, 06:08 PM
But why not being honest to the customer and saying "Our cards need idle 18W, when you only use one monitor! Otherwhise its 50W (or 60W or whatever)!"
because that isn't the full truth either. Since memory bandwidth has to match output bandwidth, I'd expect a single 2560x1600-monitor (dual-link) to draw more power than your tiny 10" netbook display (if you can fit the 5870 into your netbook - or the other way round ;))
Ati releases idle watts and TDP, those are two interesting baseline values that are useful enough to compare different cards (even against nvidia's cards, no cheating there). Marketing presentations just won't have huge tables with wattages under different conditions. Very few people would even read them.
Hasenpfote
10-17-2009, 05:35 AM
because that isn't the full truth either. Since memory bandwidth has to match output bandwidth, I'd expect a single 2560x1600-monitor (dual-link) to draw more power than your tiny 10" netbook display (if you can fit the 5870 into your netbook - or the other way round ;))Ok, this is an argument.
Ati releases idle watts and TDP, those are two interesting baseline values that are useful enough to compare different cards (even against nvidia's cards, no cheating there). Marketing presentations just won't have huge tables with wattages under different conditions. Very few people would even read them.But marketing can prepare slides to show how Nvidia currently sucks (http://tinyurl.com/yjzx42s :D ) and AMD prepares huge tables of technical information. And review sites and advanced users are interested in this technical information. Of course, if this got released in public documents, it's no fault of ATi but the reviewing sites and the users.
Well, in the end the difference in power consumtpion is no a real relevant buy factor if affects ATi and Nvidia. But I was quite shocked, that this was not discussed earlier (or I just overlooked it).
I really would like to know why ATI cards are soft limited to H264 L4.x decode and you have to google to finds hacks around that to use L5.1 (for DvXA). BD only needs L4.1 thats fine, but there are lots of L5.1 files out there which can play Nvidia cards - even using vdpau on Linux. That's a major drawback for video playback - even using the "mainstream" os.
BlackStar
10-20-2009, 04:26 AM
If by "major drawback" you mean "99.9% of the customers simply won't care", then you are right. (You and I may be downloading H264 L5.1 videos, but the vast majority will never watch anything other than BD and Youtube "HD").
That said, I'd like to know the reason as well. :)
Edit (disclaimer): I just bought a fanless 9500 for my XBMC HTPC. VDPAU works, but I'm rather underwhelmed on the whole (sync issues, mainly, as well as not-very-good 2D performance. My previous 7600GS was much, *much* better in 2D).
For youtube you should enable vsync in nvidia settings and load
nvidia-settings -l
on startup. vdpau should be using vsync - and be sure you really use the latest BETA drivers!
BlackStar
10-20-2009, 10:33 AM
For youtube you should enable vsync in nvidia settings and load
nvidia-settings -l
on startup. vdpau should be using vsync - and be sure you really use the latest BETA drivers!
Sync issues, not vsync issues! As in audio out falling out of sync with video or video speeding up / down. I don't know whether this is an XBMC or a Nvidia issue, but I've disabled VDPAU until this can be resolved (my single-core Athlon 64 can handle 720p just fine and I don't have any 1080p content right now).
That said, I also have vsync issues - this card refuses to sync to the HDMI monitor as long as a VGA monitor is connected (using separate X screens). This causes pretty bad tearing on my TV, unless I disconnect the VGA monitor first - not very fun.
I'm using the 190.36 drivers from the VDPAU team (https://launchpad.net/~nvidia-vdpau/+archive/ppa), but I've tried everything several other versions from 180.xx, 185.xx and 190.xx with no change. If I actually knew that I'd encounter these kinds of problems, I'd probably have gone with an Ati card and the open drivers instead...
You know that you can control the VDPAU vsync via an environment var? VDPAU_NVIDIA_SYNC_DISPLAY_DEVICE should be set to the devicename nvidia-settings reports for it.
BlackStar
10-20-2009, 04:32 PM
You know that you can control the VDPAU vsync via an environment var? VDPAU_NVIDIA_SYNC_DISPLAY_DEVICE should be set to the devicename nvidia-settings reports for it.
No. Is this different than the 'sync to monitor' option in nvidia-settings? (This option works correctly if I connect two VGA monitors, but not if I connect one VGA and one HDMI monitor).
I'll have to try that and see if it fixes the vsync issue, even if it's only for VDPAU (I guess I can live with normal tearing if video displays correctly). That would be a huge step forward.
Any ideas about the video-audio desync and speed up/down issues?
The a/v sync is mainly a player problem. With good input mplayer + vdpau has got the least problems with it - as long as you don't use -demuxer lavf.
evanjfraser
10-21-2009, 03:55 AM
I bought my first ever ATI card last week (XFX 5750), and I've been buying video cards since 1993. I've upgraded my desktop to karmic and I'm using the bundled catalyst restricted driver.
I bought the 5750 because it seemed the first worthwhile video card upgrade (to me) to my 7900GS that would be short enough to fit in my HTPC case.
I'm hanging out for AMD to sort out the GPU playback offloading (ala VDPAU) and If they don't get that released soon I will be quite a disappointed radeon owner... It was a bit of an impulse purchase based on phoronix's review so here's hoping they're not misleading us!
Everything is going well after the upgrade so far, halflife2 runs nicely at 1080p now.
Is there anyway to get rid of the AMD unsupported hardware overlay? I don't think there are any newer drivers out as catalyst 9.9 (latest available from amd.com) didn't work with this card at all.
Cheers, Evan.
GreekGeek
10-21-2009, 04:51 AM
Hi Yall & Evan,
this will kill the watermark thang:
echo -n 1a12e6c2f7e613b887d92e943a40799e:223cd5fbc4c82295b 7ee1ea30b7818b32a27d6f5cfd75095c68d67b409754faa292 7:2820d5a193d375dae4ed4dac087240ab782583a7c683728b b7bb19ac0b781bad2873d4f593d0758de4bf4aa30f7040ab79 24dea6c1d1718cb7e019a10e2618ff282bd4f5 > /etc/ati/signature
<edit>
Issue that cmd @ the cmd line.
Log our of X & then back in & whamo, no watermark thang...
<edit>
Kudos to Kano.
Oh and of course the ATI guruz, what writes the FGLRX code in the first place! :-)
*BFN*
GreekGeek :-)
evanjfraser
10-21-2009, 05:12 AM
Hi GreekGeek, thanks for helping!
Unfortunately it doesn't seem to have worked.
I'm new to ATI drivers except for a short unsuccessful stint trying to use ATI cards for Maya in linux back in 2001&2.
Cheers, Evan.
Hi Yall & Evan,
this will kill the watermark thang:
echo -n 1a12e6c2f7e613b887d92e943a40799e:223cd5fbc4c82295b 7ee1ea30b7818b32a27d6f5cfd75095c68d67b409754faa292 7:2820d5a193d375dae4ed4dac087240ab782583a7c683728b b7bb19ac0b781bad2873d4f593d0758de4bf4aa30f7040ab79 24dea6c1d1718cb7e019a10e2618ff282bd4f5 > /etc/ati/signature
<edit>
Issue that cmd @ the cmd line.
Log our of X & then back in & whamo, no watermark thang...
<edit>
Kudos to Kano.
Oh and of course the ATI guruz, what writes the FGLRX code in the first place! :-)
*BFN*
GreekGeek :-)
GreekGeek
10-21-2009, 05:37 AM
Hi Yall & Evan,
dang! The trick worked fine, for my 5850, with the Cat 9.9 & with the slightly newer driver, in the ATI-Stream sdk driver.
Hmmm. Do you get any error messages?
Which distro of Tux do you have?
Have you got a modifed kernel?
Have you checked the drivers that came with your card-a Linux driver may have been included?
My Sapphire card did not.
Well, at least you will not have to wait long, until 9.10 is out.
*BFN*
GreekGeek
evanjfraser
10-21-2009, 05:51 AM
I didn't even think to check the CD for linux drivers. I didn't think there would be any.
There were no errors, the echo command worked fine. I'm using ubuntu 9.10 with the standard kernel.
Thanks again,
Evan.
Hi Yall & Evan,
dang! The trick worked fine, for my 5850, with the Cat 9.9 & with the slightly newer driver, in the ATI-Stream sdk driver.
Hmmm. Do you get any error messages?
Which distro of Tux do you have?
Have you got a modifed kernel?
Have you checked the drivers that came with your card-a Linux driver may have been included?
My Sapphire card did not.
Well, at least you will not have to wait long, until 9.10 is out.
*BFN*
GreekGeek
evanjfraser
10-21-2009, 05:53 AM
Newp, xp and vista drivers only :(
I didn't even think to check the CD for linux drivers. I didn't think there would be any.
There were no errors, the echo command worked fine. I'm using ubuntu 9.10 with the standard kernel.
Thanks again,
Evan.
smitty3268
10-21-2009, 06:04 AM
The 9.10 drivers will probably be out Thursday, which should fix it.
luther138
10-28-2009, 01:54 PM
I read through the release notes for the 9.10 Catalyst drivers
it looks like there isn't any support for this card :(
Can anyone confirm whether or not the new drivers work with this card, Specifically Triple monitor support? Oh how I am waiting patiently for ATI, to make this happen for me.
Thanks
-Luther138
(Unless someone has a suggestion for a different single card that could do 3 monitor support in Linux)
bridgman
10-28-2009, 02:52 PM
My current understanding is that the code for HD57xx support was in the 9.10 posting but HD57xx was not QA tested in that release so it's treated as "unsupported hardware".
I'm trying to track down the launch driver we provided to our board partners and see if we can get that posted.
You could disable that /etc/ati/{control,signature} crap too or at least put updated updated control files to download. If you like to mark beta versions then look at nvidia, the splash shows it. A permanent watermark is stupid ^ 10.
Do amd actually release beta drivers, or are they more leaked accidentally?
bridgman
10-28-2009, 05:55 PM
We release beta drivers to a beta test group. Occasionally beta drivers get leaked to the public. .
I seem to remember beta drivers having the watermark permanently enabled on all GPUs but I haven't looked at them for a while.
bridgman
10-28-2009, 05:58 PM
You could disable that /etc/ati/{control,signature} crap too or at least put updated updated control files to download. If you like to mark beta versions then look at nvidia, the splash shows it. A permanent watermark is stupid ^ 10.
That was basically what happened at Chernobyl, wasn't it ? The alarm was annoying so they turned it off ?
The purpose of the watermark is to indicate that the drivers have *not* gone through QA on that particular GPU. Rather than updating the control files to disable the warning I would rather try to post a build which *has* gone through QA on HD57xx.
Would not be the worst idea to make public beta drivers then.
energyman
10-28-2009, 11:48 PM
can you imagine the whining when a beta driver crashs boxes?
evanjfraser
10-29-2009, 02:45 AM
Hi Luther,
I couldn't get X to load with catalyst 9.10 and my 5750. Although the stream beta drivers work, so you could try them. There was also a hack mentioned by kano to disable the testing overlay relating to the stream beta drivers in another thread here. Unfortunately I only have the one monitor and can't test dual or triple head for you.
Cheers, Evan.
@energyman
That's the "normal" case anyway when you test new features - even on "stable" drivers. Nothing new to ATI "beta" drivers. Nv "beta" drivers have in most cases production quality however.
If AMD released public beta drivers, I can think of a few non-technical reasons why this could go wrong:
1) If a bug is there, it will raise undue complaints even if it's long fixed in internal builds.
2) Support - AMD will have to deal with more people wanting support for a beta driver, even if AMD takes the stance of not supporting it properly. It won't stop people trying to get support, and may slow down driver development in general.
3) Support additional: AMD will have a harder time tracking down customer problems if they're using beta builds. It's better to have defined milestone builds to help sort out problems rather than going over several beta builds with various changes to find a problem that may no longer exist.
Did you ever get fglrx SUPPORT? You get one driver per month, no matter if it works or not.
Bridgman hangs around the forums, and he does have an interest when there's proper complaints of it not working.
I'm not a workstation customer.
As I mentioned, not much will stop people contacting AMD trying to get support, and dealing with that could slow things down.
It never happend that you got a new driver when something did not work as "full" release. Only U managed to get prerelease drivers which they only published "half", no real original installer file was available.
V!NCENT
10-29-2009, 08:32 AM
Traditional software development; BETTER! FASTER! RUSH IT OUT THE DOOR NOW!!!!! (took 20% of total development time and costs)
Ohhhhhhhhhhhh shit! It doesn't work... full of bugs. Ten thousand users complain...
Fixing the codebase = 80% of total development time and costs.
... *sigh*
The problem with fglrx is that it lacks about 2-3 month behind in several aspects:
a) kernel support
b) xorg support
c) hardware support
Nv tries to fix all 3 issues with new drivers in a way it does not take so long. There you get often kernel suport even for eary rc kernels, ati thinks they are forced to wait till the kernel is stable before they are allowed to download and test it. same applies for xorg. For hardware support issues the point is that most hardware basically works, but they tend to annoy people whereever they can using watermarks. Use a beta mark in logfile or as splash, thats really enough! To remove all watermarks look there:
http://www.phoronix.com/forums/showthread.php?t=19875
vBulletin® v3.8.4, Copyright ©2000-2009, Jelsoft Enterprises Ltd.