Nvidia renames old stuff again [Archive]

kernelOfTruth

08-28-2009, 06:01 PM

that's one of the reasons why I stopped buying nvidia cards:

they're not honest to their customers :rolleyes:

and guess how they do energy-savings / energy efficient cards ?
they down-clock cards and sell them as green editions; adaptive up- and/or down-clocking on non-green cards often is broken or simply not implemented

the only "joker" they're having right now is still their somewhat better binary blob but that will change soon

I'm glad I bought and ATI/AMD card earlier this year :D

deanjo

08-28-2009, 06:06 PM

and guess how they do energy-savings / energy efficient cards ?
they down-clock cards and sell them as green editions; adaptive up- and/or down-clocking on non-green cards often is broken or simply not implemented

Actually they do better then their ATI brethren at least until running in full bore premium performance mode. ATI's cards have been brutal on non premium modes for power consumption.

energyman

08-28-2009, 06:11 PM

At least they will work fully featured in linux upon launch.:p
.

but for how long until the bumps break and the chips fall off? :D

deanjo

08-28-2009, 06:12 PM

but for how lung until the bumps break and the chips fall off? :D

Those chips are not even the same series.

Back to power consumption:

http://www.tomshardware.com/reviews/radeon-hd-4890,2262-13.html

http://www.tomshardware.com/reviews/budget-radeon-geforce,2364-17.html

energyman

08-28-2009, 06:31 PM

Michael

08-28-2009, 06:58 PM

renaming, a game for the whole family. So much fun to rip off costumers:

http://www.semiaccurate.com/2009/08/28/nvidia-renames-2xx-parts-3xx/

Wow, SemiAccurate finally rolled around... I saw that site back in January of 2008.

energyman

08-28-2009, 07:21 PM

they have some nice stuff there Michael - without cutting into your benchmarking business ;)

some more on bumpgate, Nvidia's g300 and when it will maybe become reality. TSMC+40nm. Why is it so hard to get a HD4770 ....

The Magny Courts article posted a view days ago was the best I have seen so far.

benmoran

08-28-2009, 08:44 PM

Maybe this explains whey those two ex-AMD bigwigs were hired recently ;)

deanjo

08-28-2009, 09:03 PM

ugh - tomshardware...

http://www.xbitlabs.com/articles/video/display/radeon-hd4890_6.html#sect0

and nvidia's insurer - as DM - claim, that not only integrated parts are defective - desktop parts just go on a bit longer.

Ya and they are parts that are no longer available. G84's and G86's of that stepping are no longer used.

deanjo

08-28-2009, 09:05 PM

Wow, SemiAccurate finally rolled around... I saw that site back in January of 2008.

SemiAccurate is a perfect name for Charlie's site.;)

energyman

08-28-2009, 09:08 PM

Maybe this explains whey those two ex-AMD bigwigs were hired recently ;)

http://www.semiaccurate.com/2009/08/21/nvidia-finally-understands-bumpgate/

if one of these is a packaging guy - maybe ;)

oh and this is Sony:
http://www.semiaccurate.com/2009/08/10/sony-admits-14-defective-nvidia-notebooks/

>In any case, we repeat what we have been saying all along, all Nvidia 65nm and 55nm chips are defective, there isn't a good one out there. Sony publicly names the G84 and G86. Nvidia's insurance company names ten parts Nvidia has tried to make claims on, G86, G86A2, G84, C51, G72, G72M, G73, G72A3, MCP67 and NV42. To top it off, we are hearing that the G9xs are starting steep climb of their failure ramp, and some big OEMs are getting very nervous. If anyone has access to the aggregate numbers, send them this way.

2010 could be an interessting year, if you are a nvidia fanboy ;)

energyman

08-28-2009, 09:09 PM

Ya and they are parts that are no longer available. G84's and G86's of that stepping are no longer used.

and that change what when a card or laptop dies?

SemiAccurate is a perfect name for Charlie's site.;)

and he was so far entirely correct about bumpgate.

deanjo

08-28-2009, 09:12 PM

and that change what when a card or laptop dies?

and he was so far entirely correct about bumpgate.

Lol, in the long history of Charlie articles that is the ONLY article he got even close too. You fire enough shots into air eventually you hit a target.

The White Sox may win another pennant someday too.

energyman

08-28-2009, 09:17 PM

then - please give examples

deanjo

08-28-2009, 09:19 PM

Just for example the 4870 vs the GT 280 comparison...

http://www.theinquirer.net/inquirer/news/1040792/gt200-scores-revealed

On the high end, the R700 spanks them by wide margins, but those numbers will have to wait a bit.

deanjo

08-28-2009, 09:22 PM

Another bunch of BS,

http://www.theinquirer.net/inquirer/news/1050874/nvidia-trying-x86-chip

http://www.theinquirer.net/inquirer/news/1051248/nvidia-announces-x86-chip

http://www.theinquirer.net/inquirer/news/1051599/nvidia-hoodwinks-reviewers-mythical-gt275s

Shall I go on?

energyman

08-28-2009, 09:43 PM

Just for example the 4870 vs the GT 280 comparison...

http://www.theinquirer.net/inquirer/news/1040792/gt200-scores-revealed

nice out of context quote you did there

On price/performance, they lose badly, really badly, to the 770. On the high end, the R700 spanks them by wide margins, but those numbers will have to wait a bit.

is the complete part. Which changes the meaning a lot.

and the other three links... well, where is that gt275?

also - these are pretty fresh articles - when CD started about bumpgate people were calling him a liar too - also a lot later the truth came out. So.. how about a bit of patience?

edit: found it gtx 275. And HD4890 does well against them

deanjo

08-28-2009, 11:32 PM

nice out of context quote you did there

On price/performance, they lose badly, really badly, to the 770. On the high end, the R700 spanks them by wide margins, but those numbers will have to wait a bit.

is the complete part. Which changes the meaning a lot.

Compare the prices and performance.....

and the other three links... well, where is that gt275?ROFLMAO! open your eyes bud, they are everywhere. Including in my systems.
http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=2010380048%201305520548%20106792634%201067947241&name=GeForce%20GTX%20275

also - these are pretty fresh articles - when CD started about bumpgate people were calling him a liar too - also a lot later the truth came out. So.. how about a bit of patience?
I can go much further back if you like...

Zhick

08-29-2009, 03:55 AM

Actually deanjo is right, nVidia does have the better price/performance ratio in the highend sector right now. A buddy of mine just asked me what to get, a HD 4870/4890 or GTX 275. I just went by impulse at first and told him the 48X0 parts were probably better performance- and price-wise, since that's been true for the last few months, but then had a look at some benches and realized that the GTX 275, while a bit more expensive then the ATi cards, also was a good (bigger) bit faster.
ATi really needs to get out a new chip-generation... well I read somewhere that "Evergreen" is supposed to come out next week anyway, right?

Jaikers

10-01-2009, 04:04 PM

renaming, a game for the whole family. So much fun to rip off costumers:

http://www.semiaccurate.com/2009/08/28/nvidia-renames-2xx-parts-3xx/

wow, just wow. stuff like thsi makes me cry a little inside. Dont worry though it's really just on the inside.

Panix

10-07-2009, 09:35 AM

Nvidia can rename all it wants, at least it works!

ATI support in Linux sucks!

They dropped support for older hardware. I read about older Nvidia hardware handling all of the 'advanced' features of virtualization, graphics, desktop effects etc. Meanwhile, ATI is still at the same stage it was four or five years ago. You can Google and discover the same complaints and support questions for these issues.

energyman

10-07-2009, 09:43 AM

Panix

10-07-2009, 10:05 AM

oh, and nvidia did not drop support for old hardware?

btw - the 'dropped' support was only 'dropped' for cards well supported by the open drivers, so what are you complaining about?

And 'the nvidia drivers just work' is a myth.

for example:
http://forums.gentoo.org/viewtopic.php?p=5999881#5999881
Well, when you read of complaints and this topic after googling, I somehow am convinced, okay?

My RV250 Radeon Mobility 9000 hardware is unable to use desktop effects properly in several different distros. Mostly Ubuntu but I don't mind that as I am more inclined to use a different distro. But, I've read that older Nvidia hardware has no problem with desktop effects.

It's not a big deal to me as I don't need that feature but I read that it's been an issue as far back as 2004!

nanonyme

10-07-2009, 10:34 AM

My RV250 Radeon Mobility 9000 hardware is unable to use desktop effects properly in several different distros. Mostly Ubuntu but I don't mind that as I am more inclined to use a different distro. Should pretty much work out of the box with modern drivers and is getting better all the time.

Ant P.

10-07-2009, 05:19 PM

I used compiz for a long time with a 9250, so a 9000 shouldn't really have a problem. Besides, it was originally demoed on something like a 8500...

To be fair my older GeForce FX5200 felt slightly faster - until it died one day :D

Panix

10-08-2009, 09:02 AM

So, MY ATI Radeon Mobility 9000 is *different* then? :rolleyes:

I demo'd it in various distros, Live CDs and found that when I enable desktop effects, I had various instances of hard crashes. 'Had to power off and reboot the machine.

I didn't say the normal driver didn't work or wasn't installed. I'm talking about graphics and features.

There are a bunch of threads on LaunchPad about the bug or references to similar issues at least with Ubuntu.

I'll just deal with it. I do think right now, though, that Nvidia is a safer option for video/graphics, however. Or even Intel. When researching the 'issue', I came across so many complaints of ATI issues/problems so maybe they are swamped with issues to address. Maybe there's a manpower problem. I dunno but I perceive an issue with my laptop's card, period. :(

bridgman

10-08-2009, 10:12 AM

I think the other poster is saying that there may be a problem specific to your system.

Have you gone through all the usual troubleshooting steps, things like dropping the AGP bus speed until you find one that is reliable on your system ?

nanonyme

10-08-2009, 03:30 PM

So, MY ATI Radeon Mobility 9000 is *different* then? :rolleyes:

I demo'd it in various distros, Live CDs and found that when I enable desktop effects, I had various instances of hard crashes. 'Had to power off and reboot the machine.Try Fedora 12 Beta LiveCD when the beta is out, just for kinks? :) Fedora has pretty much the most bleeding-edge drivers there are. Should at least give you a pretty good impression on whether your card has been getting better support lately. (and yes, what bridgman said if you want to debug the issue for your current system)

misiu_mp

11-11-2009, 07:09 PM

I have a laptop with radeon mobility 9600 and it worked good with effects at least since fedora 8.
I do notice that the free ati drivers have considerably lower performance than the equivalent nvidia card.
On windows radeon 9600 is supposed to be comparable to geforce 4 Ti, while on linux it cant even compete with GeForce 3 Ti.
I wouldnt expect that to be fixed since these are very old cards and there is probably very little practical use of fixing that. I would like to oppose that POV because games as alienarena and Scorched3D could be made playable on that once popular hardware if the full potential of r300 was taken advantage of. I see those games as some of the most 3D-demanding free titles out there.
I see some hope in the pending galium3D r300 rewrite.

BlackStar

11-12-2009, 08:22 AM

I have a laptop with radeon mobility 9600 and it worked good with effects at least since fedora 8.
I do notice that the free ati drivers have considerably lower performance than the equivalent nvidia card.

You are comparing apples to oranges. The open source stack is not optimized for performance.

For a fair comparison, compare either fglrx vs nvidia or radeon/radeonhd vs nouveau.

On windows radeon 9600 is supposed to be comparable to geforce 4 Ti

The 9600 hardware runs circles around the geforce 4 in performance/ image quality/capabilities. This difference is caused by the drivers.

Gallium is expected to perform close to 60-70% of the closed source stack (and maybe higher for simpler programs and older hardware, like R300). To get higher numbers than that, you'd have to use game-specific optimizations and other hairy hacks that would complicate the codebase significantly. As a point of reference, fglrx contains several million lines of code; the open drivers are an order of magnitude simpler than that - you can't expect them to perform the same.

60-70% should actually be enough for all uses other than hardcore gaming or professional 3d (modeling, CAD, this kind of stuff). If you need more than that, you'll have to use the closed-source stack.

misiu_mp

11-12-2009, 08:48 AM

You are comparing apples to oranges. The open source stack is not optimized for performance.

For a fair comparison, compare either fglrx vs nvidia or radeon/radeonhd vs nouveau.

The 9600 hardware runs circles around the geforce 4 in performance/ image quality/capabilities. This difference is caused by the drivers.

My point precisely - the free drivers are not good enough. As ati claims the end of support is for 'well' supported by free drivers cards, it basically sentences the owners of those for poor performance.
Whats the point of having free drivers if they are not useful for decent performance. Performance is the reason for (discrete) graphic cards.

To get higher numbers than that, you'd have to use game-specific optimizations and other hairy hacks that would complicate the codebase significantly.
I find it hard to believe that as much as 40% of a graphic card performance is dependent on application-specific optimizations (i think that could be more true in the infamous days of the fx5xxx series).
That 'good enough' approach would mean that linux graphics will always stay inferior. Thats not good enough.

bridgman

11-12-2009, 10:00 AM

My point precisely - the free drivers are not good enough. As ati claims the end of support is for 'well' supported by free drivers cards, it basically sentences the owners of those for poor performance.

Sorry to disagree, but that is not at all what we said. We reduced support on older GPUs across all OSes at the same time; the line was drawn across architectural boundaries and had nothing to do with availability of open source drivers.

What we said at the time was that we felt that the open source drivers were the best route for ongoing support (rather than occasionally updating the fglrx driver) since the GPUs in question were already "well supported except for 3D" and that ongoing progress in the open source stack would continue to improve things on the 3D front.

I find it hard to believe that as much as 40% of a graphic card performance is dependent on application-specific optimizations (i think that could be more true in the infamous days of the fx5xxx series).

Again, that is not what we said nor is it what BlackStar said.

The 30-40% estimate (made 2-1/2 yrs ago) for performance delta on challenging apps covered a combination of things - shader compiler and memory manager complexity & sophistication, scenario-specific code paths (eg using the hardware differently when memory pressure was high), application-specific optimizations, and the resources/priorities of the open source development community.

I should mention that the shader compiler in the 3xx-5xx 3D driver has evolved to be quite a bit more sophisticated than what we used in the model for those initial estimates.

That 'good enough' approach would mean that linux graphics will always stay inferior. Thats not good enough.

So "good enough" is not "good enough". I'd like to see a natural language parser deal with that :D

Seriously though, I don't understand your logic, unless you mean "Linux graphics on EOL'ed hardware will always stay inferior", which I don't agree with but at least I understand. Linux graphics will continue to be defined by proprietary drivers on newer hardware; the only issue being discussed here is whether older hardware on a "reduced support" schedule is best supported through limited effort on proprietary drivers or through open source drivers.

rohcQaH

11-12-2009, 10:36 AM

I find it hard to believe that as much as 40% of a graphic card performance is dependent on application-specific optimizations.
Just read some of the announcements for new windows drivers. "Now 40% faster in <recent game>". Guess what happened?

That 'good enough' approach would mean that linux graphics will always stay inferior. Thats not good enough.
Well, here are a few points for you to think about:

- the drivers are new. First priority of driver development is to get it to work. Second priority is to make it work fast. Right now it's (mostly) working and needs bug reports and polish to work on exotic boards and corner cases. It also needs to make it's way through release schedules to reach the end-user's installation DVD. AMD's devs should also be busy on adding evergreen support (at least I hope they are). For now, speed-optimizations aren't top priority. (but as usual, patches should be welcome)

- a lot of the percieved slowness is not due to the drivers, but due to the outdated gfx stack in linux. Work is already underway to fix that (namely the transition to KMS/DRI2 and later gallium3d), but it takes time, and obviously can't be used by AMD until it's done. And unless you're contributing to those projects or paying a company that does, you should just sit tight until it's done instead of complaining to AMD. The state of the linux gfx stack is not AMDs fault.
But at least AMD makes some drivers that use the OS stack now. Nobody would be interested in improving the OS stack if everyone stuck to their binary blobs. Customers using the OS stack are the biggest force to make distribution vendors contribute work to it. Don't expect instant miracles though.

- the OS drivers are already faster than fglrx in many cases, see the benchmarks that appeared here a while ago. For most applications, they're superior. 3D performance is lacking, see the last two points for reasons.

- AMD is very linux friendly right now, but their goals are long-term goals: to eventually have kickass linux drivers. To reach long-term goals, short-term sacrifices must be made. In this case, legacy fglrx updates were scrapped to push forward OS driver development. It's expected that some users aren't satisfied with that decision, but again, looking at the big picture it's what's best both for AMD and for linux. But I do understand it's not what's best for you right now.

Panix

11-13-2009, 12:03 PM

After struggling with the open source driver on my older laptop, I've decided when I upgrade my desktop video card to go with Nvidia.

Imho, the attitude is 'who cares about older gpu' and that's my opinion of what is going on. I'm confused with all the various pages of people having this setting and that setting so I'm shooting at ducks to see if one configuration combo will work. I won't say any more, too frustrating....

misiu_mp

11-15-2009, 09:40 PM

So "good enough" is not "good enough". I'd like to see a natural language parser deal with that :D

I ment that 'good enough' (notice the quotes), referring to someone else's notion of 'good' is not good enough for me.

Seriously though, I don't understand your logic, unless you mean "Linux graphics on EOL'ed hardware will always stay inferior", which I don't agree with but at least I understand. Linux graphics will continue to be defined by proprietary drivers on newer hardware;

You see here our philosophies differ. I really would rather see linux support all hardware with free drivers and no compromise (well, 90% of the optimal performance would be very good). Support for EOL hardware only is not good enough. Thats because binary blobs never did and most probably never will work very well in the linux environment. Relying on proprietary drivers makes linux incomplete. Optimally amd would take a major part in developing the free drivers (like intel does) or open up the fglrx. Im very grateful for amd's commitment as shown by the published documentation, but in my opinion everybody's effort should strive towards making the free drivers so good as to make the closed ones largely obsolete.

bridgman

11-15-2009, 10:36 PM

You see here our philosophies differ. I really would rather see linux support all hardware with free drivers and no compromise (well, 90% of the optimal performance would be very good). Support for EOL hardware only is not good enough.

In some markets 90% of optimal performance would mean we were essentially shut out of the market. The reason we have binary drivers is that they let us share code across 100% of the PC market which puts far more effective resources onto Linux drivers than market share could justify for a Linux-only driver.

Thats because binary blobs never did and most probably never will work very well in the linux environment. Relying on proprietary drivers makes linux incomplete.

That's why we support open source driver development as well.

Optimally amd would take a major part in developing the free drivers (like intel does) or open up the fglrx.

Relative to our size I believe we have as many developers working on the free drivers as Intel; possibly more. Is that not enough ?

Im very grateful for amd's commitment as shown by the published documentation, but in my opinion everybody's effort should strive towards making the free drivers so good as to make the closed ones largely obsolete.

In some markets I believe the free drivers will make the closed ones largely obsolete. In other markets binary drivers and cross-OS code sharing are still a hard requirement in order to meet feature and performance expectations. Moving all of the fglrx developers onto the open source drivers would not be enough to change that.

Remember that Linux means different things to different people. Some view Linux as the premiere open source solution where free drivers are an obvious necessity - others moved to Linux from proprietary Unix environments after Linux's free-as-in-beer availability undercut commercial Unix development and don't give a rat's a$$ about openness. The second group of users are unwilling to give up even a few percent of potential performance and features -- and for those users we still need binary drivers.

If you're in the first group - ignore the binary drivers. If you're in the second group, ignore the open source drivers :D

energyman

11-16-2009, 05:34 AM

bridgman, I thought you already learnt that no matter what you or AMD does someone will always complain ;)

Kano

11-16-2009, 05:58 AM

@bridgman

maybe you need a 3rd group: those who does want that everything works - no matter what driver is used. I don't think that the oss vs. binary discussion is really useful for new 3d hardware. There the features are more important. For gamers of course opengl speed, for others watching movies is more important - and when you want to watch hd movies then you can get into trouble with both drivers.

misiu_mp

11-16-2009, 09:58 AM

bridgman, I thought you already learnt that no matter what you or AMD does someone will always complain ;)

Simply pointing out that one should not rest on his laurels. There is tons of more to do.

misiu_mp

11-16-2009, 10:31 AM

That's why we support open source driver development as well.
Relative to our size I believe we have as many developers working on the free drivers as Intel; possibly more. Is that not enough ?

I wasn't aware of that. Sorry for my misjudgement. I can only see that intel provides high-performance open source drivers and amd does not. Of course the intel hardware is in a different segment, so its hard to compare the two. It will be interesting to see how things will develop when they will start competing with amd and nvidia in the high-performance segment.

In some markets I believe the free drivers will make the closed ones largely obsolete. In other markets binary drivers and cross-OS code sharing are still a hard requirement in order to meet feature and performance expectations.

If you're in the first group - ignore the binary drivers. If you're in the second group, ignore the open source drivers :D

You might want the proprietary drivers for 100% performance. I want the free drivers to be as good to make 90% of users (even gamers) not need them. Not many will care for say, 10% of performance when the drivers are in-kernel, work out-of-the-box, well supported, well featured and you can count on major mishaps to be fixed (more) quickly.

If that is the *vision* amd has for its linux graphics support, then you deserve all the prising in the world.

If the *aim* is to only give open source support of old hardware, then well, as much I am grateful for the commitment of amd (its way better than most others), it is at best 'good enough'.

energyman

11-16-2009, 10:38 AM

'high performance drivers'? Where? Have I slept? I remember a nice Phoronix article lambasting the sorry state of current Intel drivers.

bridgman

11-16-2009, 10:52 AM

If that is the *vision* amd has for its linux graphics support, then you deserve all the prising in the world.

That is the vision. It has not been a real popular vision during implementation (everyone feels cheated when we spend time on something a different user cares about ;)), but I think it will prove to be the correct one in the end.

If the *aim* is to only give open source support of old hardware, then well, as much I am grateful for the commitment of amd (its way better than most others), it is at best 'good enough'.

Nope, we're aiming for consistent support across both old and new :

http://www.phoronix.com/scan.php?page=article&item=fedora_r600_3d&num=1

@bridgman

maybe you need a 3rd group: those who does want that everything works - no matter what driver is used. I don't think that the oss vs. binary discussion is really useful for new 3d hardware. There the features are more important. For gamers of course opengl speed, for others watching movies is more important - and when you want to watch hd movies then you can get into trouble with both drivers.

I think we have to satisfy that group with both drivers.