Driver Update Addresses Radeon Frame Latency Issues 108
crookedvulture writes "AMD has begun addressing the Radeon frame latency spikes covered previously on Slashdot. A new beta driver is due out next week, and it dramatically smooths the uneven frame times exhibited by certain Radeon graphics processors. The driver only tackles performance issues in a few games, but more fixes are on the way. In the games that have been addressed, the new driver delivers more consistent frame times and smoother gameplay without having much of an impact on the minimum or average FPS numbers. Those traditional FPS metrics clearly do a poor job of quantifying the fluidity of in-game action. Surprisingly, it seems AMD was largely relying on those metrics when testing drivers internally. The company has now pledged to pay more attention to frame latencies to ensure that these kinds of issues don't crop up again."
Open Source drivers? (Score:2, Insightful)
When will Nvidia and ATI release proper open source drivers instead of us having to install a binary blob to get our hardware working? That would really help if there were drivers that could ship in the kernel to handle ATI hardware instead of the closed source options.
Re: (Score:3)
When will Nvidia and ATI release proper open source drivers instead of us having to install a binary blob to get our hardware working?
Does Sus Volans answer your question?
Show them numbers on how this would benefit their bottom line without risk of wasting work for no return, and I'm sure they'll listen.
Re:Open Source drivers? (Score:5, Informative)
It's not just bottom line. At this point and time, they would financially benefit from open sourcing their drivers. The problem really lies in a huge tangled web of license agreements from other parties that they can't straighten out. For almost any commercial product to go open source while the company is still in operation is a legal nightmare, and that's why it almost never happens.
Re: (Score:2)
... and yet it happened with Java. It happened with StarOffice.
It is possible, even if it's just the parts that can be and not the parts that can't be (like StarOffice's DB was)
Stop making excuses for a can't do attitude.
Re: (Score:3)
... and yet it happened with Java. It happened with StarOffice.
I don't wish to trivialize StarOffice's database integration, but you're talking about ripping out one library and stuffing in another, it's not as hard as rewriting everything and if it is then the original code was shit. And Java was even less of an issue in some ways, because it was designed to be maintained.
Re:Open Source drivers? (Score:5, Insightful)
When you get all the patent holders for the patents they are violating to execute license agreements and hold harmless agreements. You think it's only software where patents are stifling things? Everyone in the industry kind of willfully looks the other way, as long as the other guy also willfully looks away, and as long as there's no source code for anyone to drag anyone else into court over.
Realize also that the interfaces used on the bottom end to interface the software for the hardware disclose substantial information about the hardware as well. Imagine the following question in the press: "if you say you support 'B' in hardware, but are actually doing 'Q' and 'R' in hardware when you are asked to accomplish 'B', isn't the claim that you have 'hardware accelerated B' only technically true in order to have that marketing checkbox checked?". There are similar uncomfortable questions.
Apart from those, the interfaces to the hardware can disclose additional hardware patent violations, which would normally be covered by the "willfully looks away" already in progress.
if you actually did come up with something clever, but which wasn't patentable for whatever reason, your competitor could just copy it, and then you would have lost your market advantage.
Finally, most hardware codec decoding, e.g. for H.264, is partially looped in software s that the license can be tied to the software instead of the hardware, and therefore be optional, and not add to the unit cost as a hardware royalty item to Sorenson (and others). By this fiction, they become an optional software royalty item where the company using the hardware in their design can choose wheter or not to use the capability, and thereby be required to pay the royalty. If it became easy to use the hardware capabilities from Open Source software, then this fiction disappears. You can argue that standards should all be royalty free or not be standards until you are blue in the face, but you are looking at approximately 100,000,000,000 DVDs total in the world, all expecting H.264 to decode them, and that requires a royalty payment.
No, these drivers are never going to be fully Open Source at the same time they give access to all the hardware capabilities.
Re: (Score:1)
DVDs are MP4 not H.264. You are thinking of BD (and the failed HD-DVD)
Re: (Score:3)
The argument still holds for non-BluRay DVDs: MPLA - http://en.wikipedia.org/wiki/MPEG_Licensing_Authority [wikipedia.org] - counts as "others" from which patent licenses must be obtained to enable the software loop-out.
Re:Open Source drivers? (Score:5, Informative)
DVD"s are MPEG-2, not MPEG-4.
Re: (Score:2)
Realize also that the interfaces used on the bottom end to interface the software for the hardware disclose substantial information about the hardware as well. Imagine the following question in the press: "if you say you support 'B' in hardware, but are actually doing 'Q' and 'R' in hardware when you are asked to accomplish 'B', isn't the claim that you have 'hardware accelerated B' only technically true in order to have that marketing checkbox checked?". There are similar uncomfortable questions.
So stop fibbing! Either say "Hardware assisted B" or say "Q and R in hardware" say both if you like.
Apart from those, the interfaces to the hardware can disclose additional hardware patent violations, which would normally be covered by the "willfully looks away" already in progress.
So can an electron microscope and/or a bus sniffer, which is something anyone who might hold a relevant patent will likely have and already use to examine your hardware and driver.
Finally, most hardware codec decoding, e.g. for H.264, is partially looped in software s that the license can be tied to the software instead of the hardware, and therefore be optional
Put the lot in hardware, and for the 'lesser' chip that "doesn't have it", blow a fuse to disable it.
Re: (Score:3)
No, these drivers are never going to be fully Open Source at the same time they give access to all the hardware capabilities.
I think "never" is a bit too strong of a word, maybe if you're talking for the current cards, perhaps, but Intel has seemed to be able to provide open source Linux graphics drivers. [01.org]
True, true, they don't make the same cards as the high end guys do, but it's closed minded to even think that disparate computing model w/ discrete GPUs is going to last forever.
IMO, I'd rather do everything in software rasterizer -- Pixel & Vertex shaders still don't give me the same sort of control I had doing everythi
Re: (Score:2)
When will Nvidia and ATI release proper open source drivers instead of us having to install a binary blob to get our hardware working? That would really help if there were drivers that could ship in the kernel to handle ATI hardware instead of the closed source options.
Get off your butt and pitch in on the R600 target for LLVM/Clang and later the Mesa 10.x roadmap so that OpenGL 4.x and OpenCL 1.2 quality is available. After all, it's already happening. Worry about general computing first. Those improvements will role into the Catalyst driver.
Kudos to AMD for owning up to this (Score:5, Insightful)
Probably one of the most important divides in engineers (the world?) is the ability to read the data, acknowledge your mistakes and fix it. It seems like most companies spend more time doing damage control than damage remediation. Reality must take precedence over public relations, for Mother Slashdot cannot be fooled (with apologies to Richard Feynman).
Re: (Score:1)
Re: (Score:2, Interesting)
They need a new standard: WFPS. Worst Frames Per Second. Find the slowest frame, and how many times that frame can be calculated in one second. Average is a bad way to judge things; perhaps part of the benchmark includes the view pointed at the sky with only 10 polygons, where it would obviously render at incredible speeds, driving the average FPS up. WFPS would give you a much better idea of what the system is actually capable of when under a full load.
Re: (Score:1)
Why not give us average, median, variance, skew... or paint a probability distribution onto the box ?
Re:Kudos to AMD for owning up to this (Score:5, Insightful)
The Frame latencies by percentile [techreport.com] graph they create now is the right way to look at this data. It's a sort of probability distribution function for slow frames. Nothing simpler will capture the complexity of the problem. You can't usefully boil the universe of rendering latency issues into any single number.
The worst frame will vary based on card and game, and the tools available to reviewers are not practical to find them. And what this debacle has shown is that even though they're limited, the tools being used by reviewers are sometimes better than internal QA at the manufacturers.
Re: (Score:3, Insightful)
You're assuming it's about ego. More likely they still have code that was produced by a contractor where they don't have the rights to redistribute or where they have a license which is only for their use, not for 3rd parties. Anybody who cares about it is likely to be more interested in the code being released, the people that wrote it likely don't get a say in it. Plus, I doubt most people really care one way or the other and are unlikely to even look. Assuming they can even program.
AMD has been releasing
Regional question (Score:2)
This is going to come off sounding kind of trollish/regionalist/xenophobic, but I'm curious if the location of silicon valley has a lot to do with this. Different areas and regions have different cultures, and in my line of work I have to deal with a lot of people in California. The immediate reaction to any problem is finger pointing, not looking for a solution, which is a stark contrast to where I live where mistakes are OK as long as they are owned and rectified. Everyone who transfers here from Calif
as a Radeon 6870 owner (Score:1)
Re:as a Radeon 6870 owner (Score:5, Informative)
This only applies to the new GCN architecture (most high end 7000 series cards) and not the older VLIW4/5 cards (which includes low end 7000 series cards and 6000 series cards)
Re: (Score:1)
Re: as a Radeon 6870 owner (Score:3, Informative)
Since when is disk I/O speed or access time relevant to frame rates in a game?
Disk I/O determines how fast a game loads, or how short the wait between levels in a game etc. Frame rates (or latencies) are determined by the trio CPU + memory bandwidth + GPU. Sure some games may load data while you're in the midst of the action, but in that case likely bite-size chunks that shouldn't affect frame rates significantly (and unrelated to the issue discussed here).
Re: (Score:3)
If your game didnt suck balls - it was already preloaded into RAM, which was loading textures and stuff into VRAM.
You are presuming that all the data for the gaming environment will fit in memory at the same time. These days this is quite often not the case. A game like GTA4 doesnt have any loading screens but its a large world with many gigabytes of texture data.
...but the person you are replying to doesnt make much sense for sure.. these frame latency issues have to do with the memory management of the video card, where the variance of the time per frame for condensing free blocks and garbage collection was quite
Re: (Score:1)
What the hell. Games never locextures to the gpu directly from the hard drive. That's what your ram does. Most games would only require say 100mb of ram tops if the textures came direct from your disk.
If you have 12gig of ram THEN YOU ARE NOT SWAPPING. No swapping means you could disable virtual memory completely.
And what is this mystery I/O chip you are speaking of?
The north bridge, and the south bridge or something else magical?
If either the North or south bridges fail or over heat you've got a instant cr
Re: (Score:2)
Re: as a Radeon 6870 owner (Score:5, Informative)
Since when is disk I/O speed or access time relevant to frame rates in a game?
When the engine is trying to grab data in real time from disk.
This is most assuredly *not* every game. But it is some games, or games in some scenarios. In MMO's you don't have enough RAM memory of all of the possible character armours these days, so you have to dynamically grab only that which will be on screen, same with any zone streaming in data from whatever area you have around you.
I can see why people would think this is a HDD speed issue. If you have burst loads of up to say 200 MB/s on a HDD, but average around 20, well then a regular drive will hiccup periodically whereas the SSD won't even bat an eye. As you say, that isn't actually *this* issue.
Re: (Score:2)
It does if you try and wait for whole model parts to load at a time between frames. Not everyones multithreaded rendering is as wonderful as we'd like.
Re: (Score:2)
Diablo 3 is like that. Putting the key files through mlink on a SSD disk or even USB key will bring incredible advantage.
Re: (Score:3)
Shitty console ports love to stream textures
Re: (Score:2)
Re: (Score:1)
Re: (Score:2)
Toms hardware answer this.
Re: (Score:1)
The HD 7870 (now almost all brands are OC'd) is a good alternative to Nvidia which seems to crap its pants, especially in SLI mode. But wait till the new processors are out and pay for what you can afford.
Re: (Score:1)
Go with a cheap 7770, or 650 Ti, up to a 7850, 7870, 7870 Le, or 660?
This lineup suggests a lack of understanding.
... 660 runs 140 watts and $215,
... 7770 runs at 80 watts and $115,
... 7850 runs 130 watts and $165,
The 650ti runs 110 watts and $140,
and 7870 is 175 watts and $210.
The only possible choice that isn't simply down to 'wattage requirements' or 'spot price' is between the 660 and the 7870 and my guess is that the 7870 will win most FPS benchmarks but its going to cost you more to operate over the long run (35W adds up over a year or two) and will probably b
Re: (Score:2)
Only if you are running something like furmark 24/7 for the whole year. Most games will not even remotely stress the card so hard. Your card will mostly be in idle most of the time and in the case of new AMD cards they can even go to deep sleep when not in use and only burn 3W (useful for 24/7 machines such a s mine).
Re: (Score:2)
We are talking about the 660 being $5 more than the 7870, and you seem to be arguing that the 7870 wont ever use $5 more electricity over the lifetime of the card which by my calculations is about 41kWh (at $0.12/kWh.) If you plan on owning the card for precisely 2 years (a reasonable lifecycle for a gamers vide
Re: (Score:1)
Beta drivers improve 57xx too (Score:3, Interesting)
In addition the security hole with the aslr being disabled was also fixed last stable release 12.10.
This year ATI also stopped releasing a driver every month and instead focused on QA before certifying drivers.
ATI really is improving as they try to stay alive. Bravo indeed and my next card will be an ATI.
Re: (Score:2)
ATI really is improving as they try to stay alive. Bravo indeed and my next card will be an ATI.
My advice is to wait a bit longer than that. Every time there is news that ATI has substantially improved their drivers, I try an ATI card. Every time I try an ATI card, it makes me want to hurt someone, preferably someone at ATI. I propose that you wait until other people have bought the next generation of cards, and you have time to figure out of ATI has learned to code their way out of a nutsack.
Given that ATI is owned by the CPU company that fired their CPU designers, I am skeptical that things can "con
Re: (Score:1)
Been running ATI for years and never had any of the problems you exhibit. Maybe you should play games that aren't complete shit.
Maybe you should log in if you're not just a shill for ATI, and you have the courage of your convictions. When you say things like this without logging in, you only grant the other side in the argument credibility. Or perhaps that was your goal, in which case piss off, I don't need the help.
Re: (Score:2)
I don't really have issues either, but the last fps that I played much was quake 3.
Re: (Score:2)
I don't really have issues either, but the last fps that I played much was quake 3.
Every third card or so, I think "surely ATI has got their shit together by now" and buy one. And I'm always wrong. I run the occasional new title, mostly old titles, and lots of FoSS so I do a little of everything, but I've had tons of problems with ATI-chipset cards which really were resolved by simply switching to an nVidia card. It's got to the point where I won't even buy a motherboard with ATI graphics onboard because I don't want to ever be tempted into trying to use it, and I don't want to pay for th
Re: (Score:2)
by Anonymous Coward on 01-18-13 8:05 (#42625979)
I am logged in,
Oh, so you just know you're full of shit, so you won't spend karma to back up what you say? Either way, you're full of shit if you won't expose a username.
and if you think someone from ATI couldn't make a fake account then I don't know what rock you've been living under.
You're exceptionally stupid if you don't think that it matters if someone logs in. But you do know; you're a shill.
Never had significant problems with ATI either (Score:2)
And I get money and free heating out of my ATI re: BitCoin mining.
Currently using a 6750 on Windows 7.
To accuse someone of being a shill simply because they post as AC (as if /. IDs are somehow validated) smacks of someone who's losing it...
Now fix Windows 8 (Score:2, Funny)
Re: (Score:1)
Now they need to fix the Catalyst driver to stop crashing on Windows 8. This is getting annoying, especially the BDOD that pop up every other day.
Have an even better idea: ditch Windows 8...
Re:Now fix Windows 8 (Score:4)
Re: (Score:2)
Re: (Score:2)
Im not getting any crashes on my win8/AMD machine. Have you tried upgrading your motherboard BIOS? There were some problems with old BIOS versions and the 7xxx series when going to sleep mode.
This has been a problem on all cards (Score:2, Interesting)
And I mean ALL ATi cards since the early Radeon 7000's. And the problems aren't just isolated to Windows. You can't get a decent read on the vertical blank timings because their cards are simply shitty and will randomly have frame drops and latency. nVidia seems to have somewhat followed them in their path lately but they were pretty good before.
This is generally not a problem for gamers to lose 1 frame or have a couple of them a bit later (you can't notice it) but when you're doing psychophysics experiment
Really? (Score:2)
Really? These people who want to put a binary into your computer, as a fully privileged user, interacting with every part of your system in order to improve graphical performance - with full access to all design documents, chipset manuals, and source code - couldn't work out that FPS could potentially be "spiky" and that just a single figure wasn't an accurate representation of how a human perceived their extraordinarily complex and expensive graphics, and that's why things didn't look or work smoothly?
And
Re:Really? (Score:5, Insightful)
And it took them YEARS to work this out? And only really weeks to "fix"?
Just because it is obvious to you doesn't mean that it is obvious to others. Really.
If there's a problem, kick up a fuss, complain, let someone know who can do something about it. This is true whether it is software, hardware, real life services, etc. There's always plenty for the people doing support to do, so if you want YOUR issues to be the ones fixed then you'd better sing up about them so that they get some priority. If you say nothing, everyone else will assume you're doing fine with no problems at all. That's the way the world works.
Re: (Score:3)
Just because it is obvious to you doesn't mean that it is obvious to others. Really.
It's their job to get this right. ATI's only job, in fact, was to make GPUs and produce drivers which make them work properly. They made the hardware, but they still couldn't make the driver for their own hardware, and have been doing it wrong for years. That means they're incompetent. That's the way the world works.
Re: (Score:2)
No, they built and tested their drivers internally to match the expectations of the public. The public was testing min/avg/max FPS, so that's what ATI built. In other words, shitty benchmarks result in shitty drivers.
And to call ATI incompetent because they didn't notice an issue almost nobody else did is quite arrogant. Especially since once the issue was discovered, they promptly fixed it. That's not even remotely incompetent, that's fantastic support and I'd love it if most companies were that responsive
Re: (Score:2)
And to call ATI incompetent because they didn't notice an issue almost nobody else did is quite arrogant.
Gross misrepresentation of the facts is a gross misrepresentation. They had a problem nobody else did, and didn't notice it, and that's why they're incompetent. Especially when people have been reporting this problem to them for years.
Oh, and the part where you claim ATI can't make a driver or that they are doing it wrong is flat out not true, not even close. Kindly go fuck yourself
Are you the butt-hurt fanboi who rounded up a posse to go mod down five of my comments? I've been experiencing windows crashes due to shitty ATI drivers since Windows fucking 3.1, so while I may go fuck myself later, it has nothing to do with you or ATI. Your internet bravery
Re: (Score:1)
Gross misrepresentation of the facts is a gross misrepresentation. They had a problem nobody else did, and didn't notice it, and that's why they're incompetent. Especially when people have been reporting this problem to them for years.
By "nobody else" I assume you mean Nvidia, as that's the only other driver that was tested.
And by Nvidia's own admission, they only noticed and started optimizing this about 2 years ago. Interestingly that's right around the time they started working on mobile platforms, which measure in time-per-frame and not framerate.
Are you the butt-hurt fanboi who rounded up a posse to go mod down five of my comments? I've been experiencing windows crashes due to shitty ATI drivers since Windows fucking 3.1, so while I may go fuck myself later, it has nothing to do with you or ATI. Your internet bravery is pathetic.
I didn't round up anyone, and your crashes aren't coming from ATI's drivers. But by all means, switch to Nvidia and experience the joys of their driver issues.
But in the interest of full dis
Re: (Score:2)
your crashes aren't coming from ATI's drivers
Of course not. That's because I'm not using ATI hardware, because I learned my lesson.
and other games? (Score:2)
So am I right in understanding the resolution is that DX9 games may be fixed by if they get around to addressing that game (dx9 requires game-by-game fixes), whilst all DX10 & DX11 games should be fixed by a forthcoming (hopefully...) general driver fix?
Linux too? (Score:1)
I has nothing to do with the architecture (Score:1)
Re: (Score:1)
Why don't you read what he actually asked before shooting your mouth off. This is the 4th(!) time the AMD frame rate issues story has been posted on /. and the second promising a fix in future drivers... the op didn't wonder where to get drivers, but how many more times is this same story going to be posted before an actual fix is available.
Re: (Score:2)
Oh no, somebody please help this lost little boy who can't figure out that AMD hosts their drivers on their website!!
No, they don't. They host some of their drivers on their website. nVidia hosts their drivers*, Intel hosts their drivers, but AMD only hosts their desktop drivers.
Get back to me when AMD hosts all their own drivers, so that you don't have to go a notebook manufacturer's website and get an outdated driver for your graphics card.
* nVidia used to not host mobile drivers, but has now for years.
Re: (Score:2)
Poor you.
Re: (Score:2)
Drivers for my notebook are on the AMD site.
But, just like AMD doesn't care about me, I don't care about you.
You don't care if I care about you, and AMD doesn't care if I care about them either, obviously, because they don't provide a download for my platform. Consequently, I don't care about AMD. I have never recommended their graphics cards, and now I can't recommend their processors. But I also have to take anyone who thinks they're great less seriously because they forgive not only their gross mistakes but also intentional abuse of their customer
Re: (Score:1)
RTFA. Next week.
This story is about "problem solved, beta works, driver coming up". It was not clear whether the problem was in the hardware or only driver related. So I assume it will interest people (even nvidia-users as myself) that the problem is already solved.
First story was "we have an issue here. Just look how much smoother nvidia is. We describe the problem technically and show it to you."
To me, it is news: Look, ATI tries to do good drivers and addresses problems rather then denying them and it di
Re: (Score:2)
Re: (Score:1)