AMD's Radeon HD 5870: Before And After A Year Of Driver Updates

[cnick6]

There's even more performance gains by using the 11.4 beta driver from the Radeon 6990 download page. Using the DC benchmark, I saw nearly 10% improvement in all areas except the CPU mark which went up only a point.

 

[treefrog07]

I've had my 5970 for six months - it probably came with the 10.x drivers. After reading this article, I think I'll update to the newest drivers.

 

[JackNaylorPE]

[citation][nom]banthracis[/nom]This article attempts to evaluate the effect of adjusting a single condition, drivers, on overall performance. Adding patches into the mix introduces an additional variable which is:1. Unnecessary, since it has nothing to do with how drivers improve performance. 2. Makes it difficult to determine how much performance gain/loss is due to drivers or due to patches. This would put into question any conclusion gained from the experiment (basic scientific method). 3. Exponentially increases the number of test's required, making it unfeasible to carry out.[/citation]

One problem. Card vendor is tweaking his driver to the patched game, not the original one.

 

[JackNaylorPE]

Questions (applies to both red and green camps):

1. Are the increases in fps due to an actual increase in driver efficiency for overall performance .... or is the fps increase the result of optimizations or the specific game sequence used in the benchmark ? Much like tuning a carburator .... optimizing fuel / oil mix for one set of elevation, temperature and humidity condition has a negative effect when any of these three parameters changes.

2. How are these gains materialized ? Driver efficiency or are we gaining better fps by dropping some pixels here and there ?

 

[tntom]

Great to see the driver improvements. Now AMD please get Age of Empires 2 working again with the new drivers. I have to disable Direct Draw every time.

 

[wild9]

Stability is more important to me than the small performance increases or features that only appeal to specific markets. But I accept that both nVidia and ATI have to be at the forefront of these markets if they are to remain competitive and raise sufficient revenue to support their existing user base.

When dealing with 3D games I usually just over-clock the hardware to get more performance, or just replace it altogether. I'd I update my own drivers (Windows Certified) 4 months months, opting for minimal installation option.

Historically I ran mostly onboard GPU's: first nVidia (Geforce 6000/7000 series), then AMD (700 series). No major complaints from either camp with regards to price, performance and stability. However hardware nVidia's naming conventions seem a bit confusing at times.

 

[weaselsmasher]

Two takeaways from this:

1. It's great that there is improvement in performance, essentially for free (though sometimes there also is harm to performance with a driver update - it's a mixed bag sometimes).
2. ATI ships fast hardware with crappy drivers (hey, if they were good drivers to begin with, the performance at the start of the lifecycle wouldn't have much room for improvement). As soon as the silicon is cool off the foundry, sell it regardless of driver state, and fix performance problems later.

I am not saying that the situation is necessarily any different for nVidia, or even that it's the wrong thing to do... the market has clearly shown that people will buy the latest fastest hardware regardless of driver maturity, and accept the "fix it in the drivers afterward" model. "It will ship when the drivers are ready" is clearly not the way to pay the bills if you are a PC video card manufacturer. It's just an interesting view into the mindset of that particular business.

I'd love to see a similar article for, say, a GTX 470.

 

[dertechie]

This'll be even more interesting with the 69X0 in a year and a half. HD5K at its core is a scaled up HD4K which is scaled up HD3K, so AMD already knew most of the tricks for squeezing every drop out of that arch, they just had to learn how to do it for DX11. VLIW4 is a new architecture, it'll be interesting to see them learn how to squeeze that for all its worth.

 

[booseek]

[citation][nom]slyphnier[/nom]ever since 10.11 ... many user got "100% GPU usage" and "top right mouse lag" issue11.4 early preview still having the issuefrom what i know the issue above mostly effect radeon 5000series (5850 and 5970)for people having those issue forced to stick with 10.10estill no news when amd will fix those issue, as from what i heard amd can't replicate the issue on their lab, thus can't fix it[/citation]

Let's not make assumptions about one another.

I also have had a 5870 in a new rig since May 2010 and every single driver after 10.4 gave me problems until the January 2011 release. And that was after uninstalling old drivers, driver sweeping them, etc. It was everything from BSODs, to jet-like fan speeds at system startup, crashing and recovering display drivers, and even Aero crashing. And StarCraft II cursor corruption, an ATI-only problem. All BSODs pointed at the display driver and everything else was a graphical glitch.

But nonetheless, I love my 5870 and would buy ATI again. But coming from a 7800 GT from 2005-2010, I did have many more driver-related problems.

 

[booseek]

^^^I meant to quote another post:

If that guy who created/posted his problem 5 times spent that much effort understanding why his rig BSOD'd when he used the latest set, he might have resolved his problem by now.

 

[Guest]

After being used to the "always update to the latest driver" mentality it has been tough breaking old habits. But AMD has taught me if it ain't broke don't fix it. In other words if your current drivers aren't crashing your games leave it alone, because chances are probably greater than 50% that the new drivers will. I'd rather have stability than an extra 5% any day of the week.

 

[iam2thecrowe]

[citation][nom]JBomb71[/nom]After being used to the "always update to the latest driver" mentality it has been tough breaking old habits. But AMD has taught me if it ain't broke don't fix it. In other words if your current drivers aren't crashing your games leave it alone, because chances are probably greater than 50% that the new drivers will. I'd rather have stability than an extra 5% any day of the week.[/citation]Amen to that, if there is no problem then don't update.

 

[user 18]

[citation][nom]tntom[/nom]Great to see the driver improvements. Now AMD please get Age of Empires 2 working again with the new drivers. I have to disable Direct Draw every time.[/citation]

Yes please. Love that game.

 

[user 18]

[citation][nom]JBomb71[/nom]After being used to the "always update to the latest driver" mentality it has been tough breaking old habits. But AMD has taught me if it ain't broke don't fix it. In other words if your current drivers aren't crashing your games leave it alone, because chances are probably greater than 50% that the new drivers will. I'd rather have stability than an extra 5% any day of the week.[/citation]

Amen to that.

 

[howiejcee]

[citation][nom]shrapnel_indie[/nom]Funny, no probs with my 5870 at all with any update... even with 11.2I suggest you go take a good long look at your system configuration.[/citation]

Funny, try playing Starcraft 2 which is a very popular game and you'll notice cursor errors with majority, if not all, drivers since 10.9a.

Also funny, the more recent drivers since 10.11 also result in more grey/brown/striped screens when playing games.

Even more funny how 10.9a is actually decent indicating truly crappy other driver releases, and also funny how my PC is running perfectly fine otherwise and when I swap out to my previous, old 8800GT.

FYI, my Q6600 has been running smoothly @ 3.4GHz which is a nice moderate OC and Win7 x64 has also been great w/ 6GB ram. Pls don't make assumptions that you cannot back up.

 

[icecreampop3]

on the test hardware page, how did Tom's manage to procure a 2500k with 8mb L3?

 

[hangfirew8]

[citation][nom]haplo602[/nom] the issue with the AMD driver schedule is that they alternate development teams. basicaly fixes from one driver release don't have to be included in the next, since the development started before the current one was finished. it's mostly visible on Linux drivers. One release fixes an issue and it's back in he next one.[/citation]

How does that explain 1 out of 3 fglrx/Radeon releases breaking KDE throughout 2010? Does AMD have 3 development teams? Or maybe they just suck at Linux?

 

[hangfirew8]

Sorry, I meant 2 out of 3 releases broke KDE. Though I'm not sure how to count 10.7, since it was primarily labeled 10.6 inside and out. It's like AMD was not sure if 10.7 was really 10.6 or not.

 

[philiprudd]

Thanks for testing this guys.I always do the same trough the years with my cards and find most drivers increases performance or fixes some issues I had ingame with alot of games.

I agree with you Elbert many people do know of the impact patches have on games the thing is there much bigger performance increases from patch to patch for games and many people do see this,a patch fixes issues in the game where you had slow downs because of unpolished places ingame the next patch fixes that and you have an big performance boost over what it was and the drivers of cards do increase performance but that is what its origanaly made to do so i see if you test on a scale like these guys do, they are testing the games with all the newest patches and they are re-testing all games with these old and new drivers so we can clearly see the difference between them.

 

[DedTV]

On the Warcraft benchmarks, you can avoid using Fraps by using the /timetest command before taking the flight. At the end of the flight it displays the min, max and average FPS in your flight log. It also causes the flight to be run at a specific in-game time (ie. the sky will always be the same, removing that potential variance. It's a daytime scene so if you use /timetest when it's night in the world, it will be daytime during your flight).
It used to also halt the rendering of other players during the benchmark but it doesn't appear to do so anymore as flying away from Orgrimmar I often saw a couple people coming in to land as I flew away. But it's possible that the benchmark waits a few seconds before halting rendering as I never saw anyone beyond the first few seconds of a flight. I haven't tested it to be sure or researched whether Blizzard has answered that already. But, you can also avoid the problem of having to deal with different characters appearing on screen between tests by using out of the way flight points where you are less likely to encounter anyone else.

 

[Guest]

This is why NVIDIA drivers are better. =\ I've never seen such incompatibilities with games as I have with ATI and AMD graphics card.. Ever since my first 9850 and my x1300 I've had nothing but problems when using ATI and AMD drivers. =( My last card was a 5870, ended up trading it in for a GTX 480 have not had a single problem. Wish I had switched sooner. =( Even used to be all wtf when my friends 8800GT would run flawlessly on games I was supposed to be getting similar performance on. I always got crash to desktops and other bs..

 

[youssef 2010]

I haven't seen the error in F1 since the days of NFS Underground 2 on my old Geforce MX 4400

 

[wanderlustx2]

[citation][nom]howiejcee[/nom]Andrew Ku forgot to mention that ATI's lousy drivers have and for the foreseeable future will always have problems.Not to say that Nvidia occasionally has driver issues, but at least they fix the problems. Only 10.9 and a couple versions before that don't give my 5870 any major problems.[/citation]

I've had 3870CF, 4870 and now a 6950 updated to a 6970. I've had a few small issues, much like I did when I ran my last nVidia card but nothing big. Do you own one or just trolling for nVidia? Sounds like the latter.

 

[mitch074]

Personally, I also thought driver updates at AMD's were a matter of varying quality: my trusty 4850 used to bomb on me if I weren't using a very precise driver, installed using very precise steps and settings (please note I run it on Linux) with a specific kernel revision.

Until I decided to replace my aging Athlon64 X2 3800+ (on s939): DDR being costly at that time and my Nvidia-based motherboard showing signs of weakness (it couldn't hold a CPU overclock any more), I decided to bite the bullet and got a new PSU, motherboard, CPU (Athlon II X4 620) and DDR3 RAM - and plugged the 4850 on this setup.

And instantly, I could run any kernel version with any driver revision I wanted, installed whatever way I preferred (and I must admit, it's much easier building RPM packages then install these than manually build and install each component) and still get VERY stable results.

What do I get from this?

- when running hardware, for maximum stability, try to use same-generation GPU and chipset;

- before you go and blame lousy drivers, look at the rule above.

Before that, the Nvidia motherboard used to drive a Geforce 6600 card; it was rock solid - except with some driver releases.

I must say, AMD made tremendous progress ever since 2008 in driver quality.

 

[Guest]

I'd like to thank Tom's Hardware for this article also. Very interesting, because I do make an effort to update my drivers - sometimes they are better, and sometimes I have been sorry I upgraded at all - especially ATIs TV Tuner software. Works fine with the original, doesn't work at all with an update. My son's computer played Age of Mythology very well with older NVidia drivers, but doesn't work well at all with new ones, so they both have their issues..