Tuesday, February 5th 2013
AMD Working on a Real GPU Dynamic Overclocking Technology
While digging through documentation for the latest version of AMD Display Library (ADL), we discovered evidence that AMD is working on a real GPU dynamic overclocking technology akin to NVIDIA's GPU Boost. Such a technology could manipulate GPU (and possibly memory) clock speeds, and voltages across multiple power states, taking into account processing load and temperatures. ADL allows third-party applications low-level interactions with AMD display drivers. Current generation Radeon graphics cards use Overdrive 5 and the feature-set it comes with, and so the new technology, along with Overdrive 6 could feature on upcoming generations of AMD GPUs.
Listed under Overdrive 6 capabilities, AMD documented three new definitions, one which indicates that a GPU's core/engine clock can be changed within a range (ADL_OD6_CAPABILITY_SCLK_CUSTOMIZATION), one that its memory clock can be changed within range (ADL_OD6_CAPABILITY_MCLK_CUSTOMIZATION), and one that monitors its activity/load (ADL_OD6_CAPABILITY_GPU_ACTIVITY_MONITOR). The three are ingredients of a dynamic OC technology in the works.
With its Radeon HD 7970 GHz Edition and Radeon HD 7950 v2, AMD rushed out a feature it calls PowerTune with Boost. This technology follows an archaic method of granting all applications maximum GPU clock (or boost state), while scaling down to nominal (advertised) clock speeds in the event the GPU is overloaded, so most applications run boost state at all time, very few apps actually slow the GPU down. Effectively, this means AMD is setting the boost clock speed too low on current cards, probably being held back by its impact on power draw.
The technology we believe AMD to be experimenting with could follow a method more similar to NVIDIA's GPU Boost. There are no definite clock speeds, but ranges, and the boost clock algorithm instructs the driver to alter clock speeds taking into account load, power draw, and temperatures. Close reading of the API documents also reveal that AMD is working on a dynamic boosting technology for the memory clocks, something that's currently not possible on NVIDIA's products. With AMD GPUs, applying a memory overclock no longer cause the screen to flicker (at least not noticeably).
Listed under Overdrive 6 capabilities, AMD documented three new definitions, one which indicates that a GPU's core/engine clock can be changed within a range (ADL_OD6_CAPABILITY_SCLK_CUSTOMIZATION), one that its memory clock can be changed within range (ADL_OD6_CAPABILITY_MCLK_CUSTOMIZATION), and one that monitors its activity/load (ADL_OD6_CAPABILITY_GPU_ACTIVITY_MONITOR). The three are ingredients of a dynamic OC technology in the works.
The technology we believe AMD to be experimenting with could follow a method more similar to NVIDIA's GPU Boost. There are no definite clock speeds, but ranges, and the boost clock algorithm instructs the driver to alter clock speeds taking into account load, power draw, and temperatures. Close reading of the API documents also reveal that AMD is working on a dynamic boosting technology for the memory clocks, something that's currently not possible on NVIDIA's products. With AMD GPUs, applying a memory overclock no longer cause the screen to flicker (at least not noticeably).
43 Comments on AMD Working on a Real GPU Dynamic Overclocking Technology
Or atleast make boost optional..
Manuel OC ftw!
I dont see why they cant make it like the current stock of CPU's with turbo when it comes to overclocking. That way, if you want to fiddle with clocks and such, you can just disable the turbo and clock it any way you see fit.
Whats also great is that apparently they will include the memory clocks in their dynamic scheme. I hope while they're on it they'll also fix the memory clock requirements for things like the UVD so any overclocks you make wont disrupt the working of UVD.
For an instance, take furmark, and a random other game that can stress your GPU to 100% load; Eventhough they can both load up your GPU"to "full", Furmark uses more power and produces more heat. Clocking a card so it wont overheat in furmark will mean clocking it lower then is possible for other games, resulting in less FPS.
If this dynamic clocking works correctly, it would mean your GPU can clock itself up, for games that dont use all the workunits, or down when the GPU would otherwise overheat.
Essentially it would mean you get the maximum performance your card can give you at more times, and would make the card less dependant on wether games are coded correctly(all other bottlenecks ignored).
I would say this development would give you more performance, exactly like you want it.
Furthermore, if they learned anything from their FX CPU lineup, they'll make them just as easy to overclock. Already there is a basic overclocking tool available in catalyst, and I can imagine them further building on that as they did when powertune was introduced.
You dont see the current amd/intel CPU's and AMD Boost cards perform differently per chip while still clocking dynamically? They'd just have different ceilings when overclocking, just like AMD's gpu's have had up till now.
Maybe in the future we can see cards with a dual BIOS where the second BIOS doesn't have any of that gimmicky crap and you can work out a stable OC yourself? That would be the best of both worlds.
Enthusiasts that set clocks of their PC are very much a dying breed in today's market, if only because hardware performance is so great that there is little need to. You simply don't NEED to overclock to get decent performance anymore.
So AMD can lock it up tight no problem. Let's see some apps that actually really push today's systems, and then I'll start to get concerned.
Sure, hard-modding is nice, but wtf am I spending 500 dollars on a GPU for if the POS can't even OC without resorting to hard-mods?
Locking them down is their way of cutting out enthusiasts (since we will get more for our money). Morons, who the F do they think buys their shiny new crap?
If the partners aren't stupid they will just start making all non-reference cards.
Until about 3 years ago, I was hardmodding EVERYTHING, even VGAs with software voltage control. This past weekend, I soldered a bunch of crap to my ASUS DirectCUII 7970 VGA.
Uh, I hate to break it to ya, but the only way to get real clocking ability, TODAY, is with hardmods. Software controls are BS, don't work often, and introduce issues all of their own.
I will relish the day when overclocking again becomes a skill, rather than clicking options already provided to you by the OEM...LIMITED options.
you like those options? You aren't aware of how truly limiting they are. Perhaps checking into what an "Infernal Device" is, legally, may help you see things as I do. It's not legal for anyone to sell you something whose use is gonna cost you more than the value of the item.
AMD Radeon HD7970 reference cards...have VRMs that can do over 1.9V. Every single one. Software options allow up to about 1.4 on average. You want more of that? Locked, limited options, introduced with limits so that they aren't selling infernal devices?
The whole reason AMD is investing the time and R&D dollars into this is because the current version, as many have already posted, is clearly broken. I use AMD cards daily, and really, any advanced features, like software clocking, BIOS flashing, Eyefinity...they all cause so many issues...
All I want is a good experience. AMD does not deliver that right now...in fact, no one does. The market isn't shrinking because of a lack of tech...it's shrinking from a lack of interest. OEMs need to appeal to end users, and obviously, by sales figures, current methods are failing, and software overclocking is one of those methods used to appeal to the consumer.
I used to hard-mod several years ago b/c it was the only way to get good clocks/control even for everyday use. But the new hardware is sufficient with software control.
If locked down we'd be stuck at 1.1v. I can deal with 1.4 just fine LOL.
The idea that we aren't already locked down, except if we pay enough, is FACT today. You buy unlocked chips from AMD and Intel, and the rest are locked up tight. NVidia GPU designs are restricted at the hardware-level, offering only minor adjustments, even so restricted that NVidia board partners have had to remove such functionality(eVBot) from products already released. AMD GPUs are limited by software.
By locking things up, they can ensure far greater quality in the parts sold, and when it comes to CPUs, boards already limit hardware physically with VRM designs and restrictions in BIOS.
My ASUS HD 7970..locked up tight, hence the mods. Look through the forum pages, you'll see many users complaining about it...especially since it advertises right on the box that these software adjustments would be there...but they aren't.
Now, they want to take advantage of those restrictions. This IS a good thing. And it's almost 100% completely deployed already. This isn't an idea...it's what's coming.
The idea that you don't need the extra performance is debatable, though... I'm getting about 25-30% more real world performance than stock 7970s due to my overclock. It makes a pretty significant difference in demanding scenarios.
And software overclocking abilities are just fine on the 7970... I can't imagine there's an air-cooled card out there that would perform better with hardware voltmodding over software voltmodding. The temps start to get pretty crazy above 1.3-1.4V and many chips don't even do their max OC at that high a level on air (I've had 3 cards and all of them hit their maximum clock speed around 1.22-1.24V).
If they lock stuff up, it creates an easier development platform, which to me, means higher-quality software. Consoles draw so much power from so little for good reason...years and years and years of software development. PC hardware...gets 12 months before it's mostly considered "obsolete", and 4 before all support is dropped entirely. 6-series AMD VGAs...get how much driver attention? 5-series? 4-series? Of waait..4-seires is already obsolete.:p
BTW, my 7970 doesn't clock worth a damn, and is locked from voltage adjustments. My 7950 gives me better performance. Why? OEM freedom in product development. You can't imagine a card that benefits from hardware mods on air, yet one sits in my machine right now, and many other machines across the globe. :p We already have two threads here full of complaints, just from this one model.:roll: