I don't think you're understanding what I'm saying. There's 2 methods of switchable graphics. There's dynamic, where the drivers choose the GPU for you, and there's fixed, where you choose the GPU to use. AMD supports both methods in their hardware/drivers. Nvidia only supports dynamic switchable graphics in Optimus. My point was that it's illogical to believe that Alienware added their own mux and disabled it for Nvidia cards for some reason instead of just accepting that Alienware is just using the AMD provided fixed switchable graphics. Here's a slide from AMD themselves. No mux.
There are no switchable graphics in the p150hm so I have no idea what you're trying to say.
-
Since M18x R1 and M17x R4 had a dedicated MUX why shouldn`t they have it now? Just google it and you`ll see that Alienware haven`t changed their design. Atleast thats what they are writing
There is very few Nvidia notebooks with MUX switching R3d. Perhaps its more of a AMD thing and maybe Optimus doesn`t play nice with MUX. I have no idea -
still, i can assure you m17x has a mux, several re-sellers have stated that it does. -
-
IMHO. -
Why would anyone put a process to calculate how much GPU must be used? GPU doesnt need to calculate utilization it just runs as fast as its allowed to run. -
Let me rephrase my statement then: Fixed graphics can be enabled on notebooks without a mux so the case could be that the newest Alienware notebooks do/do not use a mux. We don't know. But we do know that fixed graphics can be enabled without a mux. So since the possibility that Alienware does not use a mux exists, one cannot just say for a fact that the problem with the P1xxEM notebooks is a hardware problem.
So saying that the main problem with the notebook is how it is constructed like OP did would be unsubstantiated.
Better?
-
This could be useful for desktops, but I don't see much point for notebooks with switchable graphics since you'd be running on the iGPU anyway. -
ok, i'll humour you. so for whatever reason AMDs switchable graphics implementation decides to limit the GPU's rendering performance in a stupidly complex way, - rather than reducing clock speeds as it has traditionally done for years, it shuts down some shaders or other parts of the GPU... Now that this is done I should still see linear response in my fps when I increase or decrease my clocks. I should also still see an increase in fps if I decrease detail, because I am reducing the amount of processing that GPU has to do... neither of those things happen, fps stays the same regardless of detail or GPU clock settings.
If this was the root of the cause it would also be very easy to fix and we would have received drivers months ago. -
Yeah I won`t even bother using my time on this thread anymore. So many things to comment. Believe what you want R3d. Lets leave it at that
-
Meaker@Sager Company Representative
AMD might have this functionality in their own chipsets but they practically dont exist in this market. -
-
seriously, thats your answer? so you are basically saying: AMD introduced a new power gating system with their new cards, that is only enabled in switchable graphics configuration, just because? on top of that they are unable to turn this "feature" off and it seemingly has a mind of its own and disobeys laws of logic. ok, totally plausible, im out.
-
This is how i see it, of course i dont have a clue if it works like this or not, thats AMD's little secret:
Enduro works and exists to manage the switchable graphics in the laptop. We all agree what switchable graphics and its main purpose is (saving energy and prolonging hardware life). So how do you manage this simple task?
You create a software that manages when and WHY one or the other GPU is used. This software has default settings, high performance, battery saving, and custom settings, etc, etc.
What these settings basically do is tell the software how to "behave" in a general way. BUT the underlying programming has to communicate with the hardware and tell it what to do. Simply put the instruction shut down or work at x% usage has to be translated into the actual action, that translation aint as straightforward as one may think. In order to cover all sorts of applications, games, programs, etc. the software uses one or more algorithms to interpret what you are using and consequently tell the hardware to behave in one or other way.
It probably is a lot more complicated but the route is something like:
When A is on, then turn off B and use C, and so on, with degrees of usage as well.
This approach explains why when lowering graphic details the % GPU usage lowers as well, yielding similar or the same FPS as high detail settings. Why? Because enduro somehow knows you are setting to lower details and interprets that the GPU should be consequently lower its usage.
The obvious problem is that we DONT want what enduro thinks its what we want, at least sometimes. One should even consider how enduro interprets different commands at the same time, for example if you are running a game, a word document and a browser at the same time, wich one should be preferred in terms of power usage and performance?
The fact that even windows power settings affect the % GPU usage when one lowers the graphic settings (see my post about this) supports my notion. When changing from power saving to performance windows settings (in windows control panel), a user reported %GPU usage and FPS for BF3 in lower settings increased significantly, and he didnt even touch CCC in the process. This clearly shows there is no hardware bottleneck at all, only conflicting software directions telling the card wrongly what it should do in different scenarios.
So its not only a calibration issue (wich would be an easy fix) but a whole reprogramming of sorts, change or modification of algorithms, etc.
BUT its a driver fix in the end, nothing more.
regards
Voz -
Originally Posted by Marc@AMD
I want to thank all those users that have been patient in this matter, and peristent in providing helpful information and feedback to the community.
We have received some positive news from the Driver Development Team. A driver is planned to be released in the near future that will provide significant improvements to notebooks enabled withAMD Enduro™ Technology.
Please stay tuned to our support.amd.com site for the driver posting.
Source: AMD Game Forums - The Clevo 7970M Enduro/under-utilization debacle. -
dont you think it would be very easy for AMD to disable this imaginary feature, like its disabled on non switchable graphics configurations?
and so far everything, especially BonsaiScott's posts point to a some sort of bottleneck where data has to be transferred from dGPU to iGPU. Thats not to say that it cant be worked around with drivers, Nvidia did it with Optimus so unless AMD do not have the hardware capabilities they should be able to do this as well. -
Mostly @ R3D:
There's 2 ways to get pixels off an MXM: The pins on the connector for one of the video outputs (there's a lot of them), or the pins for PCIE. That's it.
With current laptops there's one way to get pixels to a monitor: the wires on the monitor cable (or LVDS ribbon in this case).
In the P150, the 'monitor cable' is always connected to the Intel iGPU. The only way to drive pixels to the display is if the iGPU does it. Thus the only way to get pixels from the MXM to the monitor is to transfer it over PCIE into the CPU's memory system, have the intel iGPU pull them out and send them to the monitor. If the drivers are optimum (NVIDIA clearly does a better job in this case) and you're running 640x480 @ 30 hz, cool. But as resolution goes up so does the useless consumption of expensive resources. Eyefinity from a single displayport is coming (here?).
I don't care if it's NVIDIA or AMD, ATI, S3, Matrox, 3DLabs, 3DFX, Qualcomm or MacDonald's. Using MXM-->PCIE-->CPU DRAM-->iGPU-->wires-->monitor is fundamentally broken. MXM-->wires-->monitor being better is FITTCO*. The Alienware M17xR4 (with the mux) has the best of both worlds. Does it work perfectly? Nope. Can you control it manually and get what I want? Yep. That's why I'm buying one. Just wish they made a 15.6" with 1920x1080 (or higher) that will take a high-powered MXM and has a mux.
What I _really_ wish is that the P150 had a user-controllable mux.
* Fundamentally Intuative To The Casual Observer -
Besides i am not making it all up, i am just taking the info and opinions poured all over.
I can see your point (i.e. you might be right and i might be wrong), you are just not seeing mine, to me enduro does more than just on/off, it wouldnt even have a name if that was the case. Probably as you say it doesnt shut down shaders or lower clocks, but it might control or "limit" some other part of the hardware in order to achieve its goal (lower power usage).
Having not made tests myself i dont know why looking at a point or more players entering affect the FPS, but i do know that has always happened with any card (switchable or not), FPS always go down on more demanding renders. The fact is that if GPU usage is not 100% then the reduce in FPS is much more noticeable, so those more noticeable changes are just consequent.
I quote the original report:
So far i have read thousends of posts and not one of them proves this hardware bottleneck really exists.
I wont argue that a shorter road should be faster than a longer or more stepped one, but that doesnt necessarily means it would be noticeable when we are talking about inches at great speeds. So to me it aint significant and doesnt explain at all why the GPU is underutilized in certain situations.
Just to show you i am not the only one with the notions i try to explain (clearly failing at it, lol) i quote Jarred from Anandtech.
i have to go work now, will continue later.
Its really interesting to have this exchanges of opinion tho
regards
Voz -
-
"Fundamentally broken" is slightly hyperbolic. It obviously works. It should not be quite as fast as a dedicated mux, but broken? Not really. -
If you run that same card in switchable graphics mode, both ZeroCore and PowerPlay are still there and they work the same, so why would then AMD waste their time and program another function and add it to Enduro when officially it only function is just to switch between GPUs ( Extend your battery life with AMD Enduro).
-
Still sometimes a fresh analysis helps clearing the mist of hard knowledge, and thats all i am trying to acomplish, bringing my point of view to the facts at hand.
At this point we can agree to disagree in only one point, and that is the "bottleneck". I truly think you know better than me about hardware, clocks and what not, but i am still not convinced there is any kind of bottleneck causing the GPU underutilization issues.
One of my biggest weaknesses is expressing my views in few clear words, so i will make my best effort once more just to fail... lol
IF a hardware bottleneck is present why lowering graphic details yields lower GPU utilization, and tehrefore the same or similar FPS? To me that makes no sense at all, since by lowering the graphic details ingame what you are basically doing is telling the game to be simpler (graphically speaking). A good analogy would be a digital picture, if you lower the pixels (graphic detail) the image weighs proportionally and even exponentially less than the original "ultra settings" image, allowing a much faster transfer via internet or between drives, etc.
In the example the bottleneck for transfer speed is bandwith (i assume something similar would be the hardware bottleneck you are talking about.).
So lowering the image details makes it lighter and therefore transfers should be faster, IF we are using the MAX bandwith available (i.e. reaching the bottleneck). If transfer is NOT faster with a lighter image then the problem is not the bandwith bottleneck but rather something else......
Going back to our game, you lower graphic details and therefore the weight of the rendered data is lighter, one expects speed (performance) to go up, but it DOESNT. And here is where you say its a hardware bottleneck, but it cant be because essentially the weight is much less and bandwith keeps being the same.
So unless the hardware bottleneck actually becomes smaller by lowering graphic demands (something physically impossible, unless we are talking about temperature effects) in order to keep the speed the same for a smaller info transfer, there must be something else causing this symptom.
This something else is ENDURO and whatever algorithm it has that produces this undesirable effect, or if its not enduro its another piece of software, but its stil not a hardware part bottlenecking anything.
That is why i quoted Jarred's comment on his own article, because it makes sense that enduro is somehow trying to keep GPU utilization as low as possible as long as a certain value is mantained. This value might be 60 FPS or something like that. BUT the problem arises when enduro "fails" doing what its meant to do, factually crippling your performance.
NOw, why enduro fails? My theories are detailed in my last post, although i really have no clue.
Oh and correct my if i am wrong but FPS its frames per second, and that is in no way a meassure of weight, its clearly not the same to render 60 "ultra" FPS than 60 "low" FPS.
AND maybe the %GPU usage is not managed by enduro at all as you say, thats just a secondary effect of enduro thinking the job is done and telling the card to "stand by". But still thats not a hardware bottleneck.
i hope you can get something because i am a little dizzy right now, roflmao
regards
Voz -
so what the hell is the op saying? that people who are suffering from performance problems associated with the 7970m, aside from bluescreeens and other problems besides under utilization have nothing to complain about? People pay a pretty penny for this product, it doesnt work as advertised. End of story.
-
Think he's saying the 7970m hardware card is perfectly fine but not software/laptop.
-
-
-
sorry, i think im OT, but could any1 tell what they are discussing about here?
[Sammelthread] 17,3" Clevo P170EM(3) Gamer Notebook (2012) - Seite 58
is it OLD? -
How much they want to trade in the 7970M for a 680M
bla bla bla
But most importantly a guy who claims to have some breakthrough with the Enduro problems. It involves changing the EnableULPS in the registry of Windows. I have know idea what ULPS is, but it involves with messing with the power management so that you somehow disable Enduro functions and that the 7970M is always active in 2D mode, aka in the desktop, when browsing the internet etc.
Its a bad translation from google, but the steps are something like this:
1. Go in to Windows safe mode (press and hold F8 during bootup with his machine)
2. Hit the windows button, insert "regedit" to enter the registry
3. Find EnableULPS there
4. Disable it by changing its value from "1" to "0"
5. Not shure about this one, but there are 4-5 entries there, and they should all have the value "0" aka disabled
6. Reboot and you are done
(I think they are using the 12.8 driver)
This fix may cause your fan to increase in speed.
Two people have tried it, one didn`t notice anything, still low GPU utilization and FPS, another one had great results
He noted that his GW2 FPS went from 30-35FPS to 36-64FPS -
Mighty_Benduru Notebook Consultant
I think if only 2 people tried it, it didn't work. If a fix works, everyone will jump on it.
Anyway, it doesn't matter what is wrong with the 7970m, hardware or software wise. 7970m plus Clevo = a lot of $ for a lousy product that does not perform as well as it should. From the consumer's point of view, it's a broken product. There are only 2 solutions. First, swap out the card and replace it with something else. Second, cross your fingers and hope AMD will provide a solution.
AMD keeping so quiet and low profile just pisses me off. Resellers are still selling the combination as a high end product. It's only creating more victims. -
Woah, GW2 FPS jumping up like that is a pretty important recording, Mighty_Benduru. Apparently this DOES do some thing. I am going to post this on the AMD gaming thread and see if Marc picks it up.
-
-
But there is a guy in that thread who got a FPS boost in Guild Wars 2 when doing that trick. I think. Perhaps it was overclocking he was talking about then? Google translate is useless sometimes
-
i've had my ULPS disabled a while back as its one of the steps to get OC'ing working on 7970m.
-
-
-
Meanwhile in the Alienware forum:
http://forum.notebookreview.com/ali...talyst-12-x-9-001-september-4-unofficial.html
A place where one can be proud to own a 7970m.
Wake me up when October ends. -
Extremely angry. I want 55+fps BF3 ultra. -
I would've been happy with 40 frames per seconds in World of Warcraft. But this was too hard for a 2000$ laptop ... Definately no problems with 7970m, suuuuuuuuuuuuuure.
-
-
I just run kombustor than msi afburner and no problems. I can even bench and game at 1060/1560 stable. Just to remind that I'm using the 2.2.1 msi afterburner paccage( I know that yuo need to ad something in th enew versions)
Also check if you have the MSIkombustordx11.exe set to high performance and also afterburner in swithchable graphics.. -
recent news guys, maybe the grief is at an end...
-
Original Post 9/12/12:
Conclusions:
1) The M17x-R4 hardware does it right, but the drivers for that seem to be less than optimal (manual works, automatic doesn’t always)
2) Clevo (and others) could have done it the same way but chose not to
3) NVIDIA does iGPU/MXM ‘switching’ (which isn’t really) better than AMD
4) New AMD drivers can make it much better
5) There’s nothing wrong with the 7970M MXM
Validation this week:
"Catalyst 12.11 Official Beta is releasing today @3PM ET. Great improvements AND Enduro fixes."
"I'm getting 70-100 FPS now in MP3. I'm liking this 12.11 a lot. "
"Dude, Battlefield 3 is running so much better. "
... -
Want a cookie? We've all been saying it was a driver issue, so I don't get your point. Is it just for you to gloat? Do you even have a 7970M? Are you even invested in this issue?
-
And it still has issues. The discrete GPU still does not work with some programs (found that out today with Dolphin, heard it doesn't work with Photoshop either).
-
-
Yes, but that doesn't change the fact that it doesn't see the discrete GPU, lol. I get worse performance on my NP9170 than on my G73jh because of this.
-
There is nothing wrong with the 7970M MXM
Discussion in 'Sager and Clevo' started by BonsaiScott, Sep 12, 2012.