580M SLI throttling Discussion Thread

extremely noisy, the fan goes on full power all the time, all the time! damn! 85 is only sometimes, its rather hot today..

how do u get m18xr3 anyway? XD

 

Can you guys clarify whether or not you are seeing throttling in games? I can't even get into the discussion of throttling in Furmark and many other synthetic tests. If you are seeing throttling in games with the latest AW drivers and BIOS, that I can escalate and address.

Dell's official statement on Furmark.

 

Yes we are, and more often the more demanding the game is.

 

I can assure you i got throttled in BF3 Beta with the latest 280.35 beta driver from nvidia, and i am using a00 bios. Throttling commences when my temps hit 82 or 83 this is only if I overclock to 740/1760 or 744/1760 if i run on stock, I did not experience throttling, yet.

 

So, OC you get throttling and stock clocks you don't?

 

I do experience throttling in games only if i OC my GPU's to the max, on stock voltage it's 750/1750 And, of course, all the synthetic tests like 3dmark and shows throttling issues even with the slightest (50Mhz on the core) clocks increment for me. I decided to undervolt my cards from 0.87v to 0.80v (which is not 100W of power draw) with the same stock clocks and Furmark still throttles. I use the latest A03 bios from Dell and 285.38 NVIDIA BETA drivers, BTW.

Bill, can you please comment if Dell or Nvidia is working to solve this issue? If it's power protection function not to fry a power brick, then why a 100W GPU in SLI along with 45W in games draws up to 313W peak in 3dmark on stock clocks???

We agreed that it's driver related problem and Nvidia is working on better optimized driver to fix it... Is it really so? tnx

 

Exactly. Not in games and 3dmarks. But it throttles severely in Furmark with no OC at all...

 

I'm using my Bosses M18x with 580M SLI here and there until my M18x replacement arrives. After extended play at max settings with NO OVERCLOCK(1-2 hour+) Crysis 2, Bulletstorm & Bad Company 2 throttle at 75C to 78C. If you let the game continue it will eventially go back to stock clocks and reset the throttle timer.

It would be nice to remove the throttle for OCing purposes....but I would at least like to game as intended at stock clocks w/o throttle. My boss is a poser and could care less about graphics....lol.

 

Re throttle and nvidia: nothing at my end. I was told I have to wait since they needed time to 'research'. Well, waiting... BTW, the only game that caused the throttle to kick in was STALKER Clear Sky. I played Witcher 2, C2, Bulletstorm, Bad Company 2 etc etc and never had any problem.

 

Nice. I hope my M18x arrives with no throttling issues. Just like yours did.

Because I have only seen throttling to some degree at stock clocks on the M18x in front of me, I can not give a good review of them in SLI.
Since you aren't really experiencing throttling can you give me your brief opinion/review of the 580M's in SLI?

 

Bill - It is sad that we have to undervolt our cards to prevent throttling in certain games even at stock clocks. We are hoping for a fix ASAP.

 

Hey Bill, the throttling kicks in if only the cards run on full load for too long (that's some 60-90 seconds). I guess this is the same mechanism that prevents overheating/over-anything in Furmark. It is engaged when 100% load of any of the cards is detected, in any game/application.

I am in contact with Nvidia about this issue, and it's already been confirmed this is the case. 100% full load = throttling down. Also note that Clevo cards are not affected by the problem, only Dell 580ms, both in m18x and in m17x, so I seriously doubt Nvidia is here to blame. You guys screwed up the cards. In the end Nvidia just provided the chip. When will this be fixed? Let's face it. We paid a lot for this laptop/these cards and how about we are actually able to game the way we want?

This issue needs to be revised and addressed as soon as possible. Super duper gaming laptop worth 3-4k USD... with throttle. This is ridiculous. (Is the throttle included in the price or is it free, BTW? As e.g. some Dell only extra rare for-the-win pro feature?)

 

Ok.....now we are on the same page.

I'm sure this M18x beast with 580M SLI is throttling. In Crysis 2 w/DX11 patch at max settings whenever there is a prolonged heavy fire fight, the 580M's clock speed drop to 71 Mhz from 620 Mhz. The exact same thing is happening in Bulletstorm, Metro 2033 & BC2. I tripple checked the temps while the 580M's were at full load. The temps never break 74C. I am certain that this is a software problem that is out of our hands.
When the throttle kicks in, it is worse than playing with the Intel HD 3000(IGPU).

Bill please continue to help us get a fix from Dell.

 

That's right, Skygunner, the throttle kicks in at different temps. As soon as the load reaches 100% it kicks in, even at 60-70*C. The throttle is neither temperature nor power related, BTW. Some of you are having troubles accepting that. The relation between temperature and the throttle is coincidental. And yes, surely, we are looking forward to dell fixing the problem.

Skygunner, did you try 285.27? Try those. I had throttle trouble in some old games and with .27 GPU loads fell from 100% down to 30% in e.g. Heroes of Might and Magic 5.

Bulletstorm isn't really demanding, is it? AFAIR the loads oscillated around 60-80% on 280.xx. Never really experienced any throttle in this game. BC2 is the same. Not exactly very demanding. Runs okay.

Nvidia people told me that dynamically managed clocks are intentional to prevent damage under extensive loads. I guess Dell screwed up the design/programming of their 580 and this function doesn't really work in the way it is supposed to - unless 'extensive' means 60 seconds. There is no throttle problem on Clevo cards (also confirmed).

 

Right now I'm using my bosses M18x until my replacement arrives. It's currently on 285.38 beta. I haven't tried 285.27. I heard 285.38 has a huge increase in performance compared to 285.27. Have you tried 285.38? I know that on the dell drivers the throttling is the same as 285.38b.
What more can we do to accelerate a fix for this?

I just called to check the status of my M18x w/580M SLI and they are saying I should receive it this Friday or next Monday .

 

getting my m18x wednesday with 580m sli's... if there's anything i can run/do to help guys let me know.

 

Now im hesitant of buying an m18x with SLI, which was what i was shooting for.

...i think i may wait and see what happens

 

honestly, i think the best thing we can all do is call and complain. everyone who has dual 580's should be experiencing this... and if they're not, they're not running demanding games... which is a huge waste of money.

we need to keep calling in as often as we can and complaining about it. they will reach a point where they increase the priority of the issue because of this.

also, people who order their m18x with sli 580m's need to threaten to RMA for a refund if the issue is not fixed soon. it's that simple.

all together now.

 

Dell needs to fix this, paying this much with these issues is wrongful treatment of customers.

 

Throttle can be implemented in a few ways.

• Driver controlled
• Chip level controlled (BIOS support may or may not be needed, so that two sub modes)

Driver controlled: would be very straight forward; look for a certain signature on an exe file more like heuristics, known app names, file size etc that could be tagged to trigger clock throttles. I read some reviewers were able to fool the protective measures by using another app that is not popular.

So Question: Does renaming the app change anything?

Chip level: One simply has to look for current draw per logic, this can be done on power gating transistor(s). Measuring the power draw per logic and then combining the power draw for all the logics in the GPU helps in a deterministic weighing of what is safe and what is not safe. Temperature has no connection here.

High 'I' (current) is the main concern as it can lead to break down of vias (interconnects) due to electron migration. The other more important factor being the VRM phases. Most mobile cards have only 1-2 phases at best. On desktop cards these days we see nothing less than 12 phases to take up the load. Hence the need to protect the VRMs more than the mobile GPU itself.

Since some review sites managed to beat the protection with an unnamed app unlike furmark, it seems driver checks with known signatures and then accessing VRM controller (via BIOS support, A02 onwards?) to issue throttled clock is the method in use.

This is a preliminary mode to the more draconian anti-overclock setup that will surely come to laptops and maybe even desktops in the near future. That is full chip level weighted power control. Here, over all power draw will be the ultimate factor not application signatures to trigger the throttle. They wont need driver triggered mechanisms at all for such a setup. it will be completely on-die. The chip communicates with VRM to directly issue the throttle. That means the end of high overclocks.

We might be able to push a little but we will hit the same roof which will be the same for anyone. Good side is excessive throttling wont be needed, but the full potential of the GPU's overclocking abilities will never be reached. The only work around will be to physically mod the VRM controller feedback path to surge more voltage into the GPU core and similarly for VRAM.

Bad for enthusiasts but GPU makers need to change their approach with more low-power draw designs to start with if they want to offer good overclocking options. If not any GPU maker, be it AMD or Nvidia, will have to start taking drastic measures to prevent high current damage to their products. There is simply not enough room to add more and more VRM phases and then increase cooling capacity to compensate.

Reducing channel width of the transistor is only making leakage increase and tricks like power gating and clock gating have their limits in containing the negative effects. Gate oxides are the main focus now in the industry replacing SiO2 (silicon dioxide) with materials like hafnium oxide and yttrium oxide.

Hardcore hardware volt-modding benchers will still be setting the blazing records, but the more casual overclocking that we have recently enjoyed is increasingly looking like its coming to an end unless power demands are contained.

 

Someone try the M15x 260M throttling fix I used a few years back. I bet that might work for the 580Ms as well.

 

can you link us to it or give us info on it Joker.

 

to be perfectly on honest..(and don't take this the wrong way), but some of you really need to not speak.

bottom line bill...forget over clocking! take that out of the picture! this thing throttles during normal use..which im pretty sure you guys knew this. since you have a gaming department who test these machines.

also..not sure how it's driver related since everyone is using the exact same drivers. (drivers from nvidia) and the clevos can still game without being hindered.(unless you have some dll files specific to dell??)

now when you say your research team is working on the problem..care to elaborate on which part of the team is working on what? example.. video bios team. driver team. motherboard bios team. game testing team. or all the above and more...

and this ovp/ocp that is in place..is it just like the protection on the desktop cards(gtx580/gtx480)? im sure that 98 percent of the questions we are asking are under nda and all...but can you at least tell us something we don't know?

thanks in advance.

side note:
over clocking is not a right, but privilege....

 

I actually never said anyone is working on the problem yet. If they do start looking at the problem, I'm sure it will be some combination of all of the above. It would be a core team made up of engineers from several teams, I'm sure.

I can't answer that question for 2 reasons: 1. not familiar with the ocp/ovp on those desktop cards, 2. not yet sure ovp/ocp is what's at play here.

I will send a note to the engineers and see what they say. As always, I'll keep you in the loop as much as I can.

 

Thank you Bill for the clarification. +rep for the help you gave with OSD and now this.

 

thanks bill for the quick reply!

just trying to get some answers...not pointing fingers..

 

No worries. I just have to clarify, so a casual reader of the thread will see that and think something is not what it really is. Expectations can sometimes get blown out of proportion based on the most innocuous things.

At this stage, I am in information gathering mode. Just about to pull some select quotes from this thread that illustrate what you guys are experiencing and send them over.

 

understood.

and if there is anything we can do in the way of help (probably not complain and call so much.. ..which i can not guarantee.. haha) just let us know.

 

Actually, calling and complaining are encouraged. We need the feedback. Positive feedback is great and all, but we can't improve if we don't hear the negative.

 

then hang around a bit...you'll get plenty of that negative feedback..concerning this and the windows areo issue.

 

Aero is fixed, no? At least whatever we owned that was causing it to crash is fixed. Am I missing something?

 

great. that's good news.

my windows areo always worked, so i havent been back to check. just remember seeing that topic always on top of the list.

 

It is fixed for the most part. On one out of the two M18x's that we own it's fixed. But both M18x's have random Aero flake outs with PowerDVD Ultra 11, when a Blu-ray is played.

I am personally writing that problem off as fixed.
Well done.
Now if we can focus all of our attention on 580M throttling

 

I will be sending PM's to gauge your willingness to do a system capture. Thanks for your feedback and assistance.

 

Are you looking for OSD/Aero problems or 580M throttling? If it's OSD don't bother. My M18x that continues to have aero flake outs with the new OSD is experiencing the "5 beeps" error. Which means CPU or motherboard failure. For the most part on the other unit Aero works perfect except for PowerDVD. You guys have done an exceptional job on OSD. I wouldn't want to waste your time with OSD. With 580M throttling I am using my bosses M18x with 580M SLI(throttling) until my unit arrives on the 6th(EDD).

Once I receive my replacement I am willing to do what is necessary for the progress of a fix.

 

This is for 580 throttling. Let us know.

 

I asked my boss if I can send his and he laughed at me. What a poser. I wish my system was here now. I'd install some games and test it ASAP to make sure it is easily found by engineering for a system capture.
Since the chance of getting a 2920XM is slim, I'm sure some 2920XM users would be all over a system capture/replacement....I know I would if I had a 2920XM.

 

Boss wouldn't have to let go of the old one before the new one arrived. We'd only ask that he delete his data and leave the OS install in tact.

 

I told him. There's barely anything important on it. He just got one with 580M's to annoy me back when I had 6970M's. He could careless about throttling. I only installed games on it because I read about the throttling and wanted to see if it was true(It is ). I also am expecting my new replacement system with 580M SLI so I figured I needed to get involved.

I'm really greatful(700.00USD 580M SLI Upgrade from 6990M CF or 1189.99USD for each 580M) just to be here among 580M SLI owners. Dell upgraded me at their expense.

 

We want awesome guys like Dell-Bill B and DELLChrisM to continue to be our friends and forum liaisons. I would recommend editing the original and quoted posts and removing Bill's PM. Being cautious and asking questions about someone you don't know is wise. However, posting the PM is not appropriate, it's not polite to Bill, and it is also against Forum Rules... see the 11th rule in the list.

• Do not post content from private correspondence, such as email or chat, within the forums. Posting any other persons private information such as email, phone number or address could result in banning.

For the same reasons John mentioned, I don't share very much detail about my private communications with Dell on this or any other forum other than kudos for their good service.

 

yeah, i just edited mine as well mr fox.

just trying to make batboys job easier...

 

I have deleted a few posts
OP
1)•Do not post content from private correspondence as per forum rules
2) Company Representative's like Dell-Bill_B are here to help
They are legitimate

 

thanks to alex for handling this. folks, you all know by now we have rules here. Its something this forum will enforce - especially when it comes to inflammatory posting.

As mentioned by fox, john, alex and many other members, Bill-B and ChrisM are both with Dell. They have a different color username which denotes a member of NBR's Reseller Group. In other words, NBR has confirmed their affiliation with the manufacturer in question and NBR invites them to assist with support of their systems.

These 'resellers' are not obligated to respond to individual PM's however they do monitor various areas of the forum and assist where they can. Dell's ChrisM and Bill-B have both played a critical role in helping the Alienware Community with swift action with several problems which have surfaced and were first reported here on NBR. This forum has had a great relationship with AW Engineering and we hope it will continue.


Moving forward, if anyone is contacted by PM and is concerned with the contents of the PM, reach out to a forum moderator and we will be more than happy to assist.

No more inflammatory posts. Without further delay, lets return to the topic at hand.

Thx guys.

 

Thanks you mods and everyone involved in cleaning things up. Would it be possible to give this thread a sticky?

 

Stickied for the time being...

 

Let's don't get carried away too far here. Nothing has been done so far, neither by Bill or any other Dell employee, and m18x with throttling 580ms is on the market for good 3 months now. If Dell really cared, the issue would be fixed already. Let's face it. It's enough to google for 5 min to read all about the infamous 'throttle'. Secondly, it is not his courtesy or kindness to help out us. They sold us a troublesome product without a single word of caution the 'throttle' problem may occur. I'd say it's their obligation to be here to listen. Think about it.

 

Agree with this.

 

I agree with you...this configuration should have been right from the start, but Bill is just one person who represents Dell. He could just walk away from this thread if he so chooses. He recently helped us with AW OSD. We owe him our respect patience and gratitude.

 

Not exactly, unless you used the word 'represents' incorrectly. Representing Dell here means taking care of its customers. It means PR, it means corporate communication, it means corporate image. He cannot walk away if he indeed does represent Dell. If he walks away, he doesn't represent anything, and if that's the case, he cannot help us, can he?

The guy is unknown to me so far. I owe him nothing. I do not 'rep' people in advance. I do that for their actually commitment.

This is getting personal and it wasn't the point, so let me repeat myself: Let them fix things first and THEN be thankful. We paid for this stuff. They owe us.

 

A lot of what you say makes sense. They can walk away. Trust me. Go to the M17x boards and look for the plugged in not charging thread and resume sleep brightness thread. All I am saying is that he is a person that has already proven himself with the M17x R1's latency and the M18x's OSD Aero problem. Regardless of him being a Rep, he is not completely accountable for the 580m throttle. Were the people that bought hamburgers from Jack & the box able to hold the cashier responsible for the deaths and sicknesses caused by contaminated food? Even though the cashier sold it to them and served it? You and I are after a common goal. I will not continue to argue with someone on the same team. I've already said enough I do appreciate your drive to see this problem rectified.