M17xR2 Power Supply Shutdown with AMD 8970m

I recently had two issues with my power supply. Haven't been using my M17xR2 for a long time, took it out and started using it. At first it was fine, then all of a sudden the power supply refused to "handshake" with my BIOS so when i turn on my laptop, i would get the infamous 2 beeps and "Non Dell AC adapter" BS. Battery won't charge, CPU & GPU slowed to a crawl and sluggish system overall. However, once in a while it would reestablish connection and everything went back to normal.

So i decided to leave the power On for 1 day continuously and i found that the occurance of "Non dell AC plugged in" message has reduced significantly. I continued with this experiment 2 days and sure enough everyhting went back to normal and the AC is now detectable almost every single time i power on despite turning it completely off from the socket and turning it back on when i need to.

Anyway, after that i upgraded my GPU from 2 very old 5870M crossfire to a single 8970M 4GB and needless to say the improvements was massive. Everything went well for 2 days until i installed Rise of the tomb raider and ran the built in benchmark. While running the benchmark, occasionally the power supply will cut off entirely which caused the laptop to crawl to a halt because it was running on battery. I knew the power supply turned off by itself because the blue LED ring around the DC jack had disappeared when this happened. At first i thought the power supply was blown because when i switch it off and on again, there was still no blue ring light.

But if i were to switch off the AC main for a few minutes and then turn back on, the power supply will come back to life with the blue LED ring. Then again, if i run the benchmark, quite often the power supply will shut itself down. What is going on? is the game benchmark causing the power supply to overload?

For your info, i have the 240W Delta PSU. My CPU is 940XM overclocked to 3.45 or 3.6GHz using TS, with TDP set to 75W and TDC at 62A. The 8970M itself is rated at 81W TDP in power tune and the GPU is undervolted, running at stock clocks. at any rate, the 240W PSU should have more than enough juice to run 74W CPU + 81W GPU with plenty of spare juice left. Again, i'm only using a single 8970M 4GB. But if i play rise of the tomb raider for 2 hours straight, everything seems fine. only in the built in benchmark the PSU will cut off by itself. this happened regardless of using stock clocks or overclocked on the GPU.

Is anyone else with 8970M 4GB suffering the same problem? Is the PSU overloaded or it's giving up? my motherboard power jack has been soldered recently when the PSU failed to "handshake" with the laptop so the jack couldnt be the issue. Any suggestions or experience sharing is appreciated. sorry for the long story.

 

The PSU doesn't seem to be overloaded, but it does seem to be failing. How long have you had it? Do you know its age? Are you able to test with another PSU?

 

Thanks for your reply.

My PSU is a 240W Delta variant. I have been using it since 2010 till 2016 and i hardly used it for almost 2 years until i recently took the laptop out to upgrade the GPU to 8970M. I might have to test it with another PSU or maybe try to mod an adapter using @Nospheratu 's guide. At the moment it's still working fine as long as i'm not aggressive with the overclock, meaning undervolted GPU and light OC with the CPU.

I was wondering those who owned M17xR2 or similar models with similar 240W PSU ever experienced the same issue with their PSU.

 

My friend owns the M17X R2 with a single 5870, and a Q720, and his never throttled, so Im going to go ahead and assume faulty PSU, or you are somehow magically reaching the limit, even though you theoretically should be just about at the maximum range of it

 

Funny thing, my PSU started doing the same exact thing right after I commented lol

 

LOL! Same symptom? shut down by itself after running something extensive like gaming benchmark and switch to battery mode. blue LED ring around the DC power jack connector disappeared. Can only recover after switching it off for a minute or so.

 

I think mine is just the PSU being overloaded, which is odd ( I have the M17X R5/M17 R1/AW 17 2014) and a 180w PSU, when I start the Rise of the Tomb Raider benchmark, without VSync for unlocked FPS, it reaches around 100-90 FPS (DX12 ofc) and then quickly slows down to mid 20s/30, CPU throttles instantly to 10W (0.8-9GHZ) GPU clocks down to 600/300 from 1075/1375 and the Battery icon is shown to have A/C Unplugged, but the adapter is on. Which is weird, considering 40-30W CPU/75W GPU draw, and estimated total draw of maybe 150w? I don't know honestly. In conclusion, I'm assuming the same thing is happening to you, A/C shutdown, because you have a 100w+ GPU and a 75W CPU, but still a lot of headroom

 

Yeah, happened when i ran rise of tomb raider benchmark. normal gaming seems ok. how about 3dmark firestrike benchmark?

How are you able to hit 100fps with 8950M on your R5 system? i'm assuming your graphics detailed is set to low? and 8950M is not capable of DX12 right? i have 8970M and i'm only getting 31fps on max graphics settings on DX11

 

8950m=Firepro M6100=GCN (768 GPU cores) I use a really ranged set of settings, High Textures, 16x AF, Med Shadows, 720p @ 85% scaling, and everything else low, I have a 900p screen. No DX12 nets me ~70FPS, DX12 gets me roughly 80-100FPS (before throttling) because of GCN. ALL AMD GPUs above the HD 7670/8670 or 7690M support DX12 Async, ex making more efficient use of of the GPU, because it uses the GCN Architecture. Remember, these are mobile GPU's so they aren't directly equivlant to the desktop counterparts. The 8970m has the same core count as the HD 7870/8870/R9 270x/R7 370x, but performs about the tier below it or in between due to underclocking. The 8950m is based of off the R7 260/R7 360, and performs about the same, having a similar clock speed, slightly lower mem speed though. DX12 Benefits all AMD GPU's based off of GCN, even if they are listed as not nativly supporting DX12. Turn on DX12 for a decent FPS boost.
EDIT: Sorry for long rant, I don't own firestrike, But i'll check free trial
EDIT #2 I remember playing R6 Siege last night, and GPU was 100%, but its was probably thermal throttling (90C+), I don't have the readout or logs from that time, and I just repasted so its probably finally able to unleash its full potential or whatever, my best guess.

 

Good to know GCN 1st gen can do DX12 too. i was always under the impression it can only support up to DX11 feature level 11_1. I'll give DX12 a try hopefully can get some fps boost.

One thing i'm puzzled, on my 8970M which has higher TDP, my GPU temp never exceeded 80C. when the fans kick in high, it will level off around 71C for GPU. isn't R5 Ranger a much newer model than my very old R2 and GPU cooling is better due to more heatpipes and better fan? you should not be getting upper 90C or thermal throttling for 8950M.

On a side note, i have overclocked my GPU to 1GHz core and 1.35GHz Memory, which ends up faster than the desktop 7870 Gigahertz edition. It was running stable too, until i discovered the PSU shutting down when running rise of tomb raider built in benchmark. now i have to dial back on all the overclocks, GPU and CPU included until i can find a solution to the PSU problem.

 

All GCN cards support Async Compute, the main feature of DX12 that allows them to perform much more efficient. The reason they are branded as 11.1 is they don't support all DX12 Features, like resouce binding (like SLI/Crossfire, except allows any two GPU's that support tier 3 binding to run together as one, here are a few interesting tidbits if you want to know more
Explanation of mGPU vs SLI/Crossfire
https://en.wikipedia.org/wiki/Feature_levels_in_Direct3D#Direct3D_12 Which GPU arch's support what tier

As to question why my 8950m is 90c+? Well, the I had it terrible paste, almost none, I repasted and it runs much cooler. As well, I don't have native fan control, so I have to jury rig HWINFO64 every boot time to max out the fan speed, which is only 3400RPM

 

As well, when doing the benchmark, do you notice high total power draw? (HWINFO64, maw CPU+GPU Wattage

 

Hwinfo does not report any cpu or gpu wattage for my very old laptop. Maybe it isn't supported at that time. On my R4 it does state the wattage for both cpu and gpu, which is why I can't tell how much power the cpu and gpu are drawing unless I have a power monitor on the AC

 

Afterburner allows you to monitor wattage, although I heard from somebody that it was just based off of voltage

 

Be sure to check the DC in Jack just to make sure its not loose or anything of that sort. Though I would put my money on a failing AC adapter.

8970m xfire also works in that machine but you will definitely need to Mod a 330w AC adapter.

 

That, or it being overloaded

 

240w wouldnt be overloaded if its functioning correctly. Even overestimating power consumption at 170w is still fine for the 240w AC Adapter and thats before considering efficiency.

 

170w+30w for other devices, and thats assuming the GPU isn't going above 100w, but yeah I agree with you. My 180w seems to be at the limit with a 75wGPU+40wCPu

 

Then you factor in efficiency (typically 80%), dont use software for determining power draw, they are practically worthless for getting accurate measurement of power draw.

 

Welp, its the best I can do

 

Kill-a-watt is really cheap

 

This became a major issue just about any time I would play taxing games, with my previous 780M. At that time, I also noticed my gpu temps lingering near, and sometimes slightly above 100C. So, I cleaned my gpu and cpu, repasted them both, and even through in a copper shim on each, and my heating issues dropped tremendously, and as a result, my power supply quit tripping as well.

Not sure if it is related, or what the average temps should be with your gpu, and what they are, but worth looking at.

 

thanks for your input. my temps are actually really good. GPU when not OC, leveled off at 71C when fans are on high. If i OC to 1GHz+1.35GHz TDP 81W, it will level off around 75-76C. CPU OC to 3.5GHz TDP 75W will max out around 80C.

What GPU are you using now? is your 240W PSU still alive and working well?

 

Just an update for those who are interested.

Got my second 240W flextronics PSU. My original PSU was delta 240W.

The flextronics PSU allows me to push my GPU to 1GHz Core, 1350MHz Memory @ 1.025V, 81W power limit with CPU @ 26x 75W TDP 65A TDC stable

If my CPU is set to 26x 82W TDP 62A TDC, even the flextronics 240W will shut down when i run 3DMark firestrike combined test.

This somewhat confirmed my delta PSU 240W is not capable of delivering enough power to the laptop. However, i'm still puzzled at why even the flextronics 240W PSU shut down when i set the CPU TDP to 82W. 240W PSU should be more than capable of running a single 940XM overclocked and a single 8970M overclocked especially when 8970M has a power limit of 81W by default.

GPU 81W + CPU 82W = 163W

Let's say the GPU is capable of drawing more than 81W because it's overclocked, at the speed and voltage i'm running the GPU could be drawing 130W?

GPU 130W + CPU 82W = 212W.

Even so, that should have plenty of room left to drive the rest of the system components like LCD, HDD, SSD, Motherboard etc. This can't be right. i've seen forum members running TWO 8970M and 940XM no OC on a 240W PSU and non of them ever reported power supply shut down.

The old delta 240W PSU is only capable of running the GPU undervolted to 0.975V, 925MHz 1350MHz 80W power limit and CPU 26x 62W 62A. anything more will cause the PSU to shut down.

I'm not from the US, so getting a kill-a-watt is not easy, not to mention the plug and voltage is not compatible even if i import one in.

 

Pro tip, 8970m is a 100W GPU, not 75W. Yeah, who knows why your stuff is shutting down, it could be that your DC/IN jack is slowly failing, that seemed to be a prominent issue with other M17X R2 users. My own stopped being able to charge, and I'm still figuring it out

 

Oh I forgot to add, you can't just factor in CPU+GPU and call it a day, you also have to keep in mind HDD/SSD's, Screen, Motherboard, RAM all take up power, around 20-30W total

 

yes i did factor in other components as i had stated in the previous post. and also, i had my jack resoldered when i started having PSU not recognized problem. but even after resoldering it, nothing changed. how's your psu?

 

So youre probably just reaching over the threshold of the PSU then. The people with dual 8970ms have the 330W PSU with a 240W chip ID, it cant support 330w natively, but it can with the mod. There is no way you can run a OCed i7 940xm or 920xm with two 8970ms on 240w AC, its impossible. I haven't resoldered my DC/IN jack, Idk if I am going to tbh. PSU worked with my ranger, but couldn't charge it

 

Just for reference, I ran 3dmark 11 to give you a baseline. 940xm running at 25-25-25-25 and a single 8970M running at 900/1250 undervolted to 0.975V (instead of stock 1.025V-1.050V) pulled 185W from my power meter. Safe to say you are probably tripping the power shutdown since you can't power everything with just a 240W power supply pushing the CPU to 26-26-26-26 and 1.025-1.050V 8970M.

 

thanks for your reference input. that helped to answer a lot. hard to imagine even an undervolted 8970M with a 25x CPU is already pulling 185W. If that's the case i've definitely blown past my power budget. Right now i'm running stable on 26x 75W TDP 65A TDC CPU and 1GHz 1350MHz GPU with vbios TDP limit 81W.

Just to check with you, when you pull 185W with your CPU @ 25x, what is your TDP and TDC in throttlestop?

 

100w/100a, I don't want the CPU to throttle down clock due to power limit so I let it consume whatever it needs.

 

Its mainly due to the fact that 45nm is so inefficient with power

 

wow.. setting it at 100W/A for both means you are getting 25x consistently. it's a shame we can't check the power used by CPU with hwinfo for 940xm otherwise we'll know exactly how many watts the CPU is pulling.

right now i'm prioritizing GPU performance more than CPU cause i'm running on a single 8970M. So i'm limiting my CPU to 75W TDP so there is more headroom for 8970M to run at 1GHz 1.025V. i tried 1V but even at 950MHz it wasn't stable and my ASIC quality is 78%. would be great if it can run 1GHz @ 1V.

thanks for your input.

 

Can't really compare it with modern CPU. 45nm nahalem was top of the line back in the days of 920/940XM.

 

Of course not, I'm just saying its extremely inefficient, but its still a decent performer