*OFFICIAL* Alienware 17 R5 Owner's Lounge

I just noticed that when you select a stored OC profile (or create your own) in ACC and then you test it, it resets the undervolt you have done with Intel XTU (probably throttlestop too) !

 

I have the feeling this is because the ACC is using the XTU library for it to work.

 

I found that OC profiles in advanced view give you the option to select manual voltage for the CPU (which is the VID?) and the voltage offset (which is what you undervolt). Does this mean that the VID can actually be changed from the ACC bypassing the factory defaults?

I created an OC profile to test.

https://imgur.com/a/RpFhh35

https://imgur.com/a/o3THq3M

 

YOu shoul dbe right, but I havent tested the ACC yet so I cant comment on if it actually works.

 

Undervolt through ACC OC works for sure! I rebooted the machine. ACC loaded the created profile and I saw the undervolt without changing anything. I am not sure about the VID because it changes all the time. Can you check it too please? I do not want to misinform

 

Hello, I have registered to the forum to add my results to those of doofus, captn.ko and aristotel https://imgur.com/a/td031Lk

As you can see I can reach 1400cb at x45 for all cores with -144mv adaptive (HWMon was opened after bench was finished and I redacted my pc name )

I have stock paste still and I am using bios 1.2.1

My power consumption was ~90W peak and not below 80W during bench so I think I have room for improvement with better paste

 

I just run it with your settings (-0.145V at x45). I scored 1402, Finally! But, still too low for x45.

https://imgur.com/a/LWGrOwK

So you did not get any thermal throttle with default paste?

 

Forgive me, but to be clear you mean since I am using bios 1.2.1 and you are using 1.1.6 (?) my "starting" CPU voltage is higher than yours, making my bench less impressive?

 

can you ran HWmon during CB please? And with temps please

 

Also enable graphs for all throttling and TDP in Intel XTU and run CB again with same settings while having the XTU open. Do it to see if any throttling (thermal?), not the score. This is the only way. According to what people say 1400 is low for 45X. Mine is low too for x45. Still happy I passed the 1400

 

As requested https://imgur.com/a/Oia2Rvy

Don't know what was different this time as i didn't change anything, but got scores of 1407 and then 1419 as you can see

 

No, I am using bios 1.2.1 as you. Unfortunately. VID is related to the CPU itself not BIOS. From your readings, it looks like you have lower VID which is better...

 

This is an official reply regarding current limit throttle on bios 1.2.1. The conclusions about the changes made on last bios are yours.

 

1400 si ok-ish for 4.3Ghz for BGA CPUs (captn.ko had 1400 at 4.2Ghz: http://forum.notebookreview.com/thr...5-owners-lounge.815492/page-140#post-10761383 ).

I don't recommend using xtu to monitor anything (it is a waste of cpu resources and can also interfere with other settings while xtu is active and can cause weird behaviors like power throttling even at a lower limit than the already low power/dynamic limit that some encounter or it can interfere with the readings of other monitor apps), you have all the info you need in hwinfo.

If you really want to see the behavior in more detail, use the log option in hwinfo and create a log file for the tests.

Also, don't rely on the first CB run for a good estimation of the current performance level, instead do multiple consecutive runs (at least 5 runs) without pause between them and save the scores for each run. If you have a limitation that isn't triggered during the first seconds, the first score might not be affected to much, but the next one will be.

I was tempted to reply to some of your previous posts, but I saw that you managed to find (how the OC profiles affect the voltage settings) what Papusan wanted to find with the pics he asked you to provide.

I presumed that you knew that there is a relation between the voltage, clocks, cpu load and power draw (Papusan already mentioned that in the previous posts), I already (indirectly) provided you that info when I previously asked you to test with an uv and you managed to run at higher clocks at the same power draw in the same test or you managed to achieve a higher power draw with higher clocks before you were current limited again (but that time, the power draw was already much closer to the 110w PL limits).

Also, the current limit in amps and the Pl1/Pl2 are separate power limits and act differently, so the PL1/PL2 limits didn't interfere with you previous run (you didn't reach those limits) and as you already found out (with your recent run), by uv-ing you managed to reach the PL1/PL2 limits without being current limited.

I'm glad you found out that the OC profiles are interfering with you previous undervolt settings and they override your previous settings, that is why I previously asked you: why do you need those OC profiles since TS can do the same things better, with more control and more options.

I already demonstrated that your voltage is one of the main reason for you clocks and current limit, when I asked you to uv and your clocks were higher, your cb score was closer to the "average score" and your current limit was triggered at a higher power draw.

After that I told you that from there it's a matter of fine tuning and you didn't have the time to invest for that.

I didn't think that you'll use the default voltage settings again when you want to OC or that you'll use the OC profiles for OCing and not UV again since your default voltage was already too high for stock clocks. At least now we know that you OC profiles are using different voltage settings.

It would be even better if you could post some pics of the available options when you want to create a custom OC profile in acc.

 

raz8020 thanks for your time and the huge post . Too many things to see really.

By the way, the speedstep set to 0 in TS does not work and getting 800MHz during idle and windows high performance. The unlocked options in windows high performance profile also did not work. Nothing I can do about it, it maybe the new bios because doofus has seen that too. But happens during idle and not during CPU load. I guess this may save power too (independently of the windows profile) and save power when using battery too.

hwinfo running can also reduce the CB score especially when logging . I have also seen that hwinfo is lagging when running CB (which is normal anyways).

The current limit throttle I am getting is finally because the system reaches and tries to exceed the PL1/PL2 and the system keeps it to the limits by lagging/reducing the performance by a reverse mechanism affecting the scores. I can play within the limits by slightly changing the UV and frequencies. So, the problem now is not if I can reach those limits or not but actually check if I can reduce the VID (power consumption) to reach higher frequencies at the available limits (because temps are not an issue now).

My VID hardly reaches the 1.5 but to set a PL1, PL2 to 100W I must select an OC profile which sets VID to 1.5V I guess. I do not know how to check the OC profiles regarding the VID. Here is a screenshot from OC1 and OC2 at their default settings.

Overclok profile 1
https://imgur.com/a/Foxjfnr

Overclock profile 2
https://imgur.com/a/gPJiVjo

You can always set your own profiles and manually adjust those but I am not sure if the voltage actually does anything for the VID as it goes up and down.

What I am looking for is if changing this voltage value (set manually) can slightly reduce the VID or not.

All my CB runs lately have been without using TS (only playing with Intel XTU UV and multipliers step by step before I reach PL1 and PL2) and ACC was uninstalled. So, there was no interference. I just installed ACC to play with the OC profiles and because I saw the manual options in them trying to reduce the VID (if the voltage there does anything).

I do not think that going from 1350 to 1430 or more (using same multipliers) is a matter of fine tuning. I think that the CPU VID is more important.

My OC profiles are really using different voltage settings compared to other R5s? How is that possible? In that case, can I change them like editing a file or something? By the way if the OC profiles are missing, the ACC asks to download them. So, I guess everyone downloads from the same place? But usually the OC profiles are installed with the driver for OCC (Alienware-OC-Controls_YWHD7_WIN64_1.1.8.0_A03 is the one I am using from the Dell drivers page)

My advanced options for editing OC profiles are below. This is actually the one I created. I uploaded to see the available options. I am not sure what the green point means. If it is the default or recommended value. I am also not sure what happens when moving the voltage point left or right (affects VID?). If you have anything to recommend please let me know

https://imgur.com/a/eaIKqUS

In addition, if I overclock too much may this mean there is not enough power to reach certain frequencies? My CPU is stable untill about -155mv! But, lower scores may be caused because of over UV too?

 

@Aristotelhs2060

What you need to do, is to find out the load voltage/cpu clock (with hwinfo, but set the sensor refresh to 1 sec if the default is higher) during the same test with the same 100% cpu load with different OC profiles. After that, select the profile that gives you the lowest voltage for those clocks and try to lower the voltage even more by using TS.

Or can you just select the performance option in the BIOS and use TS to manage both the OC and the undervolt so you don't have to use the OC profiles?
Can TS override the voltages that are set in the OC profiles?

I don't know (since I don't have an AW) what are the minimum necessary settings that you have to make, to benefit from the high power limits with minimum interference with the default voltages.

The current limit has nothing to do with the PL1 and PL2 limits.

 

Ok what is the current limit throttle then and how to solve it? It shows up only when the TDP is reached. In fact I have not seen it for ages now because I have learned how to keep the limit below TDP (or trying to) so I am only getting power limit now when exceeded.

Is the voltage you talk about the same with the one in ACC OC manual/advanced options? Does it have to do anything with the VID?


By the way, whatever you wan to use Intel XTU or TS or whatever you can have ACC installed. Open ACC (it may ask you to wait for 30sec for the service to run because I think it has a delayed startup). Close it. After that you can change what you want using TS or Intel XTU bypassing ACC options. But I only care for the VID now

 

@Aristotelhs2060

The current limit doesn't show only when you reach the PL1 or PL2 limit (please use the power limit term when you're referring to these, because TDP means something else and is related to the cooling design specifications).

You can be current limited before you reach other power limits like PL1 or PL2 or you can reach PL1/PL2 without being current limited.

The EDP flag is associated with the current limit:

Electrical Design Point Status (R0)
When set, frequency is reduced below the operating system request due to electrical design point constraints (e.g. maximum electrical current consumption).

Here you have a detailed explanation for the TDP, power limits and their behavior:

http://forum.notebookreview.com/threads/tdp-and-power-limiting-haswell.766743/

You can also read about this limits in Intel® 64 and IA-32 Architectures Software Developer’s Manual , go to 5.1.3.1 Package Power Control.

Mind you, that VID is not the cpu core voltage (vcore), it is just the voltage that is requested by the CPU and the voltage that the cpu is getting can be different depending on other factor like: the voltage offset, the internal calibration mechanisms (I don't know if the OC profiles alters the way the voltage is calibrated during the load).

I don't know what calibration settings does AW use, or if is uses similar ia ac dc loadline settings to the msi GT/GS series. @Falkentyne might know more about the calibration settings that AW laptops use, but if I remember correctly, you can't unlock the Aw bios due to the type of protection that it uses, so you don't have access to those settings.

And to answer your previous question, you only have to concentrate on the load voltage (compare them in the same benchmark/stress test for each clocks/multiplier) and find the profile or the settings that gives you the lowest voltage for the same clocks and further tune that voltage with TS.

 

I have repasted today but did not use "paste". I noticed that the GPU heatsink is being pushed up by the surrounding pads, from the factory. I was too lazy to spend time replacing pads. It is my feeling that some pads are too thick.

I have run some first tests and maybe I am marginally better than before, maybe I can extract 10W more before thermal and other limits cut in. I need to run a lot of tests to see if I have made an improvement over the previous attempts.

I have also fitted a slim heatsink over the PCH but it still touches on the plastic cover, creating a limitation of airflow. I am monitoring the PCH to see if I have made a difference there too - previous attempt was a 1" square aluminium piece which failed to keep temps below 100C. Undervolting the GPU to 950mV I think makes a difference.

 

You have really good knowledge of things but the best would be to have an AW too. I still have a number of question (if you find the time). Excuse me if I am asking the same things.

1)So, the current limit throttle is due to EDP? What does the electrical design point constraints mean in practise? Is there a problem with my motherboard? And why am I the only one getting this?
2) Regarding the load voltage, I guess OC1 should use the same or less load voltage than OC2. So, why bother checking those values? I could select OC1 directly and change the multipliers etc afterwards.
3)What if I select the overclock profile with the least load voltage? What is going to change regarding frequencies and scores? What if I already use the lowest load voltage?
4) If the VID is the voltage requested by the CPU and I undervolt it, then the CPU may not get enough power to reach higher frequencies?

 

The CPU requests a specific voltage when it is running at a specific frequency. Since there are two or three mechanisms by which the Intel CPU jumps/changes frequencies, and also throttles which lower the frequencies, you typically see the frequency up and down all the time, and the voltage too. Actually the frequency can change many times per second.

I presume that the BIOS will feed the CPU whatever voltage the specific CPU requires for a specific frequency. This information must be stored on the i9 chip and that the BIOS then reads, because it is different from CPU to CPU. This information must exist in the form of a table (or curve if you like).

When you undervolt/overvolt the CPU, I think you are applying a blanket offset to the curve, it is not an intelligent operation. In contrast, with MSI Afterburner, you can modify the frequency/voltage curve as you wish, not only with a fixed offset.

Yes if you undervolt too much the CPU will crash because it will be unstable at higher frequencies. The faster it goes the more volts it needs (and the more power it spends). You just need to find an undervolt level that the CPU is stable at, and leave it there.

 

Stable means that the CPU does not crash (I can set it to -155mV and still be able to run CB) or that the CPU must run at the specific frequencies? Even at stock without changing anything, I get this current limit throttle.

 

1. There is nothing wrong with you MB (and other people are getting this too, but I believe that there aren't many users that test their AW laptops or are aware of this).

You can consider it just another power limit, but instead of limiting the power in watts, it limits the maximum current consumption in amperes, but this limit usually has a range of values that are lower than what the power delivery system can handle.

If you want to know more about the relationship between voltage, amps and ohms, try the next article:
https://learn.sparkfun.com/tutorials/voltage-current-resistance-and-ohms-law


2. You assumption should be correct, but you should still test it to be sure. Also you need to further undervolt with TS and find the lowest stable voltage.


3. With a lower voltage you can get more mhz in the same power draw, or a lower power draw with the same clocks.


4 Usually, if the voltage is too low you runt into instabilities such as crashes, BSOD and freezes.

There are rare instances when there are no crashes and the performance is lower because the cpu is starved, but this is easily diagnosed by the considerably lower clocks and scores. As long as the uv is stable in idle and in stress tests and the scores in benchmarks are improving or are at least the same to the previous run, then the voltage/clock ratio is optimum.

 

Thanks for your reply
1) So my understanding is that my machine has set a really low maximum current consumption for some reason? Could this be inside the new BIOS? Is it lower than others or is it my CPU VID which wastes more power for certain frequencies?
2) Does the drop in frequencies (800Mhz) when CPU is not under load is a result of the current limit throttle? Or may it be due to the new bios or due to windows itself? I get this on every UV or not I have tried.
3) I have TS UV set at the lowest value for both voltage and cache (-125mv) and seems stable for x43 but also for higher frequencies . But the low score at CB means that it is not? Should I try with less undervolt? But then the power throttle comes together with current limit throttle which result in lower score anyways.

 

I have some very good results and news that I think will interest everyone.

As you know I repasted today and this is my third attempt.

Here are the "before" and "after" scores.

**************************************************
BEFORE - repaste with Arctic MX-4

room 28C, 140mv undervolt, Prime 95, small FFTs, 10 threads.

3.5GHz P95 10t, 67W, 90C no throttles (140mV undervolt)
3.6GHz P95 10t, 72W-73W 95C but rising, throttle kicks in

**************************************************
AFTER - reworked
room 26C-27C, 140mv undervolt, Prime 95, small FFTs, 10 threads.
​

3.5GHz P95 10t, 66-68W, 76C-80C no throttles (140mV undervolt)
3.6Ghz P95 10t, 72W-74W 80C-83C
3.8GHz P95 10t, 93W 90C-96C eventually power limit throttle

******************************************

As you can see BEFORE I was throttling at 72W-73W, now I am throttling at 93W - I GAINED 20W !
Edited typo

Ask me how I did this

​

 

How did you do that?

Enviado de meu Pixel 2 usando o Tapatalk

 

Some Cinebench scores:

settings 43x, 110W

BEFORE:

with a COOL machine, 1390cb - but later 1320-1340cb continuous

AFTER:

1386, 1380, 1386, 1388 continuous

 

1. I think is due to the high voltages and the way the voltage is boosted during loads.

2. No, this is a result of the power settings that are used, which have a bias towards power saving vs maximum performance, so the clocks are reduced as much as possible when the load is very low or there is no load.

3 You already had a low CB score for those clocks and if the sustained clocks are not lower, then it is not a symptom of too low voltage (also, when you get to the point where the voltage is too low, you'll most likely have crashes or freezes so you'll notice them quickly).

I already mentioned that, the first run of CB might not represent the current performance levels and the score that you should compare to, is from the second or after more runs if it keeps getting lower with consecutive runs without pause.

 

Well, you have the same power draw as before for 3.6Ghz, but your temps are lower, so you probably managed to improve thermals.

 

I need help with my Alienware. Two days ago I got it repasted & new heat sync. It was running fine at 4.0ghz 77c. After two days it's back to 90c running only 3.6ghz. I don't know what to do does anyone have any suggestions on why this happened ?

Sent from my Phone using Tapatalk

 

What did you repaste with? Thermal paste or LM?

 

I used this paste here as I'm not really comfortable with LM. I will use it if it fixes the issue I'll get an engineer to apply it.

Sent from my Phone using Tapatalk

 

1. Then I cannot do anything about that and I have to live with that?
2. Windows high performance is selected and I also changed the hidden settings in high performance profile according to the guide in another thread. So where is the bias coming from?
3. I think the clocks are a bit lower. In addition when I test (CPU load) an OC profile in ACC, I get much lower maximum CPU frequencies (especially on high frequencies). But when not in load, core frequencies can individually reach the frequency selected.

 

1. You can make the most with what you have.

2. If you did that, you need to disable the use of the win power plan and use the speedshift values in TS to manage the settings (0 means max perf, higher values are towards more power savings). It is possible that the ACC has its own settings for power management and maybe it overrides the win profiles.

3 The clocks are a bit lower with what settings in what bench? If CB what are the scores for each consecutive run?

When it is not in load, it doesn't matter what frequency you can reach, also the voltage changes with the cpu load, that is why I mentioned that you should be interested in the load voltage.

You should use the settings that use the lowest voltage in load (or the highest frequency for the same power draw), but as I've said, your VID is not the same as vcore, so you need to find the lowest stable voltage (that doesn't crash) with TS.

 

Well of course, that is the idea behind the experiment! I reduced temps by 10C-14C and gained 20W headroom vs my previous setup which was Arctic MX-4, which in turn was already much better than Dell's original toothpaste and rubber stamps! I believe I could improve further if I replaced the (thick) thermal pads for something more appropriate (ie thinner pads). Maybe on the next try

I did not use LM. No fear of leaking or spilling and which way to rest the laptop in case it leaks, and it will not scar/corrode any surfaces. Maybe not quite as good as captn.ko's system, but he is already one in a thousand.

This is what I used:

http://uk.farnell.com/webapp/wcs/st...estType=Base&partNumber=2723038&storeId=10151

No need to use (dangerous) LM in my opinion. And unlike thermal grease I think this lasts better, it does not dry out and become brittle. Time will tell. I always have the Arctic MX-4 and half a dozen greases in case

 

I'm glad that you have better results with the graphite sheet! I don't know how long it will last. I have a feeling that you'll eventually repaste with LM and change the pads.

LM is not dangerous if proper safety measures are used (nail polish/temperature resistant electrical tape and foam dams). LM won't corrode your HS.

 

This is anecdotal but i tried backdating to 1.1.6 and went from the 1420 i was getting in cb down to ~1350

After installing 1.2.1 again i am still stuck with ~1350

 

I got a 1396cb once during one run when laptop was cool (just switched on). When I repeated the tests i was then getting 1320-1340.

I think 1350cb is a perfectly decent score. If you are getting throttled then try to reduce the multipliers so the test runs without throttles, on my laptop this is achieved at 43x and 140mV undervolt. The throttle cuts the CPU performance too much and you end up with lower scores.

Frequency by itself is not a great indication or determinant of heat/temperatures, because heat depends on what you do (what instructions the CPU is running). Better to look at watts. if you want to see how the heatsinks perform compare watts vs temperature. With a badly fitted heatsink you will see few watts and high temps. With a perfect heatsink you will see high watts and low temps.

Prime 95 Small FFTs generates a lot of heat, my favourite test is Prime95 Small FFTs with 10 threads.

 

What I have learned from various re-pastes using thermal paste and LM is that this machine is unpredictable. If you want to use a thermal paste on this machine you have to use the best ones. I would recommend MX-4 (same as doofus is using) which seems to be better than thermal grizzly or phobya which are regarded to be the best. This is only my personal opinion from my own tests. But temps wont drop like if you use LM especially for the CPU. I would also recommend to check if the white pads are thicker than they should. But to find that out you have to check visually and see if you get any temps spikes in addition to high average temps.If you only get high average temps, this must be just due to the thermal paste itself.

 

Bios 1.2.1 changes are irreversible. So, there is no meaning to downgrade. By the way, bios 1.2.1 is really buggy. Even the performance settings can be lost with simple restart (not always but it does happen often so I have to check the power limits everytime to be sure). In addition, sometimes you see options disabled in performance (without being able to change them as you should) and you have to press restore defaults and then the option is re-enabled. I hope they will release a new bios to fix their bad changes (and bugs) which seem to be to avoid the thermal issues (I bet many machines were sent back because of the temps).

But the score difference you get is about the same with the score difference I should get. Getting (1350-1372) at X43 while it should go up to 1420 something.

1. When I find time to play with all that because of their mess.
2. For some reason, TS speedshift at 0 does not prevent drops to 800MHz at idle. So, clueless. If it is not the bios then I wonder what is causing the power saving bias.
3. I will again check this in more detail

 

Unless you fix it yourself

Sent from my SM-G960F using Tapatalk

 

But it does it on 4.3 now as well which I've had rock solid for nearly a year. I think it's that new cancer BIOS I installed

Sent from my SM-G960F using Tapatalk

 

Definitely don't do that. You'll want a pro to apply LM or you risk bricking your laptop

Sent from my SM-G960F using Tapatalk

 

Its no rocket science to apply lm, but you certainly need to know what to do. Watch/read tutorials till you feel safe enough, its not that hard. But ye if you have shaky hands - ask a professional.

 

I know how to apply it I'd use electrical tape. I just don't wanna use it coz taking it off is a pain in the ass. I did use thermal right paste bit as I said after two days went upto 90c. I'm not sure why this is could be that heat sync needs tighten up again? I also have AC in my room not sure if that affects anything.

Sent from my Phone using Tapatalk

 

Take of LM even with the isopropyl alcohol tissues provides in the package or using arctic thermal paste remover using you own tissues. They both work fine. Just be careful when you remove it, so the LM does not touch something you do not want. You can also use ear plugs with a bit of thermal paste remover on them to be more precise when removing. It does not take so long though. And yes, cover with 3M tape is the first step before doing anything with LM. Some also use nail polisher but I did not because 3M tape covers everything really well.

 

Graphite pads may be the answer. Clean and tidy. Should last for ever. Approaches LM temperatures with none of the contamination issues. No wipes, no cotton buds, no solvents and other cleaning products required, no fear of corrosion/discolouring of surfaces and no fear of spilling and killing your machine.

Here's is the datasheet:
http://www.farnell.com/datasheets/2...80.925703696.1532116838-1658495713.1532116838

Can it be subtle differences at the base clock from motherboard to motherboard? I think there is a way that the CPU modifies its clocks, I do not know exactly how, is it always an integer multiplier of base or is it fractionated and is the base rock solid fixed or does it fluctuate?

 

Hm, that might be the case, I'm on February's 1.2.4 (SKL1.2.3_KBL1.2.4 - 2/18/2018).

 

I have a graphite pad of 0.1mm. I have not used it and I do not know how to. I have not seen something like this before. Loool

 

Basically how I'm feeling rn lol I wanna buy it but idk what ones to buy or the thiccness.

Sent from my Phone using Tapatalk