ASIC quality reference

Perhaps have everyone post with the same format to keep it organized. It is much easier to read that way.

It would be nice to have a base as well, however, seeing the best scores obtainable would be more meaningful IMO in ratio to the ASIC's to see how much meaning the ASIC %'s really have to the overall results.

OP if you want how about updating the OP to use the format in which I've posted above? Including the Bolds etc...and for the ones that have posted already, they could always go and edit.

 

Interesting you said that, was thinking about it too. However, I undervolted my 6820HK CPU by 68mv and have all bins at 3.8GHz so nothing wild there. The mobo VRMs should have had some relief in this case. So, I am not too sure about this theory, there might be a cap on the total power draw as I had the GPU sucking as high as 170-175W of power in Firestrike at 1440Mhz before throttling occured. I also get a lot of power and thermal throttle on the CPU once I go with it above 3.8Ghz, and I have played with voltages a lot. I am afraid the whole system may have a cap closer to 240W dictated by the size of the ac adapter. The 330W one while being recognized by the BIOS and working fine otherwise, is not an officially supported adapter/accessory for our 15R2s. Additionally, while overclocked with high numbers like the 1440GHz GPU and 3.8Ghz CPU using the 240W adapter, the system on ocassion started to discharge the battery, a clear sign of not enough juice supplied by the 240w adapter. Now, on the 330W adapter this is not happening anymore but I get the throttling once I go extreme. So, in the end if no unfotunate bug with Prema's BIOS then it might be a main system's BIOS limitation/issue which in its current state (1.2.15) is unable to allow usage of more than 240W of power in the system but allows more power in a hybrid mode albeit in a buggy way (hybrid mode still throttles but perhaps only thermally?).

 

ASIC is a very misleading way of judging a GPU quality, you can only test the limits of one GPU by overclocking it.

970M prema 79.5% ASIC
1.2v 1485/1510 game stable


I would safely say that if your GPU is randomly throttling, it's a problem with your VRM heat. Back down the overvolt, or change your thermal pads to 17w/mk like iunlock did.

As for the CPU, check your power limit(s).

 

I've done some extensive testing last night and paid very close attention to the numbers via HWINFO, CPUID HW, XTU etc...

Here is my theory:

1. BIOS: Our ability to maximize the hardware in general is being crippled by the BIOS. ie..due to the lack of adjustability of particular voltage settings not present on the AW BIOS. Dell seems to have used a cookie cutter approach to applying generic settings that does no justice to the end user, especially with the existence of the silicon lottery.

2. Dell's Short Cuts: The stock toilet paper thermal pads on the VRM's do cripple the AW a lot, let alone the chewed bubble gum of the stock thermal paste. So if one has not repasted or have changed our their thermal pads, you're already being held back by a lot.

My liquid metal and 17.0mk/w thermal pads keep my temps at bay fairly well to where it is allowing me to rule out a good % of the heat factor. Ex. I have a ~20C difference from my peak CPU temps vs Stock CPU temps on stock thermal paste. Yes, for real.

3. Power Limits: Due to the inability to fully tweak the power limits like we can on a desktop, a lot of the throttling is occurring from the power throttle limit as it is hitting the TDP of 45W of our CPU's. So it's not necessarily a heat issue for me or for others who have properly repasted & repadded. This is where we are being crippled. Even upping the TDP and the max boost duration would be very risky given the laptop environment, as the cooling ability at the higher watts would not be sufficient and dangerous for our laptops. It is what it is...laptop = compact = has its limits.

4. PSU?: Yes and No. The 240W should be sufficient and in most cases it is. However, when pushing things to the extreme like we're all doing, it does require more juice. Let me explain... just because a PSU is rated at 240W for example, does not mean that it is efficient to be pushed to the max limit of the power brick. There is a general rule of thumb of keeping things under 70% of the maximum limit of the PSU. I've briefly and casually tested this theory myself with measuring the temp with an IR heat gun pointed at the PSU under varies conditions, and when it started to reach heavy loads beyond ~70% of the 240W capacity, the temps soared. Very interesting indeed...and I am looking forward to doing some more tests later with my friends 15R2 w/ the 180W PSU...I also have a 330W brick from my previous AW18, but no AW18 to test it with lol...

Therefore, even if you're using a 330W PSU on your AW, the juice will be there, however, there's something really funny happening with how that extra power is being managed and delivered to the system. I know, it's weird, but the BIOS is definitely crippling the whole operation, and perhaps even the mobo itself (integrity wise) could be a contributing factor as well.

6. Heat: Well this is obvious, but like I've said above, even with the proper repaste and repadding with the best of the best, I've been fortunate to be able to rule out heat as being the primary culprit compared to a stock machine on stock paste and thermal pads.

7. Silicon Lottery: We all know what this is and have learned to accept the nature of this madness lol... why can't we all have the golden chip right? - Well what's interesting is all the different voltage numbers across the same platform running the same multipliers on the CPU. This alone proves the inconsistency that we all start off with from the get go.

In conclusion, my theory is as follows:

BIOS -> Mobo Integrity -> Power Management and Delivery -> Handicapped by stock paste and pads -> Silicon Lottery -> PSU

OK time to sleep now...lol

Powered by: Quad Core Exynos + 6820HK

 

You cant simply jus tkeep overclocking as long if the cooling allows it and the chip can run stable on a set voltage.

The VRM's supply a certain watt to feed these chips. It feeds 100watt and some headroom of power to the 980M for example. WHen you cross that power limit it will throttle. Just as on desktop cards by the way. I have a MSI Geforce 970GTX desktop card which I can overclock till no end, until I hit that power limit. Because of that I cannot bread the 1600mhz barrier even though the chip quality is excellent, the thermals are great etc. The card just cant manage more juice because the VRM's cannot pull more from the system. Dont think that a 330watt powersupply can suddenly fix this. It needs to have the power circuitry inside the laptop as well to be able to distribute that amount of watt

I think there is a hard limit to prevent overloading the circuitry which has been verified for 240watt but not 330watt. There are only 4 VRM's for the GPU if I remember correctly. That is just enough to overlock it around 25 to 30% on average. If they followed a similar design to the reference boards.

 

With unlock bios you can break your current power limit, it's not VRM limited, it's power limit (watts).

The power supply (VRM) for the 970M board and 980M board is the same, as with the CPU 6700HQ and 6820HK (Unlocked).

What is weird is that I throttle at 1.2v 970M - pulling around 125w at average, 140w at max. 120w average is what you would see with 980M with very slight OV and moderate OC applied. So this might be a concern if you own a 980M model and want to push the performance.



There are 6 VRM chips for the GPU on my 970M model, so it should be OK for overclocking. VRMs either hit a temp limit (they derate at high temp) or a hard current limit (physical design limitation).

I'm pretty sure that my 970M (and @VICKYGAMEBOY 's 980M) that throttles at 72c is because of VRM overheating. Not hitting the hard current limit.
How can I be so sure you ask? Because I can sustain 140W for about 2-3 minutes at start, but as the cooling system heat up, it throttles down to P1, and bounces between P0 - P1 much more frequently afterwards. If I scale back the overvolt down to 1.15 or 1.125v, I can sustain P0 indefinitely. So it must be a VRM problem.

What causes this problem? I'm pretty sure it's because of the bad thermal pad Dell uses. Probably not the heatsink. The thermal pads are like wet toilet paper compressed together, not very impressed by the quality.

I would suggest if you are planning to repaste, get the 17w/mk thermal pads 0.5mm alongside to re-pad your system, they're expensive but can save you the headache of throttle. I believe @iunlock has the info on what size you need to buy and how to cut it.


If you plan to NOT overclock, then the stock pads would probably be ok. I think...

As for power supply, 240W is enough for GPU OC alone. With 6700HQ you need to undervolt to make sure that you don't overload the PSU and break it it up like @Papusan did with his AW17.

For 6820HK, I would highly suggest you get a 330w if planning to max out the bins or overclock + overvolt.



Having the card throttle is a good idea to save the 970/980M BGA model VRMs though, because in comparison, the 980M MXM handles VRM protection differently, it just straight up blackscreen/freeze screen on solid color if overvolt too high.

@Ashtrix with his AW17 + 980M and my (previous) AW18 + single 980M had this problem.


Maybe @Mr. Fox can comment on auxiliary component cooling, he did a 200w run on MXM 980M so there must be some special cooling involved with the VRMs.

 

I got a replacement too. I have no idea what my first one's score was but it had to be better than my current system....61.7%.

 

Just to clarify, I did not imply that you'd be able to keep on OC'ing. If you look at what I wrote;

" In conclusion, my theory is as follows:

BIOS -> Mobo Integrity -> Power Management and Delivery -> Handicapped by stock paste and pads -> Silicon Lottery -> PSU"

....I listed the PSU as last, not first. That list is to show what my theory is in chronological order, which clearly indicates that I wasn't referring to being able to keep OC'ing based on power. If that was the case, I'd list the PSU in the beginning...

I agree that there definitely exists a hard limit, but more so I think the main issue here is power management and delivery. There is something going on here and I really think it has to do with the integrity of the mobo... It's like trying to turbo a car with an engine made out of paper mache. It ain't going to happen and it'll blow up / melt in no time... I've officially fried 4 mobo's already from extreme OC'ing... I have two 17R3's and it's been a lot of fun...

Also, in all honestly there really isn't a need to OC to the extreme limits we're all trying to push the AW...it's more for the fun factor for me than anything....at least for me it plays games just fine when running at my daily 38x on the CPU and most of the time stock clocks on the GPU. When you're getting 100+ fps in games, gaining a few frames at the expense of more unnecessary heat isn't worth the trade off, especially when you wouldn't be able to notice the small gain in fps anyway.

It'd be interesting to see what types of fps gains people get from OC'ing....

Ex. When I OC my GPU to +135/+200/stock voltage, I gain about ~4-6 fps on average. Not bad for not really pushing the GPU, especially at stock voltage.

IMO due to the limitations of a laptop environment, it is ideal to try to reach your max all on stock voltage without adding any additional power. I'm a firm believer now, after my own experiences of bbq'ing several mobo's that the integrity of the AW mobo is not the greatest. Aside to the power delivery / management issue, there exists a mobo integrity issue as well.

 

Here you guys go....well actually I'll retrace it on the lined side of the index card so that you guys can have an accurate reference.

I cut these out and trace the pieces over the new thermal pads with a sharp point permanent marker. It makes it easy to cut.

All you need is the 2x3 (0.5mm) pads found here. This one piece will be enough and then some for your entire application.

Enjoy...

(I'll post the new outline soon...)



Powered by: Quad Core Exynos + 6820HK

 

I have 82,6% ASIC quality with my GTX 860m.

 

980m here ASIC quality = 69.1%

 

I thought you returned your AW?
I think it's a wise decision to initiate another return and get your money back. Pascal is right around the corner and you'll miss out on a lot of improvements if you keep your current machine.

 

Nope i've been working with DELL to get this sorted. Frankly i got this for blogging and a bit of gaming, while i could wait for the 1080 I could be waiting a while and i need a laptop now. Thanks all the same though.

 

@judal57 @DeeX @Papusan @rinneh @Mobius 1 @iunlock @orancanoren @suyash691
we need to create a thread regarding 4710,4720,6700HQ,6820HK undervolt results,

im loving my 6700hq, after a repaste with ICD7 (-175mv Core,Cache ) no sound at all, its like macbook pro running on idle, ambient 24c

 

the image is broken

 

this forum is broken.. refreshing page few times it works