No more overclocking on Nvidia mobile GPUs

It is indicative of default voltage

 

Show me where Asic has significant indication of voltage affecting performance. Some of the best results I have seen posted have low Asic with no issues with voltage or performance. Its junk number.

There have been multiple threads every year discussing Asic and not one post has demonstrated that it is anything more than junk number.

 

By far. In some cases it's 10C+ drop but you have to be extremely careful with it. It's liquid metal which is why it's so efficient but that also makes it problematic if you aren't careful. It is largely composed of gallium so if you get it near any aluminum in any amount, it will destroy the aluminum.

See: Gallium Induced Structural Failure of an Aluminum Sheet:

Since this is off topic I'll put the rest in a spoiler for you.

Spoiler

The second issue is that it's metal so it will conduct electricity. In that video you can see how it beads up? Yeah those little beads on a flat surface will roll around and if you aren't careful you can fry your card, CPU, motherboard, etc by shorting it out without even realizing it. If you take your time and paint it on, then roll any extra beads towards the center it shouldn't be an issue (it's a good idea to paint the heatsink first because a little bit goes a very long way and you don't want extra for obvious reasons).

The third problem is removing it. First thing you will notice is that it permanently stains your heatsink with a greyish silver hue. Next, if it hasn't dried out, you have to remove it. It doesn't like to come off and those little beads will come up all over the place. I used qtips and gently rolled a rubbing alcohol qtip over the metal to avoid this and had a big bead stick to the qtip and as I was pulling it away from the machine, the bead fell off! I was extremely lucky. It landed on the SLI cable and I was able to get it off without incident. Because of that though, I doubt I'll use it again. Great performance but a risk.

Sent from my Nexus 6 using Tapatalk

 

Are you actual troll or just slow?

ASIC quality = leakage. Lower ASIC, less leakage. Higher ASIC, more leakage.

Default voltage is scaled based on ASIC/leakage to stay within TDP. Lower ASIC/leakage, higher voltage. Higher ASIC/leakage, lower voltage.

http://forum.beyond3d.com/showpost.php?p=1344008&postcount=29

I said ASIC quality is indicative of default voltage because GPU with specific ASIC value is assigned default voltage based on table.

http://forums.guru3d.com/showpost.php?s=f0d2213687f1af806557a1c0d189f2fc&p=4222314&postcount=131

I never said ASIC quality has correlation to performance or overclockability; that's all silicon lottery. Stop putting words in my mouth.

 

Im extremely certain that GTX 970M cards from say Clevo have all the same voltage and TDP, based on the vbios they make for their own chips. Likewise, Alienware run the soldered 970M/980M at lower voltage due to the design of their notebooks and may have better requirements for their chips to reduce heat output and power consumption, than say Clevo with MXM modules and their models

ASIC may affect leakage and heat output and overclock performance, but they all run at the same voltage thats been coded in the vbios. Core voltage doesnt fluctuate in each individual chip based on ASIC

 

From what I've seen in the past that number has been pretty meaningless. Users with lower ASIC end up clocking better than those with much higher ASIC at same voltage. I haven't seen any definitive correlation. I know in thoery what it means, but in practice, it's pretty meaningless.

 

My 980M's default is 1.0500V.

 

Ima say it again: ASIC quality is a parameter to adjust VID accordingly to stay within the narrow TDP range that Nvidia/AMD prints on the spec sheet. Otherwise, you have high leakage cards with higher actual TDP than low leakage cards at the same VID and the printed TDP number is meaningless.

It has no bearing on a GPU's overclocking potential. There are low ASIC cards that OC great and high ASIC ones that are dogs, and vice versa.

If Clevo or whoever chooses to deviate from reference and lock VID regardless of ASIC/leakage, then the higher ASIC GPUs will have higher TDP than the lower ASIC GPUs. Unless Clevo is using chips which all have around the same leakage.

But usually, there is some voltage variation like this:

 

@Cloudfire I did some more research and don't think Clevo necessarily locks VID. For example, NotebookCheck's P651SE 970M is 1.018V, while HTWingNut mentioned here that his P650SE 970M stock is 1V. vBIOS version is 84.04.1F.00.BD on both.

 

I don't remember how to have Nvidia Inspector auto overclock/overvolt at boot. Can someone link a tutorial?

Thanks

 

My ASIC is 68%.. Some OC FS scores... My 240W brick isn't proving to be enough for some reason with my CPU @ 4.3GHz.. Need to do some tweaking and I should beat HT's score

Stock Firestrike Score: http://www.3dmark.com/fs/4108223 (6709)
OCed Firestrike Score: http://www.3dmark.com/fs/4108447 (8511)

 

check the OP of the batcave, i included a lil tutorial for underclocking at boot, its the same for oc/ov

Sent from my Nexus 5 using Tapatalk

 

Have you tested with a kill-a-watt meter? Turn off all lights on the computer and monitor when you run benchmark tests. I must because a 240W power supply is too little when running processor and graphics overclocked. Also a standard hdd uses more watts than an SSD. Dell should delivered previous Alienware 17 models with 330W power supply. The true Alienware 17 models should not have been delivered with a flimsy and weak 240W Power supply.

 

Yup. I have never seen any post on any forum that has demonstrated ASIC to be more than meaningless. Has not been demonstrated to be indicative of performance or overclockability.

 

The only evidence of ASIC being important for me, is merely anecdotal. The only recent GPU to die on my within a month of purchase, had very low ASIC quality, and it wasn't overclocked. My current GPUs are 66.6% ASIC and they work great so far, even overclocked, with low temps.

 

It's not meaningless. It's just not correlated to performance or overclockability. Which you would already know if you even bothered to read and understand my posts. But then again, that's too much to expect of Zymphad.

 

Actually your response pretty much says it all, it's useless. It's meaningless for me and anyone else who only cares about performance/overclocking. I couldn't give a crap what NVidia or AMD wants it for, they do what they need to do and if it works, that's all I care. But I guess it's too hard for you to understand why for many including myself Asic is meaningless useless number and why your explanation for it is also of no worth/meaning. Spout all you want and quote as much technical jargon as you want, for my machine and what I use it for, it's meaningless. Not trolling, it's just meaningless, useless, worthless number.

 

OK if default voltage is so worthless, why we got so many people posting theirs here and everywhere?

I think people put more faith in actual facts than whatever it is that comes out of you.

Sorry I should stop feeding...

 

No, real response is, you should stop insulting and realize what you think is important is not important to all. You keep failing to read when I saw, Asic is useless, worthless and meaningless TO ME. ME ME ME. And clearly my opinion of Asic is supported by others in this thread, why you are latching on to me when others have responded with similar view of Asic is your own issue.

This is not a forum of engineers debating over facts and far as I can tell, you're quoting Wikipedia. People can put faith in whatever your spouting or what I am, not really up to you to judge what they will read and consider.

But would love to see how many people will care if their Asic is 60% or 80% if it has no indication of overclocking and this is a thread discussing overclocking. NVidia/AMD would have taken into account of worst case scenario for voltages when they engineered our chips.

 

I don't appreciate thread-crapping trolls with a history on this board, that's all. You still haven't apologized to @thegreatsquare for being the reason his useful post got deleted during your instigation earlier in this thread.

 

No need to. I haven't been name calling or throwing around accusations or insults. My opinion is my opinion. NBR is a forum for people to post their opinions. If some other posters want to throw around insults and take discussions way out of hand and cause a Moderator to step in and clean up, that's their problem, not mine.

Keep up the insults and moderator probably will have to step in again since you're taking this way too seriously.

Just because you're not happy with my opinion of Asic doesn't mean I'm thread crapping. You still haven't proven to me why I should care about Asic, even a little bit. I've been overclocking and gaming on laptops for over 10 years and Asic has never had an impact on my use of gaming/workstation laptops yet. And I have never seen an explanation of it or a post on this forum, or others yet that convince me that Asic is anything more than worthless to me.

 

I didn't know Dave Baumann and Unwinder work for Wikipedia. And Beyond3D? Who posts there, anyway?

Zymphad in a nutshell

 

You might want to keep your opinions to yourself then, or at least tone down the rhetoric two or three notches. I don't know if you've noticed a pattern yet, but good things never seem to happen when you let loose. And it's not just me.

 

What's the difference? You're quoting other people, what is Wikipedia? It's community collaborative maintained encyclopedia. Your use of those above forums amounts to the same. It's hearsay through other users, not "fact" or from personal knowledge as an engineer who has first hand experience with NVidia/AMD GPU boards.

Agreed at times my tone may be a little off, but do my best not to throw around insults, name calling or accusations. But I'm not going to keep opinions to myself on a public forum just because you may not like them when I'm responding to your opinion. I haven't seen a discussion of "facts" among engineers and academics yet.

 

I thought I understood a little about ASIC, but now I'm not so sure after reading a few posts in this thread now. If default stock voltage is often related to ASIC quality, with High ASIC = lower stock voltage for a given frequency, then how would this not related to overclocking ability. Wouldn't it follow that High ASIC quality would hit a higher overclock frequency for any given voltage then? In fact the blurb in GPUz says that Higher ASIC Quality = Higher Overclocking on Air. I can understand there might be some variation from sample to sample within that relationship, but on average wouldn't High ASIC Quality result in a likely higher stable overclock for any given voltage?

(ASIC Quality of my 670MX is 76.6%, it overclocks quite a bit better than an average sample for any given voltage. 76.6% seems like a high ASIC Quality to me too, it's not that often that I've seen ASICs in the 80's and above - not that I've looked into it extensively!).

 

I had a 675MX with 58% Asic (maybe lower) and I overclocked the core by 315 and memory by 550. If Asic was an indication of overclocking, 58% seems awfully low and I did not see on NBR overclocking threads where my results were correlative-ly low to my Asic score.

My current 980M is 68%, yet my overclocking to 200/400 with slight 12.5mv overvolt seems to be on par with others, for stable gaming for a few hours in Crysis 3. The voltage issue with 980M I found an explanation that I can understand.

I like Khenglish's explanation for 980M voltage issue, rather than Asic.

Spoiler

The chips you are referring to are high-side and low-side power FET combo chips. Usually one chip is just a low side-FET or a high-side FET, but Nvidia uses chips with a high-side and low-side combined on one chip. These are involved with voltage regulation, but they are not the VRM. The VRM is located on the back side of the card. Power FETs are responsible for powering the core, while the VRM controls the power FETs. All current to the core travels through the power FETs, which in the case of the 980m are 3 combo chips.

Now that that's cleared up...

The 980m has 3 power phases while the 680m, 780m, 880m, and 970m all have 2. On the 980m each phase has just one chip, while older cards and the 970m have 2 on each phase. This does mean that each chip must pass more current, but overall there is less strain on the voltage regulation. There are 2 reasons for this. One is there is a 3rd inductor. Inductors dump a lot of heat out onto the card because they are fairly resistive, and having a 3rd to split the current load reduces heat output significantly. Other reason why the 980m is better is because having 2 chips in parallel is not a purely positive thing. The 2nd chip on the phase means that the VRM must push and pull twice as much charge to turn the FETs in the chip on and off. You want to switch back and forth between high-side being on, and low-side being on, and vice-verse as fast as possible. The longer it takes means lower efficiency and more heat put out. The 2nd chip per phase slows down the switching speed.

So combining the positives of the 3rd inductor and faster switching speeds, the 980m seems to me to have better power delivery than the 970m or earlier cards.

 

The difference is that Beyond3D is a moderated forum comprised of many industry veterans, while anybody can post ish on Wikipedia. Even you, Zymphad.

I'm pretty sure that Unwinder has prolific first-hand experience with both AMD and Nvidia hardware, given that he's the developer of RivaTuner, MSI Afterburner, EVGA Precision, and RTSS. And Dave Baumann as well with AMD hardware since he's their technical marketing manager, basically the counterpart of Tom Petersen at Nvidia.

 

And Wikipedia is not moderated or contributed by industry veterans? lulz?

Either way, I'm more interested in people's experiences on a forum like NBR, as I can use google and use search functions on forums, or even try to wrap my head around some technical paper. But my personal experience, theory and user experience aren't always the same.

 

You'd have to correlate it with voltage used for overclock achieved though, because different models of laptop & even different units of the same model of laptop have different cooling capabilities, so absolute max overclocks will vary independantly from ASIC if temperatures are the limiting factor.

So, if you were at +315 at 1.05V, then that would be a poor overclocker for example, considering mine is +524 at 1.05V. If you got +315 at 0.925V stock voltage, then that would be as good as mine. Depends on the voltage.

 

Why? I did not alter voltages and I was not reading posts of people altering voltages on Clevo 675MX. Some with higher Asic had lower overclock. Temps? Yeah I would think temp would be a bigger issue, and that seems to be luck of the draw with the chip, not the Asic score of a silicon board.

Overclocking on the 980M seems to be the same deal to me. It's not Asic score of the board, luck of the draw with the chips you got. Some people can OC a lot on the core but can't with memeory, (memory controllers at fault?). Asic still hasn't provided me a good explanation for this other than binning, luck of the draw.

 

Ah, in that case, if you were at a stock 0.925V & got the +315, then that's as good as mine with a 76% ASIC, and your ASIC was only 58% you say. If that's the case then our particular example definitely does not prove there's a link between ASIC quality and overclockability for any given voltage used.

 

ASIC quality is binning. It's a number assigned at the fab based on electrical leakage, and then VID is tuned accordingly to fit TDP.

 

I always thought the following:

1) Low ASIC Quality = High Leakage = High Voltage Required for any given clock = for some reason better overclocking with Water or LN2.
2) High ASIC Quality = Low Leakage = Low Voltage Required for any given clock = better overclocking (on air), due to less voltage required to attain an overclock = less heat produced for any given overclock.

Not right?

So, a high ASIC Quality part could have a higher frequency for any given TDP because the voltage could be lower.

 

You got it backwards. Lemme refer you back to my previous post.

ASIC Quality has no correlation to overclockability.

 

Chip binning as in overclocking potential, couldn't care less about Asic binning.

 

Exactly. Electrical leakage is one parameter by which chips are binned.

 

OMFG, OVERCLOCKING. I was talking about overclocking potential, couldn't care less about your Asic binning.

How does ASIC have any impact on this discussion of overclocking and NVidia's policy on overclocking? Are users here having issues with ASIC causing thermal or systemic damage to their laptops?

 

And lemme repeat for the umpteenth time: ASIC quality has no correlation to overclockability

 

Why do you insist on talking about Asic when you yourself are admitting it has no bearing on overclocking and this is a discussion of overclocking?

Even Robbo99999 was only interested in how Asic relates to overclocking yet you keep talking about Asic binning when I'm talking about chip binning regarding overclocking potential.

You well know what I'm talking about and you're complaining earlier about me trashing a thread and trolling? C'mon, this is ridiculous.

 

Thanks, yep, I did remember reading that post of yours, but it was contradictory to my understanding. I'll look into it another time to see if I can find any solid sources. Have you got any definitive sources you could link? (A quick google search didn't give me anything definitive)

 

Wikipedia, or supposedly superior Beyond3D.

 

I'm not arguing. You are. And shouting at the top your lungs, too. Amusing.

You state ASIC Quality is meaningless. I explained what the number actually means with the disclaimer that it has no affect on overclockability, which I repeated many time. And then you decided to go off again.

If you don't like the direction this thread is going, you're welcome to leave at any time.

 

Both Zymphad and Octiceps have been warned about arguing and have ignored those warnings, so thread closed.

The discussion continues here: http://forum.notebookreview.com/threads/nvidia-clockblock-part-deux-vbios.771900/