980 Ti Asic quality - Makes no sense

Sprayingmango

Sprayingmango

[H]ard|Gawd
Joined
Jan 20, 2012
Messages
1,259
So I have a reference MSI 980 Ti and an EVGA SC+ 980 Ti with reference blower on it. The bone stock MSI card is as vanilla as it gets, yet it has an 89% ASIC Quality. The EVGA card is supposed to be a bit more premium but has a 69% ASIC score. :confused::confused:

What gives? Is EVGA binning crappier chips? Wouldn't you think it would be the other way around?
 
ASIC quality makes no sense to me and many others. I don't think it's any kind of proven science either. According to what GPU-Z says about my 780 it shouldn't be an overclocking beast, yet my EVGA GTX 780 @ 64% overclocks better than I've seen most other people manage to get theirs to stable.
 
ASIC % doesn't have a scientifically-supported correlation with overclocking potential.

However, no SKUs have any "binning" worth mentioning. 1300MHz with stock voltage is pretty easy.

Lower ASIC chips have higher stock voltage which is why they should easily make it into the cut.
 
I had an MSI Gaming with 63% and have a classified with 70%. there is no binning. Just silicon lottery and I lost twice.
 
Gotcha. Thanks for the responses. I've wondered how that score is even calculated?? Is GPU-z referencing it to some chart?
 
friend'scatdied said:
ASIC % doesn't have a scientifically-supported correlation with overclocking potential..

Lower ASIC chips have higher stock voltage which is why they should easily make it into the cut.
Click to expand...

This is all there is needed to know about ASIC value.
 
friend'scatdied said:
ASIC % doesn't have a scientifically-supported correlation with overclocking potential.

However, no SKUs have any "binning" worth mentioning. 1300MHz with stock voltage is pretty easy.

Lower ASIC chips have higher stock voltage which is why they should easily make it into the cut.
Click to expand...

Yes, from what I've read recently on this, the ASIC quality is really just a pragmatic measure of how much voltage it'll take to maintain stock speeds, for whatever reason. That's it.

While you might sensibly expect a lower stock voltage to equate to higher overclock potential (largely the case for CPUs), it doesn't pan out for GPUs as they are so parallel in nature. E.g. When overclocking a quadcore cpu, you only need 4 cores to be good to maintain the speed. GPUs have like, 2,000-4,000 'cores' all of which need to maintain whatever speed you throw at it, so your chances are massively reduced. It's for this reason ASIC quality means nothing when overclocking. It should simply be ignored!

Regards to binning, I think possibly Asus's Matrix platinum range really does get 'binned chips', but the binning is done on measured overclock performance, not ASIC quality AFAIK.
 
You guys seem to forget that higher leakage parts (ones that require more voltage) are BETTER for getting really high clocks. Yeah, it's not intuative, but thats the truth.
 
extide said:
You guys seem to forget that higher leakage parts (ones that require more voltage) are BETTER for getting really high clocks. Yeah, it's not intuative, but thats the truth.
Click to expand...
Typically yes, and this holds more true as you decrease load temperature (e.g. high-leakage being preferred for LN2 and subzero suicide runs).

ASIC % is allegedly some approximation of chip leakage but there hasn't been any evidence or professional citations to even support how accurate or correlated it might be to the chip's actual leakage levels (someone should ask an Nvidia engineer).

For now the real-world implications of ASIC % in practice are mostly limited to determining the load voltage of the card. Much of the other benefits/consequences are conjecture with limited support and a fair amount of contradiction.
 
Last edited:
extide said:
You guys seem to forget that higher leakage parts (ones that require more voltage) are BETTER for getting really high clocks. Yeah, it's not intuative, but thats the truth.
Click to expand...

That apparently only really holds true under ice, and even then, there's plenty of exceptions to the rule.

In the end, the value is irrelevant.
 
seems like stock cards score higher ASIC scores while premium or overclocked cards score lower...ASIC means as much to me as the Windows Experience Index score...it's just a fun thing to look at
 
I've got 68.1% on one card and 69.4% on the other. Haven't made any attempt to OC just yet. Suspect it'll net me almost nothing as I'm likely bound by CPU speed and PCIe gen/lanes long before the GPUs.
 
It's not a reliable indicator of overclock. All it really tells you is it needs a bit more voltage, but nothing about how far the clock speed could go ultimately. If I understand correctly.
 
I've got 2 980 Ti's One has an ASIC of 83% and the other is 68%. I put the 83% card on top as the main and have both cards overclocked to 1450 for daily driving and if I wanna punch it for a few more FPS, I can move up to around 1490. Take ASIC with a grain of salt....The silicon lottery...It is what it is...
 
My 63% asic reference 980ti card overclocks and is more stable at 1500 core than my 980 ti MSI 6g card with a 78% asic....
 
People are marking up used units due to their ASIC scores, which is funny.
 
66.9% and mine will overclock to 1460mhz. I think ASIC is shit due to way too many variables.
 
Well if evga can charge a premium for cards with high ASIC scores then I'm thinking there is some validity to having a higher score and being a better overclocker
 
Sprayingmango said:
Well if evga can charge a premium for cards with high ASIC scores then I'm thinking there is some validity to having a higher score and being a better overclocker
Click to expand...

I think there's some validity to people seeing a higher number and thinking that automatically equates to being better. That's about it.
 
ASIC matters for Maxwell as I pointed out in this post.

But since you're probably too lazy to wade through that wall of text, I'll cut right to the chase:

Non-Maxwell cards: High ASIC = lower volts at default clocks, but also higher leakage = cannot overclock as far as low ASIC cards since voltage response not as good

Maxwell cards: Doesn't voltage scale worth shit until subzero, highest overclocking potential typically achieved around stock volts, thus higher ASIC = better
 
Mine is 70%, but because of my VBIOS I'm locked at 104% Power Limit, so I am at 1214 core / 1350 boost.
 
ASIC is just an indicator that it MIGHT be good.

A really, really high ASIC is likely something a bit special in the overclocking department.

That said, I have a regular 980 with a 69% that clocks like a beast on air. 1540 and 7950 stock cooler 24/7 gaming stable.

So it's NOT all about ASIC scores.
 
Last edited:
You must log in or register to reply here.
Back
Top