ASIC quality rubbish

[Kaapstad]

I have just looked at the ASIC quality of my GPUs using both GPUZ 6.2 and GPUZ 6.6. The strange thing is I got totally different results between the two versions of GPUZ and it was a big difference 20% - 30%.

I have never taken ASIC quality serious anyway and this is the nail in the coffin. How can I get totally different results between the two versions of GPUZ.

 

[Yaayuh!]

I find it inaccurate and pointless.
My GTX580 had a very low ASIC rating but overclocked a dream. Yet my friend's had a high rating and was rubbish.
I really don't understand what black magic it uses to dish out this ASIC propaganda.

 

[pgi947]

With newer versions supposedly comes better reading of ASIC quality.

Its a load of tosh tbh, proven by my 62.4% 7970 that benches in excess of 1300mhz...

 

[Kaapstad]

I find it inaccurate and pointless.
My GTX580 had a very low ASIC rating but overclocked a dream. Yet my friend's had a high rating and was rubbish.
I really don't understand what black magic it uses to dish out this ASIC propaganda.

+1

My worst clocking card has got the best ASIC score.

 

[DigitalPrime]

I have a 83% rating on my xfx 7970 black edition, I cant breach the 1200 mark that everyone seems to be doing. It wont go past 1130 without artifacting. So to me this rating means squat as ive seen people with really low ratings. yet get immense OCs.

 

[borley]

i have an ASIC on my 7970 of 199% but my core clock only goes to 27mhz

 

[drunkenmaster]

Are people really still doing this, the first day this appear, I said it was a complete joke, haven't run it, won't ever run it, know it will be based on some absurd formula and means absolutely nothing at all, no amount of tweaking will make it anything other than a guess.

Its as useless as making your gpu overclock stable for furmark..... which will only lead to using lower clocks than will be stable in every other usage possible. Both are a joke, super pi(it was at one stage semi relevant code, hasn't been for 5 + years), plenty of other benchmarks, humans are odd competitive people. Put a benchmark out, its importants doesn't matter as long as someone can beat someone else

 

[555BUK]

Every single one of my 7800 series cards has overclocked according to ASIC quality. I have owned two cards which overclock with complete stability above 1300MHz. One was a 7850 than reached 1360MHz stable (at 1.3v with an Accelero TTII cooler) and the other is my current 7870 than does 1325MHz stable (1.3v stock Windforce cooler). Both of these cards indicated 84-85% ASIC.

The worst card I owned was an MSI TF3 7850. It had an ASIC of 64% an struggled to run stable at 1050MHz, no matter the voltage. The remaning 3x 7850's I have owned (I build systems for other people), overclocked somewhere between 1150 to 1250MHz and had ASIC's between low 70's to high 70's.

Now, ASIC is not the only determining factor. If you have poor components on the PCB, poor power regs, perhaps a poor mobo or PSU, this may restrict a even a great GPU, but I believe higher ASIC "improves the chances" of obtaining a high overclock.

One thing that is CERTAIN is that AMD determines the default voltage supplied to the GPU by ASIC quality. This indicates that lower value ASIC's require more voltage to run at stock speeds. This would also indicate that higher ASIC's will have more headroom for overclocking due to cooler running temperatures and more voltage headroom. A 7850 GPU with 1.218v default volts can only be overvolted by a small percentage, but one with 1.050v has a much wider margin.

I consider myself in a good position to judge because I have tested multiple GPU's with multiple ASIC values within exactly the same system. It cannot be coincidence that all of my 7800 GPU's scale according to ASIC. If anyone can show me a 60-70% ASIC that can match either of my 84/85% 7850/7870's I will eat my words.

I also believe that ASIC does not apply to such an extent (if at all) for NVidia 6 series cards, mainly because all of my 6 series cards have indicated 100%. Perhaps GPU-Z simply reads these incorrectly, or maybe the reading is simply mute.

Remember that the GPU is not the only component which determines GPU overclocks. The PCB is critical in the same way that good or bad motherboards will affect CPU and memory overclocks. ASIC is measured by overall quality of the silicon wafer, and GPU location within that wafer (quality tends to improve towards the centre of the wafer). You can still get defects (areas of lower quality) within even the best wafers, so even GPU's taken from the centre of the purest wafer may not be the best overclockers in the world. They just have a better chance.

 

[Broken Hope]

I also believe that ASIC does not apply to such an extent (if at all) for NVidia 6 series cards, mainly because all of my 6 series cards have indicated 100%. Perhaps GPU-Z simply reads these incorrectly, or maybe the reading is simply mute.

This is fixed in newer versions of GPU-Z, 6 series cards no longer all read at 100%

 

[Gregster]

Asic quality is the biggest pile of poo I have seen and a waste of effort and time even clicking it.

 

[Twirlywoos]

ASIC quality doesn't determine how well a card can overclock, no matter what anyone says, that includes you 555BUK

 

[Tonester0011]

The only time i've checked asic was when i thread a while back asked for it.

 

[mikeo]

Totally meaningless value as far as I can see. Each of the last 4 releases of GPU-z has given a different value for my current card.

As far as I was aware. GPU-z does not do anything clever here. It just merely takes the voltage for your card and reads an ASIC value off from an internal table hard coded in GPU-z. Hence why the value "appears" to change between versions of GPU-z (IE. the table has been updated between releases). Quite happy to be proved wrong here of course.

 

[Locky]

The card with the highest ASIC quality is the coolest card, but not the best clocking, mine at all 71 to 83% and all do 1250mhz, the uber one does 1350 (depending on benchmark).

Wish they all clocked that well, but 1250 is fine and 1200mhz 24/7 is perfect.

 

[555BUK]

ASIC quality doesn't determine how well a card can overclock, no matter what anyone says, that includes you 555BUK

I say ASIC helps and I have yet to see proof otherwise.

Please explain why AMD set higher ASIC GPU's to run at lower default voltage if they are not like to be superior? Surely, chips capable of running the same speeds at lower voltages have two advantages:- i). they run cooler, ii). they have more headroom for voltage tweaks. Why would AMD run some GPU's at 1.050v and other at 1.218v unless there was a fundamental reason (ie lower ASIC's are less likely to be stable at stock speeds unless default volts are ramped up).

I am not saying that high ASIC guarantee a better overclock or that low ASIC guarantees a poor clocker. I do believe (through owning and testing many AMD cards to their limits) that high ASIC's give you a better likelihood for high clocks.

Once again, please show me a low ASIC 7800 card that can match the overclocks of my two highest ASIC cards? It is also strange that I found my lowest ASIC card to be by far the poorest clocker, even though it had one of the best cooling solutions (MSI TF).

 

[wazza300]

id say it plays a part to a certain degree,theres other things like gpu memory that plays a part in how well a gpu overclocks

just like cpu's some are pure voltage hogs yet run cooler,some use little voltage yet run hotter