Has Nvidia "hidden" the ASIC ratings of the GTX 1000 series cards?

[defred34]

Just curious. ASIC was the definitive determinant in ultimate GPU Boost 2.0 speeds in the Maxwell GTX 900 Series. I'm pretty sure it still plays a role in determining the ultimate GPU Boost 3.0 speeds, but has Nvidia "hidden" this value from being read by programs such as GPU-Z?

 

[Toothless]

As far as i know W1z has been having issues getting the asic info.

 

[Vellinious]

Yes, ASIC will not be supported on Pascal.

 

[rtwjunkie]

Yes, ASIC will not be supported on Pascal.

Do we know this for fact? As late as last week W1z said he was still working on it.

 

[defred34]

Yes, ASIC will not be supported on Pascal.

Is because Nvidia hid this value, so that people don't know the true quality and whether they hit lottery on their GPU? I'm sure on the manufacturers side they should still know the ASIC score.

 

[EarthDog]

ASIC really wasn't a 'bible' for how a GPU overclocks. It would show a trend, but there were plenty of exceptions. ASIC quality for those outside of LN2 benchmarking is almost useless. Kind of like the batch numbers on CPUs... Back in the day, it actually meant something, but since around IB days on forward, it really was hit and miss. There are some links at Hwbot which show that (for CPUs).

 

[defred34]

ASIC quality does show a definitive trend when it comes to out of box GPU Boost 2.0 clock speeds though.

I know this for a fact.

 

[the54thvoid]

ASIC quality does show a definitive trend when it comes to out of box GPU Boost 2.0 clock speeds though.

I know this for a fact.

Meaningless statement unless backed up by empirical data. Vince (Kingpin) stated that all 980ti Kingpin cards would hit above 1500 (some maybe mid 1500's) regardless of ASIC. The higher ASIC cards were meant for LN2 (as @EarthDog) stated. The voltage intolerance of Maxwell meant the ASIC had little real life merit. Almost any extra voltage caused instability at water or air induced temps.

 

[EarthDog]

ASIC quality does show a definitive trend when it comes to out of box GPU Boost 2.0 clock speeds though.

I know this for a fact.

It does? Take a look at some reviews and see where the same card ends up boost wise.

ASIC has nothing to do with the boost clocks. Boost clocks are a function of temperature, and power used/load. There are boost tables in the bios of the card that basically state, if you are under xx temperature and xxx power with at least xx load, boost to xxxx clockspeed. The quality of the silicon really has very little to do with it.

As the54th mentioned/alluded to, nearly all 980Ti's would hit 1500Mhz. That, I would say is the peak of the bell curve, perhaps a bit more. Anything above 1525Mhz was above average, anything below 1475Mhz was below average. Anything in that 50 Mhz range would be what most consider normal. You could get a bit further with a bit more voltage and a modded bios (for power limit increases), but they petered out pretty early without significant modifications to it.

 

[rtwjunkie]

I have to agree with @EarthDog. I have cards with ASIC of 64% that overclock and boost as well as cards with 78%.

ASIC quality is overrated. Sure it makes a nice Gee Whiz thing to publicize if it's high, but that's it.

 

[cdawall]

I have to agree with @EarthDog. I have cards with ASIC of 64% that overclock and boost as well as cards with 78%.

ASIC quality is overrated. Sure it makes a nice Gee Whiz thing to publicize if it's high, but that's it.

People seem to forget lower ASIC quality means higher leakage which means higher overclocks when cold...

 

[defred34]

Guys, I am NOT talking about overclocking at all. I am specifically talking about the maximum out of box boost clock speeds (call it GPU Boost 2.0/3.0) of identical GPU models. Yes, temps and power are a limiting factor and can impact the observed boost clock speeds, but really it's not at all difficult ti determine your maximum out of box boost speed.

Read this article. Eleven EVGA Reference GTX 980 Ti. The higher the ASIC, the higher the observed out of box maximum boost clock speeds. http://ocaholic.ch/modules/smartsection/item.php?itemid=3946

Show me proof that two GPUs of the EXACT SAME MODEL, where the one with the higher ASIC has lower out of box maximum boost clock speeds.

Sometimes, this value may be different due to different BIOS versions, and sometimes I have seen a lower ASIC card get a slightly higher max voltage to compensate for the lower bin (but the higher voltage would only account for one extra bin)...but for all intents and purposes for the GTX 900 Maxwell series, for two identical GPUs, a higher ASIC meant a higher maximum out of box maximum GPU Boost 2.0 clockspeed. I would love to be proven otherwise.

 

[Toothless]

Are we only talking about the 900 series because my two 780s have a good asic gap and boost the same.

 

[Vellinious]

ASIC really wasn't a 'bible' for how a GPU overclocks. It would show a trend, but there were plenty of exceptions. ASIC quality for those outside of LN2 benchmarking is almost useless. Kind of like the batch numbers on CPUs... Back in the day, it actually meant something, but since around IB days on forward, it really was hit and miss. There are some links at Hwbot which show that (for CPUs).

That wasn't true with Maxwell.....ASIC was a VERY good indicator with Maxwell, on just how well a card would overclock. The various exceptions, like specific cards that were just poorly put together....the GM204 STRIX cards come to mind almost immediately.

 

[Vellinious]

Do we know this for fact? As late as last week W1z said he was still working on it.

I had seen a post a few weeks ago, where (whomever), said that Pascal wasn't supporting ASIC quality. The guy was supposed to be the GPUz creator or something.

Found the post on reddit:

https://www.reddit.com/r/nvidia/com..._holy_grail_100_asic/?st=iqcmdbz2&sh=6d08763c

 

[rtwjunkie]

I had seen a post a few weeks ago, where (whomever), said that Pascal wasn't supporting ASIC quality. The guy was supposed to be the GPUz creator or something.

Found the post on reddit:

https://www.reddit.com/r/nvidia/com..._holy_grail_100_asic/?st=iqcmdbz2&sh=6d08763c

That's our own W1zzard, the site owner. And notice what he says further down, which is the same thing he said here, that it WILL happen eventually, it's just difficult right now.

*And yes, he is the creator of GPU-z.

 

[defred34]

Are we only talking about the 900 series because my two 780s have a good asic gap and boost the same.

I don't know, do 700 series cards have GPU Boost 2.0 aside from the Maxwell based 750/Ti? I speak from my personal research, and so far I've had a GTX 750 SC and a GTX 960 SC and for both these cards compared out of the box max boost values and found them to correlate well with the ASIC rating of the card.

In theory, it could extend to overclocking as well, but overclocking results are all over the place and there are so many factors and for my specific card I never found much else to compare to, so I cannot comment on it. But I know for a fact I can take my single-fan, 6-pin GTX 960 SC to 1541MHz stable (unstable over time at 1550MHz) on stock voltage, and with voltage increases it could have gone higher.

 

[BiggieShady]

I had seen a post a few weeks ago, where (whomever), said that Pascal wasn't supporting ASIC quality. The guy was supposed to be the GPUz creator or something.

Found the post on reddit:

https://www.reddit.com/r/nvidia/com..._holy_grail_100_asic/?st=iqcmdbz2&sh=6d08763c

 

[truth teller]

The guy was supposed to be the GPUz creator or something.

lol, thats the funniest thing ive seen all week, for real
also, what @rtwjunkie said

 

[dorsetknob]

The guy was supposed to be the GPUz creator or something.

HE DID NOT LIE OR CLAIM FALSE CREDIT

Credit Due to @W1zzard its a well respected program and an "industry Standard"
Available for download on many sites

 

[Vellinious]

I wasn't making the assumption that he was....I just didn't remember his name, nor did I care to. /shrug

 

[EarthDog]

Guys, I am NOT talking about overclocking at all. I am specifically talking about the maximum out of box boost clock speeds (call it GPU Boost 2.0/3.0) of identical GPU models. Yes, temps and power are a limiting factor and can impact the observed boost clock speeds, but really it's not at all difficult ti determine your maximum out of box boost speed.

Read this article. Eleven EVGA Reference GTX 980 Ti. The higher the ASIC, the higher the observed out of box maximum boost clock speeds. http://ocaholic.ch/modules/smartsection/item.php?itemid=3946

Show me proof that two GPUs of the EXACT SAME MODEL, where the one with the higher ASIC has lower out of box maximum boost clock speeds.

Sometimes, this value may be different due to different BIOS versions, and sometimes I have seen a lower ASIC card get a slightly higher max voltage to compensate for the lower bin (but the higher voltage would only account for one extra bin)...but for all intents and purposes for the GTX 900 Maxwell series, for two identical GPUs, a higher ASIC meant a higher maximum out of box maximum GPU Boost 2.0 clockspeed. I would love to be proven otherwise.

That article is pretty ironic, particularly when you consider what a high ASIC is supposed to mean according to GPUz...(low air overclocks, high LN2 overclocks). Why would it boost higher if the ASIC was better at LN2?

 

[Vellinious]

That article is pretty ironic, particularly when you consider what a high ASIC is supposed to mean according to GPUz...(low air overclocks, high LN2 overclocks). Why would it boost higher if the ASIC was better at LN2?

"low air overclocks, high LN2 overclocks"

This was more true with Kepler, but not so much with Maxwell. Things changed with Maxwell, and they're continuing to change with Pascal. With the smaller processes responding less and less to additional voltage, the higher ASIC cards end up being better all around. I would bet that IF he gets found what he needs to, to display ASIC quality with Pascal, that we'll find the same thing there as we did with Maxwell. That the higher ASIC cards are going to be the better overclockers across the board.

Just a thought....