Overclock.net › Forums › Graphics Cards › NVIDIA › A question about ASIC quality, why is lower better for water?
New Posts  All Forums:Forum Nav:

A question about ASIC quality, why is lower better for water?

post #1 of 15
Thread Starter 
My issue is my 980 ti 74% ASIC does great at 1500core on air, about your average max OC.

My 290 is ASIC of 80% and its under water and gets a max OC of about 1160 core and 1600 vram, which is above avg for a 290, yet due to my high ASIC quality this actually shouldn't be happening, it should be a better OC on air according toe gpu-z ASIC definition.


Can anyone explain this to me, cause I call bs biggrin.gif
    
CPUGraphicsRAMHard Drive
i7-7820HK Kaby Lake @ 4.2Ghz GTX 1070 @ 2Ghz boost @ 9008Mhz Vram 16GB 2x8 DDR4 2133 Samsung 960 EVO 500GB NVMe 
Hard DriveCoolingOSMonitor
5TB Seagate 2.5" 5400RPM Grizzly Conductonaut on GPU and CPU Win 10 Bloatware Free Edition 17.3" 1080p IPS oc'd 100hz GSYNC 
  hide details  
Reply
    
CPUGraphicsRAMHard Drive
i7-7820HK Kaby Lake @ 4.2Ghz GTX 1070 @ 2Ghz boost @ 9008Mhz Vram 16GB 2x8 DDR4 2133 Samsung 960 EVO 500GB NVMe 
Hard DriveCoolingOSMonitor
5TB Seagate 2.5" 5400RPM Grizzly Conductonaut on GPU and CPU Win 10 Bloatware Free Edition 17.3" 1080p IPS oc'd 100hz GSYNC 
  hide details  
Reply
post #2 of 15
I wouldn't be surprised if Asic % has no correlation with potential overclocking.
post #3 of 15
Quote:
Originally Posted by caenlen View Post

My issue is my 980 ti 74% ASIC does great at 1500core on air, about your average max OC.

My 290 is ASIC of 80% and its under water and gets a max OC of about 1160 core and 1600 vram, which is above avg for a 290, yet due to my high ASIC quality this actually shouldn't be happening, it should be a better OC on air according toe gpu-z ASIC definition.


Can anyone explain this to me, cause I call bs biggrin.gif

There's a whole lot more in play, than just the ASIC quality of the silicon. All the ASIC quality does, is measure the voltage leak in the GPU core. The VRMs, voltage ripple, power supply, power limitations, pcb quality....there are literally a hundred different things that play into how well your card will be able to overclock. Only one little tiny part of that, is the ASIC quality.
post #4 of 15
Thread Starter 
ok, when you go to gpu-z and read ASIC quality, it specifically defines lower overclocking association with higher ASIC, and on air with higher the ASIC better overclocking, so I guess the GPU-z definition of it is wrong.

biggrin.gif that settles it then cheers mates.
    
CPUGraphicsRAMHard Drive
i7-7820HK Kaby Lake @ 4.2Ghz GTX 1070 @ 2Ghz boost @ 9008Mhz Vram 16GB 2x8 DDR4 2133 Samsung 960 EVO 500GB NVMe 
Hard DriveCoolingOSMonitor
5TB Seagate 2.5" 5400RPM Grizzly Conductonaut on GPU and CPU Win 10 Bloatware Free Edition 17.3" 1080p IPS oc'd 100hz GSYNC 
  hide details  
Reply
    
CPUGraphicsRAMHard Drive
i7-7820HK Kaby Lake @ 4.2Ghz GTX 1070 @ 2Ghz boost @ 9008Mhz Vram 16GB 2x8 DDR4 2133 Samsung 960 EVO 500GB NVMe 
Hard DriveCoolingOSMonitor
5TB Seagate 2.5" 5400RPM Grizzly Conductonaut on GPU and CPU Win 10 Bloatware Free Edition 17.3" 1080p IPS oc'd 100hz GSYNC 
  hide details  
Reply
post #5 of 15
Jay did a video on ASIC quality here: https://www.youtube.com/watch?v=S4sLDvMlfMs
post #6 of 15
Quote:
Originally Posted by Dimensive View Post

Jay did a video on ASIC quality here: https://www.youtube.com/watch?v=S4sLDvMlfMs

Dear lord no. Dont credit jay.
post #7 of 15
Thread Starter 
Quote:
Originally Posted by Dimensive View Post

Jay did a video on ASIC quality here: https://www.youtube.com/watch?v=S4sLDvMlfMs

plus rep, watched it on 1.5x speed, that video answered everything, lmao you are my hero - lets snugz biggrin.gif
    
CPUGraphicsRAMHard Drive
i7-7820HK Kaby Lake @ 4.2Ghz GTX 1070 @ 2Ghz boost @ 9008Mhz Vram 16GB 2x8 DDR4 2133 Samsung 960 EVO 500GB NVMe 
Hard DriveCoolingOSMonitor
5TB Seagate 2.5" 5400RPM Grizzly Conductonaut on GPU and CPU Win 10 Bloatware Free Edition 17.3" 1080p IPS oc'd 100hz GSYNC 
  hide details  
Reply
    
CPUGraphicsRAMHard Drive
i7-7820HK Kaby Lake @ 4.2Ghz GTX 1070 @ 2Ghz boost @ 9008Mhz Vram 16GB 2x8 DDR4 2133 Samsung 960 EVO 500GB NVMe 
Hard DriveCoolingOSMonitor
5TB Seagate 2.5" 5400RPM Grizzly Conductonaut on GPU and CPU Win 10 Bloatware Free Edition 17.3" 1080p IPS oc'd 100hz GSYNC 
  hide details  
Reply
post #8 of 15
ASIC works quite differently for Maxwell as Kingpin explains here. http://www.overclock.net/t/1565834/evga-com-evga-to-offer-ability-to-purchase-980ti-kinpin-cards-by-asic/120_30#post_24201543 (wall of text warning)
Warning: Spoiler! (Click to show)
Quote:
Right on..

Based on the comments in the thread, maybe best its best to let you guys know(who don't already) a little bit about what ASIC is and how it relates to this card or any other card for that matter past or present.
I'll hang out here a bit and check back often to answer any questions by OC.net'rs about the card, ASIC, pricing, air performance, ln2 performance, ANYTHING smile.gif.
I would rather respond with the correct information directly to someone, than see that person repeating the wrong information. They haven't started selling just yet AFIK..
Ill answer some of the questions I've seen asked already to get a start.

ASIC is not a new "measurement" , its been around for a very long time. Only just recently is it available to be seen through public utility though(gpuz). It represents a few measurements on a GPU, not just one as many think. Basically it measures the performance ability of a gpu at a given voltage. There is another measurement of leakage as well. These two values represent "ASIC" as you guys know it.
Without talking about numbers or percentages yet, higher ASIC quality means a GPU will require less voltage to run at say default specs. What does this mean roughly? It means that this GPU is using less voltage and is generating less heat per clock than a lower ASIC counterpart. HOWEVER it also means the voltage limit of what it can take on air as well as the voltage response getting weaker/noisy. Here comes the leakage part. Highest ASIC gpus have also have the highest leakage, low ASIC gpus have lowest leakage. The two values scale linearly. This is why the lowest ASIC cards are the ones that can take loads of voltage on air, and the response is good. Usually these low asic cards can OUT OVERCLOCK their higher asic counterparts because they end up scaling higher on clocks maxed out with the benefit of the voltage increase that the high ASIC cards cannot. They are stuck on lower volts because the leakage is already very high.

How does ASIC relate to air overclocking? Typically a higher ASIC gpu will overclock nicely on default voltage/air cooling, yielding highest overclocks WITHOUT any voltage increase. Less voltage/less heat. Lower ASIC gpu will need to use more voltage for a given clock as the higher ASIC one. Back in the Kepler days, this meant great air/water overclocking on our 680's and 780's. I remember posting a thread at evga explaining to people that they needed to use 1.4v+ (at dmm) to max out their classified gpus on air. God I miss those days smile.gif. Back then, the best gpus on air/water were the low asic ones, they could always oc/ov the highest. Times have changed, and this doesn't apply to maxwell however.

How does ASIC relate to Maxwell Air overclocking? With maxwell gpus the above definitions of ASIC do not apply Well you guys know maxwell 980,980ti, titanx have ambient cooling voltage limit on what v's you can give it on air/water. That's just the way it is. Its been proven over and over on every single manuf brand 980,titanx, 980ti. kp980 owners as well, no different. These gpus don't like voltage on air over 1.22-1.23v usually max. Just green garbage all over the screen with more, no better clocks. Best clocks usually achieved with stock voltage or maybe slightly higher.
So given what we know about ASIC quality and the voltage scaling capability of 980,980ti,t-x on air/water(NONE), it indicates the best gpus on 980Ti, will be the ones that can overclock the highest on default voltage or near default voltage. Ever noticed why almost every single review of 980ti (any brand) is around 1500mhz? The reviewer never can never add much voltage for better OC result.
I'm mostly an Ln2 person, but some users complaining about 980kpe not overclocking on air prompted me to try and make a better bios for ambient that would allow voltage. I managed a slight improvement that works on some cards, that's about it. But I learned a lot about the scaling of Maxwell on Air during that time and how we could if anything improve on this with KP980TI.

How does ASIC relate to LN2 overclocking and more specifically kp980ti?
As explained above, higher asic = higher leakage. Leakage is actually a good thing and can be contained on LN2 cooling. Cards with more leakage will run a bit hotter, usually extending cold limits on gpus and getting more core MHZ on LN2. Every overclocker wants every last mhz right? Higher ASIC GPUS also have better memory controllers and typically can overclock the memory very high on LN2. Lower ASIC gpus usually are not so good at driving the memory on Ln2 and the overclocker will lose a lot of MHZ when going cold.
Lastly, every serious overclocker knows the highest ASIC gpus use the least amount of voltage for any given clock. This means these gpus will always have the highest potential for scaling to the absolute highest clock on LN2, because most 980ti gpus max out around the same voltage level on the high end max max ln2 as well. Wouldn't you want the most clock you could get for that voltage smile.gif. KP card pushes them all the way. Unless its a lemon gpu (cant test every single one on ln2 lol), it will max out on this card.

Does high ASIC guarantee highest clocks on air? NO. The other part of ASIC which is Leakage is high on these, so that can actually hold back some high asic gpus on air. This doesn't mean its bad on Ln2 as well, and usually the contrary. I Tested around 15 pieces or so of KP980Ti these days, all different asic levels. Some as high as 81% all the way to 64% (which we wont even sell smile.gif the average clocks on air were roughly 1550mhz Lowest was 1526mhz, highest was 1592mhz . Seemed like every card went to 1539mhz or so smile.gif Most of the higher asic cards did as expected and hit the upper 1550's. None could pass 1600mhz, but some came really close! Those were mostly higher ASIC%'s. You should know that every kp980ti is binned gpu and even the minimum ASIC level for any card is very high compared to average. A 70-72% asic card is a great card. Reviewers should be hitting low to mid 1500's on avg and some cards hitting close to 1600. If your an air/water guy and don't plan to run ln2, I think no matter what asic level of card you get, it will do mid-low 1500's and there is a chance on all kp 980Tis regardless of asic to hit the magic 1600. Still need some luck too, leakage can limit this. For the hardcore users or the ones that may run ln2, I would think more seriously about asic and the time/money you can save buy getting something closer to what you want. Being an Ln2 overclocker myself, I feel this buy is mostly for you.

Does ASIC % guarantee highest clocks on LN2? No it doesn't, it is only an indicator of what to expect. EVGA is giving the chance for customers to zero in on exactly on what you want. Some users will try many cards to find the highest asics for best LN2 overclocking, its not a new thing. The highest ASIC gpus will almost always be the best ones on LN2 as explained above. These are the users we mostly are targeting with this. The ones that end up buying and trying lots of cards to find the one gem, almost always a high ASIC card. They will end up spending much more than the price difference of kp980ti high asic and wasting lots of time in the process. This is geared for them.

Are we binning gpus away from other cards to make this one (classy or other)?. LOL I wish, but no way that's possible or anyone would even let me do that haha. For sure there will be guys on classys or other cards who find a high asic gpu here or there and im sure they will let us all know they paid XXXX lower than what someone paid for their KP and still got a high asic.
KP980ti is very special card in many ways hw wise. This is just one special added buying option for our more hardcore users on first few batches directly from us, that's all smile.gif

Anything I didn't cover or you want to know something specifically about the card, I'm happy to answer.

But the short version is this:
Quote:
Highest ASIC gpus have also have the highest leakage, low ASIC gpus have lowest leakage. The two values scale linearly. This is why the lowest ASIC cards are the ones that can take loads of voltage on air, and the response is good. Usually these low asic cards can OUT OVERCLOCK their higher asic counterparts because they end up scaling higher on clocks maxed out with the benefit of the voltage increase that the high ASIC cards cannot. They are stuck on lower volts because the leakage is already very high.

This is why the 780 Ti Kingpin was actually binned for LOW ASIC, because the voltage response for Kepler is very good, and more volts = more overclock even on air. Whereas Maxwell doesn't voltage scale worth a damn until subzero, as Kingpin explained above, so everything is basically reversed.
post #9 of 15
Just watched the video....he essentially said the same thing I did. Except he doesn't give it as much credit as I think it deserves, especially with Maxwell and the restrictions that NVIDIA put on them. In some cases with voltage restrictions set at the VRM, can have a pretty big impact in the overclocking headroom of the card. Speaking pretty specifically about any of the cards that are locked in at 1.212v at the VRM (hello STRIX, I'm talking to you). Which is just barely workable, imo. /shrug
post #10 of 15
Is that a true hardware lock to 1.212V, or is the software just not reading Vddc correctly?

Asking because when I had a 980 Ti Classy, no matter how I flashed the bios or played around with the sliders, AB would not read over 1.212V. Yet when I whipped out my DMM and read the actual core voltage it corresponded to whatever value I had set in the bios. looniam has also observed the exact same thing I believe. Asking because the Strix uses an Asus rebranded controller, so it may or may not be reference. If non-ref then it would appear 980 Ti cards with non-ref voltage controller suffer from this 1.212V bug.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: NVIDIA
Overclock.net › Forums › Graphics Cards › NVIDIA › A question about ASIC quality, why is lower better for water?