Hi, I'm having some trouble, I installed EVGA Precision software, my card should be the reference one (Precision displays the correct reference frequencies), but while gaming, the gpu boost bring the core to 1250mhz, it should be 1178. Not a big problem, but for keeping the gpu under 75° I must force a steep fan curve, which sounds very loud. Any reason for this boost? Thanks!

Oh thanks, understood. But if I'd like to restore the defaul gpu boost clock, what can I do? Just for comparing performance. I think too it's better to customize fan curve, I'm from a gtx 660 FTW Signature 2, and this is really much noisier, at 75°C the fans must run at 2400+ rpm, I'd like to find the right balance of noise/cooling :)

Thanks a lot for your answers!

I think this is the first time I read a thread post where someone buys a high performance GPU and complains it clocks to high...
 
I know you're not really complaining.  It just made me chuckle a bit.  As stated above, spend some time dialing in your fan curve and enjoy...

jonkrmr
It has nothing to do with ASIC quality of the GPU on how high the GPU will boost. Max boost of the card is set in the BIOS on the card. The GPU will boost based on three factors. Power target, Voltage limit and Temp limit. The GPU will boost to maximum boost value set in BIOS until any one of or a combination of those factors are reached at which point the GPU throttles back on the boost until limit conditions are no longer maxed out.
 
The max boost level the card will do set in the BIOS of the card is not the max boost speed you see in an app like GPU-Z. That is why you are seeing your card boosting faster than the max boost you see rated for the card. This is a good thing and nothing to worry about.
 
If you increase what ever factor (Power target, Voltage limit, Temp limit) is causing the card to throttle, it will stay at max boost indefinately when under load.
 
ASIC quality comes in for overclocking and is another subject altogether on how much it really applies to how well a particular card will overclock.

+1
 
Also, if you want to lower the clock, use PX or AB to give a negative GPU clock offset.

OK, the manufacture might use the ASIC quality of the chip to determine the max boost set in the BIOS on that card, but I have have not seen any difference in the max boost clock set in a particular model card's BIOS between different cards in that model and I've look at a bunch of different card's BIOSs in a same model all with varying ASIC quality between them. I have seen different revisions of a particular model's BIOS where voltage, power limit, max boost level and fan curve have changed but suspect that was to make that particular model from the manufacture more stable (too many RMA's for stability at "stock" advertised speed). The manufactures really don't factory overclock the cards aggressively anyway unless you are getting a custom PCB version with beefed up VRMs on it and even then they aren't factory overclocked that high. They leave the extreme overclocking to you. That is part of the point, right? Also, no guarantees on how well a particular card will over clock.
 
The GPU has to have a physical max boost otherwise, if all of the three limiting factors are never met say like under LN2, the GPU would boost until it burned up internally. That is what is determined by the manufacture based on the general quality of GPUs used for that particular model and then set in that model's BIOS.
 
A card's particular ASIC quality might determine how hot the VRMs will run on the card which in turn would determine the max boost of the card depending on when it hits it's thermal temp limit. Lower quality ASIC means hotter running VRMs which equate to the card hitting its thermal limit sooner?
 
ASIC quality is more a determining factor of the general quality of the GPU itself. Higher ASIC means a better quality GPU die which usually means better overhead potential. All GPUs/CPUs produced on a silicon wafer do not come off that wafer with the same quality. It is called wafer yield. The ones off the wafer that are perfect quality will get binned for the highest speed/feature set, the lower quality ones get binned for slower speed/reduced feature set. This way the manufacture can maximize the use of the chips coming off the wafers in manufacturing with the least amount of wasted chips.

ManBearPig
I have 2 970 FTW+s with ASIC qualities of 72.8% and 79.6%.  They are the exact same cards with the exact same BIOS and where bought at the exact same time.  My 72.8% card has a boost clock of 1404 MHz while the 79.6% has a boost clock of 1430 MHz; 2 bins higher (1 bin = 13 MHz) from the higher ASIC.  When I had my 970 FTWs, they had ASIC qualities of 77.7% and 78.3% and both had boost clocks of 1430 MHz; same clocks as my new FTW+.
 
If you look here http://forums.evga.com/GTX-970-FTW-Users-Post-Your-ASIC-and-Boost-Clocks-m2261389.aspx you will see that the higher the ASIC quality is, the higher the stock boost clock is.  And that they are increasing in increments of 13 MHz.
 
If you have multiple cards, you can try this as well.  But it was nice in that thread with compiled information from multiple users all showing the same thing; the higher the ASIC, the higher your stock boost clock will be.

Actually, I've had the same experience...

jonkrmr
You have FTW+ cards that have a custom PCB with better quality, beefier VRM design and a custom BIOS that has higher power target and voltage limit than the reference PCB models. These cards are binned by the manufacture so it would not surprise me to see them come with custom tweaked BIOSs specific to that particular card in your hands. That is not neccassarily the case for the reference PCB cards. Those are not binned, have Nvidia reference VRM designs and the manufactures keep the BIOSs the same on those across that model run for stability reasons so all of them will reliably operate to the specified speeds for the model.  The rules that apply to the reference design PCB models do not apply to you :)

The 970 FTW was not a custom PCB though (used a 760 PCB), but that still doesn't matter as that isn't how GPU Boost 2.0 works (700 series and up).  And these cards are NOT binned by the manufacturer (EVGA); the first card to actually be binned by EVGA is the new 980Ti Kingpin.  The rules for GPU Boost 2.0 work the same for ALL cards, reference or not.  All cards from Nvidia have a preset base clock that all cards must meet.  GPU Boost 2.0 then will boost your card even more depending on your ASIC quality (given you aren't being limited by temperature).  Factory OCed cards run under the same rules, except their base clock is a little higher.  As you can see from the link I provided, every single card in there ran by the same rule; higher ASIC equals higher stock boost clock.
 
Can you show me examples where lower ASIC quality cards have higher stock boost clocks than cards with higher ASIC quality?

it's always better to do some research prior to writing, it's not that difficult.

I hate to necro a dead thread, but i have this card. Purchased from Best Buy for 220 bucks when it was on sell. I have precison x set to +215 on the core and +325 on the memory; highest stable boost clock I can get is around 1450mhz; it usually sits between 1399-1438mhz depending on the scene its rendering. In my Enthro Pro M case with just a rear 140mm, temp sits at 80C (default fan curve). The reference cooler is very quiet. ASIC quality is 65.5%
 
Just for reference on OC; Unigine valley score is 2580 (min 31.7 max 116.6) i5-6500/Asus Z170A/EVGA Supernova 550w G2