HabitualRitual
New Member
- Total Posts : 14
- Reward points : 0
- Joined: 2020/09/04 16:46:58
- Status: offline
- Ribbons : 0
Hello. I am new to this forum. I have a RTX 2080 ti kingpin and I am curious about memory voltage, I know TiN's guide says stay below or don't go above 1.50 volts ambient on the memory with stock cooling. The guide doesn't mention the use of the mem load line switches on the back of the board itself, which I believe there is two of and they both add 25 mv each for fine-tuning and to help stability, it isn't expected to be used in a daily overclock however this doesn't mean it can't be used to help stability in a daily overclock. I am curious if I could set something like 1.50 volts on the memory or FBVDD using the classified tool and also have both of the mem load line switches on to increase stability? Would this be sort of a safe thing to do on a daily overclock? I am worried about it being 1.55 volts if the switches do indeed add + 25 mv each, so that would potentially set the voltage 0.50 mv higher? or am I interpreting the meaning wrong? I did switch them on and didn't notice a difference in the OLED readouts but I would like to make sure this is a fine thing to do without worry. - I know in the guide TiN mentions the 1.50 memory voltage limit and even states that 1.55 voltages on the memory could degrade the memory chips/ICs if used for over extended periods of time, e.g weeks or months, hence the question and concern.
"Do NOT set anything over 1200 mV for any type of near ambient cooling. Even with chilled watercooling and antifreeze your GPU temperatures under load will be positive and giving more juice to poor chip will only throttle it. It’s easy to see by lack of performance scaling (such as 3Dmark score / game fps)." - On stock cooling, or even a water block, does this mean the max limit is 1.20 volts for the GPU core or NVVDD? as in I can set it to 1.20 volts in the classified tool? Again I will want to run this daily. - How does the switches for the OFSG and GPU LL on the back of the board contribute to this 1.20 volt limit? and does the GPU LL switches interact in a way of sorts with the NVVDD load line in the classified tool? what would you consider the safety limit on the NVVDD voltage for ambient cooling, e.g stock cooling or custom water cooling? would you bother with the switches on the GPU for daily or what would you use it for? I want to run this kingpin long-term and safely while maximizing the overclocks without degradation issues. The normal bios and overclocking bios can only pull about 380 watts average, though the max watts the kingpin 2080 ti can pull is 520 watts, however the LN2 bios has no overcurrent protection and temperature limits, etc. if you want to run a hefty overclock and perhaps use that as a daily, would you be able to turn on the LN2 bios to achieve this? I know it's not recommended and could be considered dangerous according to TiN's guide, though I am still curious, in theory would this approach work for more heavy overclocks that wouldn't be possible on the normal/oc bios because of more safety limitations? I am wondering about some other options within the classified tool, PWM frequency, PEXVDD and +1.8 V, what exactly are these for and what would be a good approach to utilize these options? I know the PEXVDD is supposed to adjust the voltage for the PCIE lanes though I'm not sure what that does and how is it supposed to help with overclocking? I'm thinking it might help to keep the PCIE lane not stay frozen under LN2 or be too cold however I do not know and I am only speculating, would someone please clarify this for me? I have no idea what the +1.8 v means besides it's something to do with 1.8 voltages. PWM frequency is the switching frequency of the VRM? does this have any use in overclocking? as a daily or what is it used for? does it actually help stability of overclocks in any way and how? My main focus on this thread would be the memory overclocking related questions, especially the mem load line or mem ll question and max voltages, the second most important question is about the GPU voltages or NVVDD questions, everything else is what came to my mind when reading and wondering about how this RTX 2080 TI kingpin works.
|
Sajin
EVGA Forum Moderator
- Total Posts : 49131
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
Re: RTX 2080 ti kingpin question
2020/09/04 19:48:43
(permalink)
If you want to keep it safe keep it stock. 1.2v on the core and 1.5v on the men isn’t suppose to be used 24/7. Those are the max recommended voltages for the stock cooler when overclocking.
|
Nealx01
Superclocked Member
- Total Posts : 108
- Reward points : 0
- Joined: 2018/11/21 14:18:56
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/04 20:57:56
(permalink)
@sajin i told OP to post here and ask some of the other guys who have used the classified tool for longer than i have. he meant to post this down in the overclocking lab, maybe you can move it down there? we were hoping to get more advanced answers from folks who have pushed their cards. thank you👍
|
Sajin
EVGA Forum Moderator
- Total Posts : 49131
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
Re: RTX 2080 ti kingpin question
2020/09/04 21:21:48
(permalink)
If the op wants me to move it to the overclocking section I can.
|
HabitualRitual
New Member
- Total Posts : 14
- Reward points : 0
- Joined: 2020/09/04 16:46:58
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/05 07:24:16
(permalink)
Yes, please move this thread to the overclocking lab.
|
wimpievanzyl
New Member
- Total Posts : 21
- Reward points : 0
- Joined: 2019/07/01 05:06:46
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/07 01:51:57
(permalink)
I have tested the card in games running the 1.2v limit but only for the time the game was running. Will then revert back to stock voltage once done gaming. I did not see any noticeable difference between running at a clock speed of 2190 or 2205mHZ during game play or running at stock voltages at say 2130mhz. Although it is very interesting to be able to run at those clocks in game it is more of a niche thing to say that you can but is not needed to get more performance. The kingpin 2080ti card is already performing in-game better then a RTX titan card at stock so you are getting the best performing card in the 20 series you can buy already. The switches on the back of the card does not work with the classified tool. They will add fixed voltage to the card and if you use them in conjunction with the classified tool you will most probably over volt the card and damage something if you are not careful. I have used them but they make no noticeable difference on stock cooling. As the guide states it is more for using while running ln2 for stability reasons. When it comes to the memory overclock you will easily hit a +1400 overclock without doing anything on stock setting. Adding about +50mv with the classified tool will help me reach +1500 mhz overclock. The only issue is you will have to find the sweet spot where you still have good performance with such high memory overclocks while gaming because it can cause crashes while in game. I found running my card at about max +1100mhz for daily use in games have worked out pretty good plus it is already a huge overclock compared to normal 2080ti cards in the market. To sum up the switches on the back will not give you any noticeable performance differences and should rather be left for the day you want to bench the card on ln2 or DICE. Secondly rather avoid running the card for extended periods of time at anything from 1.2V and above on the stock cooling system. And lastly do not bother raising the memory voltage to high because at a certain point the memory IC`s stop scaling with voltage increases and you will start having negative returns. Go have a look at Gamers Nexus on youtube. He did extended testing with the card and you will learn a lot from those videos. Hope this helped a bit.
|
TiN_EE
Yes, that TiN
- Total Posts : 377
- Reward points : 0
- Joined: 2010/01/22 21:30:49
- Location: xDevs.com
- Status: offline
- Ribbons : 14
Re: RTX 2080 ti kingpin question
2020/09/07 10:22:33
(permalink)
Higher voltages on above-ambient cooling aren't going to help much, as you making heat issues even more prominent. It's way more helpful to keep GPU temps under +50 and memory temps under +60 for overclocking. Also none of the watercooling or aircooling is "heavy" in any meaning , as card will throttle a lot instead of running faster. You might get bigger numbers in MHz, but your performance will not increase, most likely decrease instead. :D And loadline switches are not adding +25mV to set voltages, they have completely different meaning and functionality.
If you have question, please post in public forum. I do not reply PMs, so all in community can benefit the answer.
|
HabitualRitual
New Member
- Total Posts : 14
- Reward points : 0
- Joined: 2020/09/04 16:46:58
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/07 14:38:47
(permalink)
I noticed that my temperatures never go above +50 degrees celsius for the GPU cores and the memory tends to cap out around 50 to 40 degrees celsius when on stock settings and also when the power limit and voltage are maxed out in EVGA precision, with these settings at stock configuration I can get +1200 mhz on the memory stable and +100 mhz on the core frequency stable. When I'm using the classified tool to adjust voltages, for example at 1.15 to 1.16850 set NVVDD for the core I can get +160 mhz stable on the core clock which actually improves my performance/fps in-game when I'm playing a game with maxed out graphic settings, quite significantly, the temperatures are fine as well, it's around 50 degrees celsius or even lower, nothing too concerning or high. I leave the NBVVDD load line at 0, though I haven't tried adjusting it much and seeing how that affects stability of the overclock. The memory voltage is a bit different, I can get it to actually boot in-game for a while to + 1400 mhz memory clock when the FBVDD is set at 1.50 volts when normally +1400 wouldn't even be remotely stable, so it adds a bit of stability and maybe even performance for a bit however all the settings with the frequency from 1300 to 1400 mhz or maybe less and the voltages from 1.450 to 1.50 FBVDD however visual artifacts and graphic glitches tend to happen after a prolonged period of time, so not much luck with the classified tool and changing memory voltage and frequency even though the temperatures stay around 60 degrees to 58 degrees celsius for the memory at these voltages and overclocks, however the memory definitely heats up a lot faster when the voltages are pushed but nothing too concerning. - My question is that if I stick to 1.20 volts on the core or lower, like what I stated earlier, e.g 1.15 to 1.16850 volts and 1.450 to 1.50 volts on the memory with the MEM load line at 2 (two switches turned on) with the temperatures I have with these settings, would this be fine to be used as a daily without running into degradation issues? - I would have to check if my memory voltage and frequency adjustments using the classified tool is causing thermal throttling, which I doubt it is at these provided temperatures however I think I should check it before I say it is or not, if it isn't being throttled then would the 1.50 volts on the FBVDD be fine along with the other settings I mentioned?
|
Sajin
EVGA Forum Moderator
- Total Posts : 49131
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
Re: RTX 2080 ti kingpin question
2020/09/07 15:30:39
(permalink)
1.093 is the max nvidia allows on the core. Anything above that could cause the core to degrade faster. Nvidia has even stated running the gpu at 1.093 24/7 could lead to the gpu only lasting a year... https://forums.evga.com/U...o-1-year-m2856901.aspx1.50 on the memory is only suppose to be used when benchmarking. If you want to run elevated voltages daily that is your choice.
|
HabitualRitual
New Member
- Total Posts : 14
- Reward points : 0
- Joined: 2020/09/04 16:46:58
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/07 17:33:37
(permalink)
The video you mentioned refers to a regular Nvidia card, and not the EVGA kingpin that is a custom card with a PCB designed for overclocking with cleaner power delivery and lots of high end VRM/mosfets to ensure better overclocking and life span, etc. The video is about 24/7 usage as well too, I believe, which is quite different from normal, regular everyday usage as a daily computer. Another thing that factors into life span is the temperature of the die/cores and memory. The turing cards like a regular RTX 2080 ti could run any where from 70 degrees celsius to around 90 degrees celsius which would also have a huge impact on life span compared to something like the RTX 2080 ti kingpin which runs around 50 degrees celsius because of the modified, hybrid cooling system, copper heatsinks, etc. It does so right out of the box. - Note the term "reliability" was used in the video instead of life span. The goal of this thread is to get people that used the classified tools for the RTX 2080 ti kingpin and get their results and information from messing around with the settings and translate that into how reliable and long the card is able to handle the custom settings, and at what settings those are, so we have an idea if this kingpin could be ran overvolted with the classified tool "safely". I am also curious what results people got with overclocking GDDR6 memory and the voltage tolerance of these type of memory chips. - The kingpin is a hand selected piece of hardware, something to keep in mind about it's ability to tolerate overclocks.
|
Sajin
EVGA Forum Moderator
- Total Posts : 49131
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
Re: RTX 2080 ti kingpin question
2020/09/07 17:49:43
(permalink)
The thing is people who buy this card understand that using elevated voltages is only recommended when overclocking. If you bought this card to run elevated voltages daily you’re doing it wrong.
|
TiN_EE
Yes, that TiN
- Total Posts : 377
- Reward points : 0
- Joined: 2010/01/22 21:30:49
- Location: xDevs.com
- Status: offline
- Ribbons : 14
Re: RTX 2080 ti kingpin question
2020/09/07 21:40:09
(permalink)
who buy this card understand that using elevated voltages is only recommended when overclocking People who buy ANY card and ANY hardware and doing overclocking, should already understand that elevated clocks/voltages are only for educational purposes only. It's ethically wrong to kill hardware by pushing hard and expect same warranties and free manufacturer replacements. You killed it, you own it If you bought this card to run elevated voltages daily you’re doing it wrong. There is nothing wrong to use card that was bought as owner please, it's not crApple PC world here, when one step to unapproved policy direction will be punished. That being said, seems like OP got everything already sorted out and have nice running GPU. Your temps seem to be alright, so if you confident about it, run it with higher volts. I'd expect card to work just fine for years to come, way past it's physically and morally obsolete by RTX 4080 or whatever. Obviously asking for warranties at such case is also unreasonable, nobody will guarantee what is outside specifications. Just keep in mind that different apps and games load system differently, and if its stable and you see perf increase in game X and Y, it does not mean that card is 100% stable in games/apps B, F and G as well. YMMV.
post edited by TiN_EE - 2020/09/07 21:44:39
If you have question, please post in public forum. I do not reply PMs, so all in community can benefit the answer.
|
Sajin
EVGA Forum Moderator
- Total Posts : 49131
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
Re: RTX 2080 ti kingpin question
2020/09/08 12:31:17
(permalink)
Well I guess that settles it then. Have fun running elevated voltages daily.
|
Zeddivile
SSC Member
- Total Posts : 761
- Reward points : 0
- Joined: 2017/09/30 11:15:15
- Location: @ the tipity top of FPS mountain
- Status: offline
- Ribbons : 7
Re: RTX 2080 ti kingpin question
2020/09/08 12:41:35
(permalink)
@TiN Yes if you bought it and killed it, you need to take accountability / ownership of the mistake.. That being said EVGA should be willing to make that assessment in the service center and not default to denying claims and intentionally delaying or ignoring email and RMA requests. For example, from the consumer perspective It doesn't really matter who broke it, why/how it is broken. Or who is financially responsible. Many enthusiasts may not be able to discern if they are at fault or are working with a faulty product... People just want their VGA's to work. If I send a VGA back to EVGA and the assessment is made that the owner is at fault, there should be a path to repair at the cost of repair payed by the consumer instead of an immediate rejection of a claim, Many manufacturers provide a path to repair regardless of fault. Fault being only the determination with regards to financial responsibility and not with regards to repair..... @OP On my KPE's anything above 1.45v @ambient on the VRAM and the cards run into stability issues.
post edited by Zeddivile - 2020/09/08 12:46:05
"This stuff breaks my tiny often dehydrated and carb deprived hamster brain" -2019
|
HabitualRitual
New Member
- Total Posts : 14
- Reward points : 0
- Joined: 2020/09/04 16:46:58
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/08 12:50:36
(permalink)
@Zeddivile, What about core voltage? I mean the goal of this thread is to ensure people don't kill or degrade their GPUs trying to overclock it because of lack of information out there on safe voltages or not enough information to make that decision but doing it anyways. I think that would save more cards in the long run. - Good to know that 1.45 volts on ambient might be the threshold of stability on ambient cooling for the memory.
@Sajin, are you not a fan of overclocking? You seem really against overclocking.. especially for a GPU, if you overclocked a CPU it's sort of the same concept with overclocking the kingpin.
@Tin, I will see if I can get some benchmark results that show scaling on ambient with increased voltages applied via classified tool vs EVGA precision overclocking and also a benchmark for stock with power limit and voltage slider maxed and one with all default settings and then I can measure the benchmark score and fps to get a percentage increase or decrease in performance, temperature will be monitored as well.
|
Zeddivile
SSC Member
- Total Posts : 761
- Reward points : 0
- Joined: 2017/09/30 11:15:15
- Location: @ the tipity top of FPS mountain
- Status: offline
- Ribbons : 7
Re: RTX 2080 ti kingpin question
2020/09/08 13:13:28
(permalink)
After a year of benching KPE's at ambient and sub ambient obviously can confirm TiN's guidance. Which by the way never needs to be confirmed as he is the source... For example while benching @ ambient GPU core @ >40C and vmem @ >50c 1.18v core yields better perf than 1.19 or 1.2v - Attributed to less thermal "noise" LLC plays a Significant role here. I run 0 LLC all the way to 1.1875v and VCP disabled For gaming and daily I see no reason to push more than 1.10v and the KPE's deliver plenty of perf @ 1.10. This is inline with what Sajin and NVIDIA state. (this statement does not appear to account for additional powerplant and cooling solution implementations on the KPE for instance...) For ambient and sub ambient benching the limit is 1.2v For XOC Bench w/ LN2 @ -120c and colder 1.45v + | This is were the line of delineation regarding voiding warranty becomes relevant that is unless you ran XOC BIOS but with lower voltages ect… Which would mean you have already voided warranty?
post edited by Zeddivile - 2020/09/08 13:20:12
"This stuff breaks my tiny often dehydrated and carb deprived hamster brain" -2019
|
Sajin
EVGA Forum Moderator
- Total Posts : 49131
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
Re: RTX 2080 ti kingpin question
2020/09/08 13:16:44
(permalink)
I’m not against overclocking at all. But I’m not going to run my gpu into the ground either by running elevated voltages daily. Do you think evga is going to rma your card for you after you tell them you constantly ran the card daily over 1.093mv? They won’t. They probably wouldn’t even rma the card if you told them you only ran the card once at 1.2v. If you push the card beyond safe voltages deemed by nvidia (1.093mv) and it breaks down evga shouldn’t be held liable for it just because you bought a card that was built for overclocking and you broke it by going above safe voltage levels.
|
Zeddivile
SSC Member
- Total Posts : 761
- Reward points : 0
- Joined: 2017/09/30 11:15:15
- Location: @ the tipity top of FPS mountain
- Status: offline
- Ribbons : 7
Re: RTX 2080 ti kingpin question
2020/09/08 13:24:35
(permalink)
Sajin I’m not against overclocking at all. But I’m not going to run my gpu into the ground either by running elevated voltages daily. Do you think evga is going to rma your card for you after you tell them you constantly ran the card daily over 1.093mv? They won’t. They probably wouldn’t even rma the card if you told them you only ran the card once at 1.2v. If you push the card beyond safe voltages deemed by nvidia (1.093mv) and it breaks down evga shouldn’t be held liable for it just because you bought a card that was built for overclocking and you broke it by going above safe voltage levels.
This statement is logical when considering stock power plant designs. Logic goes out the window when you consider the wildly overly engineered power plant and cooling solutions, not to mention the binned silicon that was intentionally binned to run 2195-2205 @ > 1.2v look at Silicon lottery for example, they sell you a binned and suggested overclocked voltage / clock rate ect and then warranty that chip for 1 year assuming you follow their guidance. I don't really see what EVGA is doing with the KPE' products as much different, they are heavily binning die's and then selling them at a marked up price...
post edited by Zeddivile - 2020/09/08 13:28:15
"This stuff breaks my tiny often dehydrated and carb deprived hamster brain" -2019
|
HabitualRitual
New Member
- Total Posts : 14
- Reward points : 0
- Joined: 2020/09/04 16:46:58
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/08 13:27:39
(permalink)
@Sajin, I understand the bit about the RMA part. I'm not going to RMA it if I voided the warranty or otherwise ruined it by myself on my own doing.
|
Sajin
EVGA Forum Moderator
- Total Posts : 49131
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
Re: RTX 2080 ti kingpin question
2020/09/08 13:33:44
(permalink)
If the cards were meant to run at 1.2v daily without issues then evga should have just set them like that straight out of the box.
|
HabitualRitual
New Member
- Total Posts : 14
- Reward points : 0
- Joined: 2020/09/04 16:46:58
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/08 13:38:07
(permalink)
@Sajin, Good point, however note that its possible for it to start thermal throttling at those voltages with ambient temperatures, which are highly variant from region to region by the way, or even house to house, hence why TiN mentioned setting voltages too high can cause throttling and negative scaling on the performance itself, however my temperatures seemed reasonable. Obviously you don't want a product that can thermal throttle right out of the box or that needs too much tinkering out of the box.
|
Zeddivile
SSC Member
- Total Posts : 761
- Reward points : 0
- Joined: 2017/09/30 11:15:15
- Location: @ the tipity top of FPS mountain
- Status: offline
- Ribbons : 7
Re: RTX 2080 ti kingpin question
2020/09/08 13:40:30
(permalink)
@OP also. I do not think there is a lack of info regarding safe voltages. https://xdevs.com/guide/2080ti_kpe/ The only thing that might not be spelled out is what voids warranty. From what I've gathered it seems pretty simple really: 1. Anything EVGA decides is a valid reason 2. Any modification period 3. Any Vasoline anywhere on VGA or MB 4. Like Sajin stated if you applied any voltage above NVIDIA guidance 5. If you flashed XOC BIOS 6. If an EVGA tech does not like you as a customer/person - IMHO
"This stuff breaks my tiny often dehydrated and carb deprived hamster brain" -2019
|
Zeddivile
SSC Member
- Total Posts : 761
- Reward points : 0
- Joined: 2017/09/30 11:15:15
- Location: @ the tipity top of FPS mountain
- Status: offline
- Ribbons : 7
Re: RTX 2080 ti kingpin question
2020/09/08 13:46:58
(permalink)
@SAJIN what part of this statement does not compute?
"For gaming and daily I see no reason to push more than 1.10v and the KPE's deliver plenty of perf @ 1.10. This is inline with what Sajin and NVIDIA state. (this statement does not appear to account for additional powerplant and cooling solution implementations on the KPE for instance...)" or the next statement "For ambient and sub ambient benching the limit is 1.2v"
"This stuff breaks my tiny often dehydrated and carb deprived hamster brain" -2019
|
TiN_EE
Yes, that TiN
- Total Posts : 377
- Reward points : 0
- Joined: 2010/01/22 21:30:49
- Location: xDevs.com
- Status: offline
- Ribbons : 14
Re: RTX 2080 ti kingpin question
2020/09/08 13:52:52
(permalink)
Q: What maximum voltage is safe for daily use on KPE before warranty is void? A: Yes :) KPE warranty meant to be voided
If you have question, please post in public forum. I do not reply PMs, so all in community can benefit the answer.
|
Sajin
EVGA Forum Moderator
- Total Posts : 49131
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
Re: RTX 2080 ti kingpin question
2020/09/08 13:54:06
(permalink)
Remember evga doesn’t make the core of the chip. The core has limits which it can take. Overvolting anything will accelerate electromigration as it causes faster degradation of the electrical pathways. Just because your temps are fine doesn’t mean you aren’t causing damage. Cooler temps just slow the damaging process.
|
HabitualRitual
New Member
- Total Posts : 14
- Reward points : 0
- Joined: 2020/09/04 16:46:58
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/08 14:00:39
(permalink)
@Zeddivile, Perhaps I should of said lack of documentated consequences or not quite black and white aspect of following the voltage guidelines that are in TiNs guide, like what If you ran those max voltages at ambient for extended periods of time.
|
Zeddivile
SSC Member
- Total Posts : 761
- Reward points : 0
- Joined: 2017/09/30 11:15:15
- Location: @ the tipity top of FPS mountain
- Status: offline
- Ribbons : 7
Re: RTX 2080 ti kingpin question
2020/09/08 14:30:58
(permalink)
@HabitualRitual Ya I understand. Most of the EVGA lackies seem to know all these lesser known details about consequences ect. As a costumer for the entirety of EVGA's existence I didn't and still don't really know these details and dislike how EVGA customers are treated when asking these types of question on the forums. I just never had a reason to crawl around the forum until 2019, Was to busy gaming, and to EVGA's praise did not have an issue with any EVGA product for 19 years... It seems to me that there is a tough balance with suggested documentation, customer expectation ect. EVGA try's to walk a line were they do not want to dissuade from using the product as intended (XOC KPE) and reasonable ethical warranty claim resolution. I can't imagine its an easy walk, and I am sure there is historical presentence of unethical claims that have lead to EVGA's current behavior / SOP. It's unknown at the time of documentation what happens if you run the KPE under specific conditions for extended periods of time... Vince and TiN have their anecdotal evidence as do the rest of the KPE owners. Non of which may be relevant or correlated to your specific environmental conditions... That being said SAJIN is most likely correct when suggesting that 1.2v will cause accelerated degradation vs 1.1 volts over an extended period of time. What is not known is how that equates to lifetime degradation and core clock performance degradation. Also as the KPE silicon was binned to meet a performance range and not a specific voltage / clock, even in the same environmental conditions each KPE could degrade differently and non linear...
post edited by Zeddivile - 2020/09/08 14:49:15
"This stuff breaks my tiny often dehydrated and carb deprived hamster brain" -2019
|
TiN_EE
Yes, that TiN
- Total Posts : 377
- Reward points : 0
- Joined: 2010/01/22 21:30:49
- Location: xDevs.com
- Status: offline
- Ribbons : 14
Re: RTX 2080 ti kingpin question
2020/09/08 16:53:15
(permalink)
People write PhD papers and spend decades of time on research about complex semiconductor operation, so there is no universal answer, sadly. Even more so, as was already mentioned, its NV who makes that chip and have the actual silicon datasets. What we as users can figure out is based on observation and testing, but it's only a tiny wee bit of the true story. Real question is - does it really matter? It's like adding nitro to a sport car, and asking manufacturer service to tell you what is safe limit when they still can provide free parts replacements for your engine :D. Can't have both, something gonna give. If you worry sleepless about GPU having premature death, play it safe within vendor specifications. If you just want to have fun and joy, push it to the limits while it's relevant. Sometimes high stress outcomes are clearly visible (such as Core i7 980X LN2 runs, when cpu lost margin by hundreds of MHz just from one session over 1.9V), but in most cases your chip will overheat and crash before you run into degradation. Crashing is a way to safely terminate operation and stop heat generation. When me and Vince were doing VRM tuning on cards, while running at LN2, I could tell when card throttle or crashed even without seeing the monitor, just by the power behaviour... NV was very correct on rtx3080 video, when they say every microwatt of power into GPU is translated into performance. More power in, more performance out, given you can keep thermals under control and have enough switching energy in transistors (e.g. not clocking faster than transistor voltages reliably can flip). Btw, that is also one of the reasons why we see clocks dropping every time more billions of transistors added to a chip. TSMC and etc try to finetune that spot, when you get good clocks while keeping voltages safe and feature size smaller. Multi-$B price tag for fab should give a hint how easy that is.... In my books, it's better to try walk that thin line in push of performance, than giveup alltogether without even trying like all usual criminals As/Ms/Gi do, churning out one SKU after other with just prettier RGB LEDs and more insane looking fans...
post edited by TiN_EE - 2020/09/08 17:01:57
If you have question, please post in public forum. I do not reply PMs, so all in community can benefit the answer.
|
HabitualRitual
New Member
- Total Posts : 14
- Reward points : 0
- Joined: 2020/09/04 16:46:58
- Status: offline
- Ribbons : 0
Re: RTX 2080 ti kingpin question
2020/09/08 17:17:09
(permalink)
@TiN_EE and @Sajin, Here is some quick math and tests I did for the RTX 2080 ti kingpin. Without locking the voltages in classified but maxing out the voltage slider in evga precision It pulled 200 watts at 1.05 volts, = 190.5 current *amps (Within nominal voltage range of +0.65 to +1.09 VDC) When using classified tool to lock in voltage at 1.09375 (within nominal voltage range of 1.09 +VDC) pulling upwards of 203 watts = 185.6 current *amps When voltage is set at: 1.16875 volts (read around 1188 mv in EVGA precision) (Above nominal voltage range of 1.09 + VDC) max wattage pull 230 watts current = 197 amps to 200 amps So with my current overclock settings via classified tool, the GPU core is only getting about -4.9 current in amps less than using EVGA precision tool at 1.09375 volts (locked in with classified tool) to about + 6.5 current in amps at 1.16875 volts (locked in) and + 9.5 current in amps at 1.188 volts (locked in), with approximately +30 watts being pulled from base line What about EVGA precision without touching the voltage slider but maxing out the power limit? Pulled upwards of 203 watts at 1.0560 volts = 192.5 amps Lets say at default values without overclocks we were doing: 1.033 volts, pulling 175 watts (within nominal range of +0.65 to +1.09 VDC) that is a current (amps) of 169.5. The worst case scenario, stock settings will be pulling 30.5 amps (current) less than the max classified tool overclock setting, at 200 amps with 1.188 volts and 230 watts 200 - 169.5 = 30.5 = difference. "Beware, voltages over the nominal limits are not guaranteed by EVGA or NV. These can be unsafe for everyday use, even on watercooling!" This is pretty much the classic overclocking warning that tech companies give their customers in regards to overclocking advice. Tell me if that seems right to you I used a online calculator because I don't have an educational background in being an electrician or working with voltages, etc. I know the load isn't accurately represented for all 2080 ti kingpin users and therefore wattage pull isn't one hundred percent accurate but thats what my system is doing in a quick heaven bench runs and the comparisons between them. so the question is, is pulling between -.4.9 to 9.5 amps more than max load going to harm the GPU cores? and is 30.5 amps over the stock average going to do significant harm? Since the mod from EVGA seems really uneasy about the GPU cores not being able to handle more than Nvidia's built in limit even when TiN suggests otherwise. Note: I sent Buildzoid this equation for his validation or not, to see what he says about the math I did above and if he thinks that much amps are going to cause significant harm or should be of concern. I will update this post with his input once/if I receive it. Of course the numbers will change depending on the voltage, wattage pull and how heavy the load on the GPU is though that can be measured and adjusted for. - I may update this post with new load numbers and current difference as well later on. @Sajin, I think electromigration cannot be fully stopped anyways, however you are right in the sense that overclocking will accelerate this, but that sort of what overclocking is, trading off some life span for speed instead. - There still is the relevance of memory voltage, temperatures, overclocking and other factors that play into the overclocking potential of the card instead of Nvidia's turing cores for the rtx 2080 ti kingpin and the limitation of NV's cores itself. EDIT NOTE: I talked to Buildzoid and this is the answer. Buildzoid mentioned that he wouldn't push past 1.1 volts on gpu core on any 16/12nm nvidia card and memory is fine until at least 1.4 volts. I guess that says something because he likes to push high voltages for testing and for daily runs.
post edited by HabitualRitual - 2020/09/14 14:05:17
|
TiN_EE
Yes, that TiN
- Total Posts : 377
- Reward points : 0
- Joined: 2010/01/22 21:30:49
- Location: xDevs.com
- Status: offline
- Ribbons : 14
Re: RTX 2080 ti kingpin question
2020/09/08 19:07:52
(permalink)
Sadly, you missing the point that your calculations aren't quite representing anything, as they are averaged out heavily here. Just look at Image 27 and 28 in my guide. Also your readings quite low, showing that card is not even sweating, in whatever benchmark you ran.
If you have question, please post in public forum. I do not reply PMs, so all in community can benefit the answer.
|