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 20:07:17
(permalink)
@TiN I wanted to do more in-depth tests soon as I been planning shortly after I made the thread which will include the other information such as performance-voltage-throttling as we discussed earlier because of the scaling I seen with higher voltages on the KP on ambient. This has got me thinking though, maybe it is scaling because of the low load and therefore less throttle. "Peak power already going to 508W, and increasing voltage any higher will cause a power cap and lost performance, even though average RMS power would be reported just 360W. Ofcourse, if your application/game stressing GPU less than 3Dmark TimeSpy, you would have some more headroom before hitting power cap, but that cannot apply to every benchmark/app." Yes, you mentioned in the guide that the PX1 only shows the averages of the power draw and wattage pull, not the peak power or power limit to performance increase without obvious increase on the average reading. That's one of the reasons why I included the current for when the card is just stock with voltage and power slider maxed out. The previous post with the calculations are still sort of a work in progress because I know GPU load wasn't maxed out but I just wanted to illustrate the point that the average power consumption when overclocked isn't much difference than the normal power consumption of stock card, I wanted to figure out how much of a current draw difference there was overclocked vs stock and maybe have some input on that to see if there's only little difference, then is overvolting the core really that bad? or significant in terms of what the cores can handle in voltage tolerance? or maybe it would show that there was a huge average current change in overclocked and non-overclocked scenarios. For example, lets say if I run my card at stock with power limit maxed out and do a benchmark then I run my card overvolted to 1.20 on the core and the power limit is maxed and I do another benchmark, I was thinking maybe over a period of 30 minutes, or however long it takes to do the test or benchmark, if the current change is negligible or not much, then how does that impact the cores in terms of how much current it can handle? I wanted to sort of point this out and get input on it and compare to NV's stock average current draw for their 2080 TIs, so we could see stock 2080 ti NV's current draw vs 2080 ti kingpins overclocked current draw except that in my calculations I only used the rtx 2080 ti kingpin because I don't want to mess around taking graphics cards out of my system and back in. - Think, how much of a change there was? and how meaningful is it on the impact of the cores itself, is it bad like Sajin suggested? or is it more on the lighter side of matters? I do understand the illustrations you put in the guide. I definitely need to do more fine-tuning and adjusting with my overclocks and also increasing GPU load under stressful situations. I should of waited to make this post and put quality into it but I just wanted to get that out there.
post edited by HabitualRitual - 2020/09/09 04:42:21
|
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 22:53:31
(permalink)
"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
2020/09/09 04:12:49
(permalink)
@Zeddivile, That being said, is LLC for NVVDD less droop at zero or more droop at zero?
post edited by HabitualRitual - 2020/09/09 04:18:29
|
sparetimepc
CLASSIFIED Member
- Total Posts : 2413
- Reward points : 0
- Joined: 2009/11/13 09:24:09
- Location: Altona, IL
- Status: offline
- Ribbons : 27
Re: RTX 2080 ti kingpin question
2020/09/09 07:38:25
(permalink)
HabitualRitual
@Zeddivile, That being said, is LLC for NVVDD less droop at zero or more droop at zero?
The lower the loadline the less droop and more heat is generally created.
|
wmmills
CLASSIFIED ULTRA Member
- Total Posts : 5679
- Reward points : 0
- Joined: 2006/01/04 20:47:29
- Location: New Jersey
- Status: offline
- Ribbons : 40
Re: RTX 2080 ti kingpin question
2020/09/10 07:59:24
(permalink)
HabitualRitual
@TiN I wanted to do more in-depth tests soon as I been planning shortly after I made the thread which will include the other information such as performance-voltage-throttling as we discussed earlier because of the scaling I seen with higher voltages on the KP on ambient. This has got me thinking though, maybe it is scaling because of the low load and therefore less throttle.
"Peak power already going to 508W, and increasing voltage any higher will cause a power cap and lost performance, even though average RMS power would be reported just 360W. Ofcourse, if your application/game stressing GPU less than 3Dmark TimeSpy, you would have some more headroom before hitting power cap, but that cannot apply to every benchmark/app."
Yes, you mentioned in the guide that the PX1 only shows the averages of the power draw and wattage pull, not the peak power or power limit to performance increase without obvious increase on the average reading. That's one of the reasons why I included the current for when the card is just stock with voltage and power slider maxed out.
The previous post with the calculations are still sort of a work in progress because I know GPU load wasn't maxed out but I just wanted to illustrate the point that the average power consumption when overclocked isn't much difference than the normal power consumption of stock card, I wanted to figure out how much of a current draw difference there was overclocked vs stock and maybe have some input on that to see if there's only little difference, then is overvolting the core really that bad? or significant in terms of what the cores can handle in voltage tolerance? or maybe it would show that there was a huge average current change in overclocked and non-overclocked scenarios. For example, lets say if I run my card at stock with power limit maxed out and do a benchmark then I run my card overvolted to 1.20 on the core and the power limit is maxed and I do another benchmark, I was thinking maybe over a period of 30 minutes, or however long it takes to do the test or benchmark, if the current change is negligible or not much, then how does that impact the cores in terms of how much current it can handle? I wanted to sort of point this out and get input on it and compare to NV's stock average current draw for their 2080 TIs, so we could see stock 2080 ti NV's current draw vs 2080 ti kingpins overclocked current draw except that in my calculations I only used the rtx 2080 ti kingpin because I don't want to mess around taking graphics cards out of my system and back in. - Think, how much of a change there was? and how meaningful is it on the impact of the cores itself, is it bad like Sajin suggested? or is it more on the lighter side of matters?
I do understand the illustrations you put in the guide. I definitely need to do more fine-tuning and adjusting with my overclocks and also increasing GPU load under stressful situations. I should of waited to make this post and put quality into it but I just wanted to get that out there.
There may be a new way to get exact power wattage use and other parameters SOON:
MOBO: EVGA x299 Dark, CPU: I9 10900X, RAM: Patriot Viper RGB 3600 32gb, SSD: Samsung 860 EVO 1TB, M2: Samsung 970 EVO+ 1TB, PSU: CoolerMaster M2 1500, CPU HSF: EVGA 240 CLC HSF~ P/P EK Furious Vardar, G-CARD:EVGA RTX3090 FTW3 Ultra Gaming w/Hybrid kit and Noctua IPPC 3000 P/P, CASE: LIAN-LI PC-V2010B w/ Window mod, OS: Windows 10 Pro 64bit, MON: Alienware AW3821DW  New EVGA product? Register it NOW with this link: http://www.evga.com/register/default.asp?affiliatecode=4QFQRAMOII Help Our Vets From K-2! [link=https://strongholdfreedom
|
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
2020/09/16 10:49:35
(permalink)
Buildzoid replied about running voltages daily, which is far from 24/7, I'd say.
"How much extra current from base line or stock would overclock for rtx 2080 ti or rtx 2080 ti kingpin would you think is acceptable to run, i.e not too significant of negative impact that would cause fast damage or degradation? Under terms of using it daily like most casual overclockers."
Buildzoid - "I wouldn't run any 16/12nm Nvidia cards above 1.1V
and the memory is fine to at least 1.4V"
- He didn't address my question specifically about samsung memory, which I think is relevant because there is different types of memory in RTX 2080 Ti graphic cards, such as micron, samsung, etc.
'Do you think the samsung memory on the 2080 ti kingpin can tolerate something like 1.450 volts on the memory instead of 1.40 volts daily?
I understand that different rtx 2080 ti's use different memory such as micron, samsung, etc.
"Since GDDR6 memory is now provided by multiple manufacturers, with RTX 2080 Ti KINGPIN EVGA is going back to traditional Samsung K4Z80325BC-HC14 memory chips, in standard 11 GByte configuration. These chips are rated for 14.0 Gb/s data rate and designed for 0-85 °C temperature range running at nominal 1.35 VDDQ ±3%. Maximum stable memory frequency can scale a bit with memory voltage, especially when well cooled." '
Sure Buildzoid knows his voltages, current, switching frequencies, etc. However I am more inclined to believe TiN's input on this as he has the educational background on this matter and he played a major role in developing the RTX 2080 ti kingpin.
Nominal value of 1.350 (+ 3 percent) = 1.3905, so 1.3905 volts the samsung memory is rated to run at, right around the 1.380 to 1.40 volt ball park the memory on the rtx 2080 ti kingpin runs already. I assume it can go past these max values a bit.
@Tin
@Sajin
I am getting good results from running vcore locked in at 1.1 volts with LLC adjusted as well, and memory voltage at 1.450 with mem LLC on 2. This gets me +1250 mhz on the memory and +140 mhz on the core clock. I think I will continue to daily this, unless I change my mind and go for 1.50 mem volts if it scales and maybe enough volts to get +160 mhz on the core as I did before when I tested the overclocks. - I might continue to daily the for sure safe voltage of 1.1 volts and bump the memory voltages a bit past stock, probably from 1.40 to 1.450.
Note: stock voltage on the rtx 2080 ti kingpin is 1.40 volts on memory and 1.068 max on vcore, with 1.097 or so maxed out using the voltage control slider in EVGA precision, so 1.1 volts is about the same as max when not messing with the classified tool but it's locked in at that voltage instead and bumping up the memory from 1.40 volts to 1.450 volts doesn't seem like much to me however I could be wrong.
I did some tests on furmark with EVGA precision overclock, power and slider maxed out and +100 mhz on core and +1200 on memory and compared that to +140 mhz on core and +1250 mhz on the memory and it scaled positively even though I set 1.1 volts on vcore and 1.450 volts on memory (more volts than 1.40 default), I still need to test out 1.16850 volts on vcore for +160 mhz and 1.50 volts on memory for somewhere between, I think 1250 to 1300 mhz on the memory.
I am going to test these settings on furmark and other benchmarks like 3D mark to see if it scales positively, so far I haven't seen negative scaling even with increased voltages.
I'm thinking I should test absolute stock values and compare that to overclocks, for example run EVGA precision with voltage and power sliders maxed out, fans maxed with memory and core mhz at default so +0 mhz on both of them and see how those compare to the above overclocked settings I tested earlier.
|
Sajin
EVGA Forum Moderator
- Total Posts : 49168
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: online
- Ribbons : 199
Re: RTX 2080 ti kingpin question
2020/09/16 20:34:32
(permalink)
|