Category 5
Superclocked Member
- Total Posts : 148
- Reward points : 0
- Joined: 2017/03/02 10:20:25
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/27 08:25:25
(permalink)
ipkha
Unlike CPUs, the GPU chips don't have extra units to short out by a winning process. That costs too much and wasted wafer space. Instead they test them for stable speeds and select a low common denominator and use that as the default clocks. Like CPUs, some chips will be does (very small percent) and some will be excellent overclockers (also small percent). The vast majority will be stable at or around the default clocks (more likely 50Mhz or so of headroom just in case).
i think this is pretty right on. Then, in a few months yields get better and the percentages sway toward the top. Still, no reason why nvidia wouldn't skim the chips from the top for their own cards if they are tested during production. Less chance of warranty liability.
|
velo-x
New Member
- Total Posts : 97
- Reward points : 0
- Joined: 2016/06/29 14:15:03
- Status: offline
- Ribbons : 1
Re: 1080 TI FE Successful stable overclocks
2017/04/27 08:33:58
(permalink)
My 1080ti FE on reference cooler does 175 core 12ghz mem on superposition benchmark and 150 core 450 mem in BF1. 90-100% fan speed, voltage to the max. It brings me around 2075-2012mhz range. I reckon putting this baby under water will give me 2100mhz on the core. Will see.
MAXIMUS 7 HERO / i7 4790K - H110i GT / 16GB DDR3 / EVGA 1080 TI Fe/ Xonar Phoebus / XFX 850w Gold+ / WD Black 5 TB / Evolv Glass / Dell u2713HM 
|
nvidyuh
Superclocked Member
- Total Posts : 106
- Reward points : 0
- Joined: 2017/03/10 08:38:45
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/27 15:21:26
(permalink)
I hope my FTW3 gets clocks with a cool temp. Don't want a flaming GPU. 
|
demon09
FTW Member
- Total Posts : 1334
- Reward points : 0
- Joined: 2016/09/16 21:18:42
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/27 15:30:03
(permalink)
velo-x
My 1080ti FE on reference cooler does 175 core 12ghz mem on superposition benchmark and 150 core 450 mem in BF1. 90-100% fan speed, voltage to the max. It brings me around 2075-2012mhz range. I reckon putting this baby under water will give me 2100mhz on the core. Will see.
:0 nice silicone lottery pull
|
nvidyuh
Superclocked Member
- Total Posts : 106
- Reward points : 0
- Joined: 2017/03/10 08:38:45
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/27 15:31:23
(permalink)
Bretty gud. 
|
Poofu1
Superclocked Member
- Total Posts : 141
- Reward points : 0
- Joined: 2017/04/27 14:16:47
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/27 15:41:12
(permalink)
i'll e able to tell you once i get mine lol
|
cmoney408
Superclocked Member
- Total Posts : 150
- Reward points : 0
- Joined: 2014/12/09 19:17:00
- Location: San Jose, CA
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/27 16:11:40
(permalink)
DaBombDiggidy
can do 165+/400+ all day it seems on valley but getting crashing in games. backed it down to +100/+200 for now until i have more time to fiddle with it. was reading more about voltage curve overclocking yesterday and will probably do it that way next time.
did you every try the voltage curve method? i read the reddit post, but still seems a bit confusing to me, as far as tuning it.
|
bryanernster
New Member
- Total Posts : 23
- Reward points : 0
- Joined: 2016/07/03 23:10:26
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/27 23:26:13
(permalink)
Ok!!!!! what is everyone doing here becu im getting power throttle and so more power throttle............120% isn't nearly enought for 1700MHZ? UM i think those other folks are full of it......I under volt'd min to 1.025 via ctl+F and get it to about 2GHZ and u guessed it 120% power limit boo throttles coming my way
|
sethleigh
SSC Member
- Total Posts : 796
- Reward points : 0
- Joined: 2015/08/12 11:27:56
- Status: offline
- Ribbons : 4
Re: 1080 TI FE Successful stable overclocks
2017/04/28 00:32:27
(permalink)
Benchmarks: Fire Strike (21,866) - http://www.3dmark.com/fs/12471586Fire Strike Extreme (13,986) - http://www.3dmark.com/fs/12471846Fire Strike Ultra (7,284) - http://www.3dmark.com/fs/12471743Time Spy (10,084) - http://www.3dmark.com/spy/1651263 This is nearly a first look at how this performs on my machine. I just got the 1080ti FE today from the step-up program (starting with a 1080 SC upgraded to 1080 SC2, which I never even opened), and it's still running on air. My EK block shows up tomorrow (Amazon accidentally sent me an EK Titan X block, not the Titan X Pascal block I'd ordered, so I returned it and ordered the EK 1080ti block from PPCs), and I should have it in this weekend and see the real numbers when it's properly cooled. I have high hopes, as my 1080 SC on water consistently held a good 50-100 MHz over what it had done on air. The machine also has a 960 SSC in it, which is running my 24" 1920x1200 monitor, while the 1080ti runs the 28" 4K. I had Precision X OC running on the second monitor so I could look at the temperatures and speeds, so it's possible my scores might be very slightly higher if I didn't do that. I'm running an 8-core cpu, though, so I doubt it would have much of an effect. My cpu was running my day-to-day overclock of 4.2ghz, not a more aggressive overclock for testing purposes of 4.4ghz or so, which I simply don't ever really need. I didn't really tweak things much, and stopped and backed off a bit the first time a benchmark aborted. These scores were with +180gpu/+175 memory. Due to thermal limits on air and the changing loads during the benchmarks I saw anywhere from 1911 MHz to 2088 MHz gpu clock during these tests. I'll see if it stabilizes more and gives me a speed I can regard as what it will actually hold after I get the water block on it. ETA: Ran Unigine Heaven benchmark at 4K and 1440p on Ultra quality with Extreme tesselation. It aborted at the same +180gpu/+175mem I'd run the previous benchmarks on, so I changed it to +150gpu/+200mem and it ran just fine through the Heaven benchmark at both resolutions. 4K: FPS: 41.1 (21.5/97.6 min/max) Score: 1036 1440p: FPS: 91.7 (33.4/194.3 min/max) Score: 2311
post edited by sethleigh - 2017/04/28 01:14:02
Happy EVGA customer. Affiliate Code: 0Y7-1VU-ATW2 GigaByte X570 Aorus Master, AMD Ryzen 5900x under Optimus Foundation block, 32gb G.Skill DDR4 @ 3800 MHz 14-14-14-28, EVGA 3080ti FTW3 Ultra under Optimus block, 2TB 980 Pro SSD, EVGA Supernova G6 850W PS, ASUS 34" 3440x1440p 120Hz ultrawide, Lenovo 24" 1080p secondary monitor, Win 10
|
DRY_ICE
Superclocked Member
- Total Posts : 160
- Reward points : 0
- Joined: 2015/06/02 06:01:14
- Location: Chicago
- Status: offline
- Ribbons : 1
Re: 1080 TI FE Successful stable overclocks
2017/04/28 10:18:11
(permalink)
Only able to get a 2025 max stable on core regardless of +0 to +500 on memory. But with EKWB temps rarely near 50C so I guess I've got that goin for me! Still a great card even though my chip seems to be on the lower end of OC ability.
|
Jszupiany
Superclocked Member
- Total Posts : 103
- Reward points : 0
- Joined: 2004/01/22 17:32:29
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/28 10:27:13
(permalink)
It's funny how overclocks almost become a 1-up game. People practically willing to set their cards on fire for that extra +5 bump.
|
sethleigh
SSC Member
- Total Posts : 796
- Reward points : 0
- Joined: 2015/08/12 11:27:56
- Status: offline
- Ribbons : 4
Re: 1080 TI FE Successful stable overclocks
2017/04/29 02:06:26
(permalink)
DRY_ICE
Only able to get a 2025 max stable on core regardless of +0 to +500 on memory. But with EKWB temps rarely near 50C so I guess I've got that goin for me! Still a great card even though my chip seems to be on the lower end of OC ability. What program are you using to determine the 2025 Mhz stable speed? I ask because watching the frequency in the OSD during benchmarks it's always jumping around between different speeds. I'm not sure I'd call the fastest one I see for a short time in a benchmark the speed it runs at. On the other hand, with a +190/+225 setting right now (got my EKWB installed this evening) I'm running the Folding client right now and seeing it run steady at 2088 MHz. Seemed to run all my benchmarks with those settings too, but I'll probably not run this OC for regular folding. I don't think I want to push my card hard for 24/7 kinds of things like folding.
Happy EVGA customer. Affiliate Code: 0Y7-1VU-ATW2 GigaByte X570 Aorus Master, AMD Ryzen 5900x under Optimus Foundation block, 32gb G.Skill DDR4 @ 3800 MHz 14-14-14-28, EVGA 3080ti FTW3 Ultra under Optimus block, 2TB 980 Pro SSD, EVGA Supernova G6 850W PS, ASUS 34" 3440x1440p 120Hz ultrawide, Lenovo 24" 1080p secondary monitor, Win 10
|
Category 5
Superclocked Member
- Total Posts : 148
- Reward points : 0
- Joined: 2017/03/02 10:20:25
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/29 06:54:19
(permalink)
2088 is a beautiful OC for a 1080Ti! If you're running that without throttling I'd say you pretty much got a golden chip. I'd be happy if my second card could run 2000 but I've actually had crashes with +0 overclock.
It seems stable in games but heaven and 3dmark stress test sometimes fail. I wonder if some of the memory isn't contacting the cooler or something. I'd RMA it but I don't have the strength to send cards back and forth and be without it for two weeks again.
Works okay in most of my Games, and VR stuff, and oddly it's much cooler than my first card so not sure what gives.
Psu is an AX1200i so I don't think that's a factor. Furmark crashes right away even at full stock settings. It just seems to not want to be pushed. Maybe it's got one bad cuda core or something.
I also noticed that sometimes gpu1 is running at 99% and gpu 2 is hovering around 50-60 and I'm getting crazy stuttering. A reboot cures it for a while. I am using nvidia's high bandwidth bridge.
|
sethleigh
SSC Member
- Total Posts : 796
- Reward points : 0
- Joined: 2015/08/12 11:27:56
- Status: offline
- Ribbons : 4
Re: 1080 TI FE Successful stable overclocks
2017/04/29 07:09:05
(permalink)
I'm not sure I've seen 2088 Mhz during a game. That's running the folding client. I've seen a lot of 2025-2063 or so during various benchmarks. I haven't done much yet but make some settings and run the benchmarks. It's not really dialed in, and I need to go back and look at the power/voltage limits and whatever else and remind myself how it all works in Precision X OC. I haven't played with it in a few months, since I got the 1080 SC originally and played with it a lot.
Happy EVGA customer. Affiliate Code: 0Y7-1VU-ATW2 GigaByte X570 Aorus Master, AMD Ryzen 5900x under Optimus Foundation block, 32gb G.Skill DDR4 @ 3800 MHz 14-14-14-28, EVGA 3080ti FTW3 Ultra under Optimus block, 2TB 980 Pro SSD, EVGA Supernova G6 850W PS, ASUS 34" 3440x1440p 120Hz ultrawide, Lenovo 24" 1080p secondary monitor, Win 10
|
sethleigh
SSC Member
- Total Posts : 796
- Reward points : 0
- Joined: 2015/08/12 11:27:56
- Status: offline
- Ribbons : 4
Re: 1080 TI FE Successful stable overclocks
2017/04/29 07:12:22
(permalink)
Oh wow. I just adjusted the voltage limit up to 80% and watched the frequency hit 2100 Mhz while folding.
I'll have to play with this some more today in benchmarks and see what I can get with this further tweaking. If I recall from a few months ago, I think the frequency I considered to be "what this card can get" is the frequency it spent the most time in during the Heaven benchmark. I'll load it up shortly and see what happens.
Happy EVGA customer. Affiliate Code: 0Y7-1VU-ATW2 GigaByte X570 Aorus Master, AMD Ryzen 5900x under Optimus Foundation block, 32gb G.Skill DDR4 @ 3800 MHz 14-14-14-28, EVGA 3080ti FTW3 Ultra under Optimus block, 2TB 980 Pro SSD, EVGA Supernova G6 850W PS, ASUS 34" 3440x1440p 120Hz ultrawide, Lenovo 24" 1080p secondary monitor, Win 10
|
sethleigh
SSC Member
- Total Posts : 796
- Reward points : 0
- Joined: 2015/08/12 11:27:56
- Status: offline
- Ribbons : 4
Re: 1080 TI FE Successful stable overclocks
2017/04/29 08:13:55
(permalink)
Category 5
2088 is a beautiful OC for a 1080Ti! If you're running that without throttling I'd say you pretty much got a golden chip. I'd be happy if my second card could run 2000 but I've actually had crashes with +0 overclock.
It seems stable in games but heaven and 3dmark stress test sometimes fail. I wonder if some of the memory isn't contacting the cooler or something. I'd RMA it but I don't have the strength to send cards back and forth and be without it for two weeks again. I had to put my EKWB on multiple times and look carefully at it from the side with a flashlight to convince myself that all the memory chips were in fact making contact through their thermal pads. I think a couple of them aren't very compressed, but it's probably good enough. The gpu itself has good contact with the block, and I used the thermal compound that EK sent with it, which is fairly thin, so it probably squished out the excess nicely. I just tweaked the voltage slider up to 85% and ran Time Spy with gpu set to +190 and memory set to +225, and it was locked in at 2088 Mhz for most of those tests. Occasionally it would dip down to 2075 Mhz for a few seconds, and once I saw it actually hit 2100 Mhz even for a short time. I don't know if this is truly stable or not, so I'll run some more tests on it to see. So far in benchmarks I've seen the gpu temp get up to 44 C max, though while running the Folding client it got up to 46 C. During Folding, however, it was also folding with 7 out of the 8 cores of my cpu, which shares the same water cooling loop as the gpu, so that could have been artificially high. Time Spy - 10,506 http://www.3dmark.com/spy/1660937 (note this reports the gpu speed as 2101 Mhz, but that's because it hit 2100 Mhz at least once for at least a couple of seconds, not because it held that constantly: it generally was at 2088 Mhz) This is also with my cpu OCed to 4.2ghz, which is my everyday speed, not the stable 4.4ghz max OC I developed before for testing purposes.
Happy EVGA customer. Affiliate Code: 0Y7-1VU-ATW2 GigaByte X570 Aorus Master, AMD Ryzen 5900x under Optimus Foundation block, 32gb G.Skill DDR4 @ 3800 MHz 14-14-14-28, EVGA 3080ti FTW3 Ultra under Optimus block, 2TB 980 Pro SSD, EVGA Supernova G6 850W PS, ASUS 34" 3440x1440p 120Hz ultrawide, Lenovo 24" 1080p secondary monitor, Win 10
|
johnalan
iCX Member
- Total Posts : 272
- Reward points : 0
- Joined: 2011/02/25 18:55:23
- Location: Mississippi
- Status: offline
- Ribbons : 4
Re: 1080 TI FE Successful stable overclocks
2017/04/29 09:01:44
(permalink)
|
DRY_ICE
Superclocked Member
- Total Posts : 160
- Reward points : 0
- Joined: 2015/06/02 06:01:14
- Location: Chicago
- Status: offline
- Ribbons : 1
Re: 1080 TI FE Successful stable overclocks
2017/04/29 09:41:16
(permalink)
sethleigh
DRY_ICE
Only able to get a 2025 max stable on core regardless of +0 to +500 on memory. But with EKWB temps rarely near 50C so I guess I've got that goin for me! Still a great card even though my chip seems to be on the lower end of OC ability.
What program are you using to determine the 2025 Mhz stable speed? I ask because watching the frequency in the OSD during benchmarks it's always jumping around between different speeds. I'm not sure I'd call the fastest one I see for a short time in a benchmark the speed it runs at.
On the other hand, with a +190/+225 setting right now (got my EKWB installed this evening) I'm running the Folding client right now and seeing it run steady at 2088 MHz. Seemed to run all my benchmarks with those settings too, but I'll probably not run this OC for regular folding. I don't think I want to push my card hard for 24/7 kinds of things like folding.
I've used superposition benchmark for initial stability testing. It'll lock in the core clock on the osd (verified with afterburner) for me on each run and not fluctuate - temps don't go above the throttle threshold. I was able to get a pass I believe at 2036mhz core on one run but it's not stable during gaming or extended use. 2025 is the max stable I can get that'll pass benchmarks and extended gaming sessions.
|
sethleigh
SSC Member
- Total Posts : 796
- Reward points : 0
- Joined: 2015/08/12 11:27:56
- Status: offline
- Ribbons : 4
Re: 1080 TI FE Successful stable overclocks
2017/04/29 11:24:41
(permalink)
DRY_ICE
I've used superposition benchmark for initial stability testing. It'll lock in the core clock on the osd (verified with afterburner) for me on each run and not fluctuate - temps don't go above the throttle threshold. I was able to get a pass I believe at 2036mhz core on one run but it's not stable during gaming or extended use. 2025 is the max stable I can get that'll pass benchmarks and extended gaming sessions. I did some more testing, and in Time Spy and Fire Strike if my gpu hit 2100 it often crashed. In Fire Strike it hit 2100 fairly often, but then it eventually crashed. I dropped my settings enough to nudge it down to 2088 and it passed Time Spy and Fire Strike, but then it eventually would crash out in Heaven. The last crash was with the voltage at 90%, 2088 Mhz and no memory speed boost at all, so it's looking like 2088 just isn't rock stable. Didn't have the time yet to drop it to 2075 and retest, but I'll try it out today and see if that's going to be stable. I'll try Heaven again and the Superposition one. Once I'm sure how high the gpu will go I'll start over with memory boost. I noticed your graphics test in Fire Strike was significantly higher than mine even though my gpu was higher clocked than yours. You seemed to have a significant memory boost. In the little bit of memory boost I tried so far it seemed that some speeds increase the performance, and some speeds reduce it. I have no idea how to interpret that, but that's what I was seeing.
Happy EVGA customer. Affiliate Code: 0Y7-1VU-ATW2 GigaByte X570 Aorus Master, AMD Ryzen 5900x under Optimus Foundation block, 32gb G.Skill DDR4 @ 3800 MHz 14-14-14-28, EVGA 3080ti FTW3 Ultra under Optimus block, 2TB 980 Pro SSD, EVGA Supernova G6 850W PS, ASUS 34" 3440x1440p 120Hz ultrawide, Lenovo 24" 1080p secondary monitor, Win 10
|
DRY_ICE
Superclocked Member
- Total Posts : 160
- Reward points : 0
- Joined: 2015/06/02 06:01:14
- Location: Chicago
- Status: offline
- Ribbons : 1
Re: 1080 TI FE Successful stable overclocks
2017/04/29 12:08:06
(permalink)
sethleigh
I noticed your graphics test in Fire Strike was significantly higher than mine even though my gpu was higher clocked than yours. You seemed to have a significant memory boost. In the little bit of memory boost I tried so far it seemed that some speeds increase the performance, and some speeds reduce it. I have no idea how to interpret that, but that's what I was seeing.
Where did you find that? I'll have to take a look at it tonight when I get off work.
|
sethleigh
SSC Member
- Total Posts : 796
- Reward points : 0
- Joined: 2015/08/12 11:27:56
- Status: offline
- Ribbons : 4
Re: 1080 TI FE Successful stable overclocks
2017/04/29 13:17:58
(permalink)
DRY_ICE
sethleigh
I noticed your graphics test in Fire Strike was significantly higher than mine even though my gpu was higher clocked than yours. You seemed to have a significant memory boost. In the little bit of memory boost I tried so far it seemed that some speeds increase the performance, and some speeds reduce it. I have no idea how to interpret that, but that's what I was seeing.
Where did you find that? I'll have to take a look at it tonight when I get off work.
My bad, it was Johnalan who'd just posted a Fire Strike score lower than mine, but with a noticeably higher graphics score, even with an (apparently) lower gpu clock.
Happy EVGA customer. Affiliate Code: 0Y7-1VU-ATW2 GigaByte X570 Aorus Master, AMD Ryzen 5900x under Optimus Foundation block, 32gb G.Skill DDR4 @ 3800 MHz 14-14-14-28, EVGA 3080ti FTW3 Ultra under Optimus block, 2TB 980 Pro SSD, EVGA Supernova G6 850W PS, ASUS 34" 3440x1440p 120Hz ultrawide, Lenovo 24" 1080p secondary monitor, Win 10
|
TRClark911
SSC Member
- Total Posts : 565
- Reward points : 0
- Joined: 2017/03/03 21:56:17
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/29 14:30:37
(permalink)
sethleigh
I'm not sure I've seen 2088 Mhz during a game. That's running the folding client. I've seen a lot of 2025-2063 or so during various benchmarks. I haven't done much yet but make some settings and run the benchmarks. It's not really dialed in, and I need to go back and look at the power/voltage limits and whatever else and remind myself how it all works in Precision X OC. I haven't played with it in a few months, since I got the 1080 SC originally and played with it a lot.
2063mhz works good for me on benchmarks... I was having issues with the card hitting the power limit and downclocking to 1911mhz... I resolved it by lowering the memory OC from +250 down to +100 and haven't had issues since. 5.0ghz 1.34v put me at 23,300 Firestrike but I prefer to run at 4.9ghz 1.27v for stability... I've hit 22,850 at that speed.
|
sethleigh
SSC Member
- Total Posts : 796
- Reward points : 0
- Joined: 2015/08/12 11:27:56
- Status: offline
- Ribbons : 4
Re: 1080 TI FE Successful stable overclocks
2017/04/29 15:08:16
(permalink)
TRClark911
2063mhz works good for me on benchmarks... I was having issues with the card hitting the power limit and downclocking to 1911mhz... I resolved it by lowering the memory OC from +250 down to +100 and haven't had issues since.
5.0ghz 1.34v put me at 23,300 Firestrike but I prefer to run at 4.9ghz 1.27v for stability... I've hit 22,850 at that speed. Just spent a couple hours tweaking and testing. Since my earlier 2088 (with occasional 2100) had run in Time Spy and Fire Strike, but not in Heaven, I reset my memory clock back to +0 and just tested the gpu to find out what was rock solid, then ramped the memory back up. In the end I was able to run all the way up to 2075/6003 gpu/memory speeds (+165/+500), but my Heaven score was actually lower than it was at 2075/5954. By watching the OSD during tests I figured out why. At 2075/5954 the gpu clock stayed pretty locked in at 2075 MHz with occasional short blips down to 2063. When the memory clock went to 6003 I saw the gpu blip down to 2063 more often, and actually drop to 2050 a few times. So I crank the gpu up to the last speed step at which it will run reliably, and then with whatever power budget is left within the 120% limit I'm able to crank the memory clock speed. At some point the memory is using up enough power that it forces the gpu to have to throttle back sometimes. I verified this by testing at 2075/5900, 2075/5954, and 2075/6003. The 5954 memory speed score was higher than the 5900, and also higher than the 6000 memory speed score. I've gone back and redone all the 3DMark tests, easily beating all of my previous runs. Fire Strike Extreme: 14,354 http://www.3dmark.com/fs/12489992Fire Strike Ultra: 7,702 http://www.3dmark.com/fs/12489921 Fire Strike 1.1: 22,232 http://www.3dmark.com/fs/12489818 Time Spy: 10,523 http://www.3dmark.com/spy/1663494 I also exceeded 100fps in the 1440p Heaven benchmark for my first time, getting 100.9fps and 2541 score on it at the 2075 MHz gpu / 5954 Mhz memory settings. I guess I should just reboot into my max stable cpu OC I'd worked up last summer, that I haven't used since (its only real use is adding 2mm to my e-peen) and make sure all this still works.
Happy EVGA customer. Affiliate Code: 0Y7-1VU-ATW2 GigaByte X570 Aorus Master, AMD Ryzen 5900x under Optimus Foundation block, 32gb G.Skill DDR4 @ 3800 MHz 14-14-14-28, EVGA 3080ti FTW3 Ultra under Optimus block, 2TB 980 Pro SSD, EVGA Supernova G6 850W PS, ASUS 34" 3440x1440p 120Hz ultrawide, Lenovo 24" 1080p secondary monitor, Win 10
|
Valtrius Malleus
iCX Member
- Total Posts : 290
- Reward points : 0
- Joined: 2017/03/16 19:46:33
- Status: offline
- Ribbons : 2
Re: 1080 TI FE Successful stable overclocks
2017/04/30 04:00:50
(permalink)
sethleigh
DRY_ICE
I've used superposition benchmark for initial stability testing. It'll lock in the core clock on the osd (verified with afterburner) for me on each run and not fluctuate - temps don't go above the throttle threshold. I was able to get a pass I believe at 2036mhz core on one run but it's not stable during gaming or extended use. 2025 is the max stable I can get that'll pass benchmarks and extended gaming sessions.
I did some more testing, and in Time Spy and Fire Strike if my gpu hit 2100 it often crashed. In Fire Strike it hit 2100 fairly often, but then it eventually crashed. I dropped my settings enough to nudge it down to 2088 and it passed Time Spy and Fire Strike, but then it eventually would crash out in Heaven. The last crash was with the voltage at 90%, 2088 Mhz and no memory speed boost at all, so it's looking like 2088 just isn't rock stable. Didn't have the time yet to drop it to 2075 and retest, but I'll try it out today and see if that's going to be stable. I'll try Heaven again and the Superposition one. Once I'm sure how high the gpu will go I'll start over with memory boost.
I noticed your graphics test in Fire Strike was significantly higher than mine even though my gpu was higher clocked than yours. You seemed to have a significant memory boost. In the little bit of memory boost I tried so far it seemed that some speeds increase the performance, and some speeds reduce it. I have no idea how to interpret that, but that's what I was seeing.
Adjusting the voltages doesn't change that voltage across the whole range of speeds. The card checks the voltage curve for the speed, at the voltage it is currently at. Increasing the voltages alone won't make it more stable. Instead, increasing the voltage percentage will allow the card to draw more power (depending on power limit etc) which it will then check the voltage curve and see what clock speed it should be running at with those voltages.  If you want the same clock speed with higher voltages you need to do something which seems counter intuitive. You need to lower the offset! but increase the voltage percent. That will allow the card to push more voltage through which will then check the voltage curve and see what frequencies it would be running at with those voltages. If you lower the offset then it will be running at the same speed, but a higher voltage, which might make it more stable. tl;dr - increasing the voltage percent allows the card to draw more volts (voltages?) which then checks the graph to see what frequency it should be running at with those voltages. Unfortunately adjusting the voltage curve directly isn't really possible with Precision X because the bar chart voltage curve only allows increments of +25 for the offset. I would much prefer more precision like the graph the fan curve uses, which afterburner also implements. But i digress.
Attached Image(s)
|
Category 5
Superclocked Member
- Total Posts : 148
- Reward points : 0
- Joined: 2017/03/02 10:20:25
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/30 06:26:15
(permalink)
Man, you guys stable above 2050 are making me jealous. I am pretty sure once I get water my first card can get close to 2100 because I've managed to see it by using the voltage curve, but it generally likes to run at 2025 during stable use.
The second GPU holds everything at around the lower 1900s during SLI, but even though I see it get up to 1961 it needs 1.09V to do that. Card one is nowhere near the power target when clocked at 1961, with its voltage at .9
Seems stable in gaming at the lower 1900s in SLI, but I mentioned before Heaven can crash that card even with no overclock. Seems Heaven is taxing the the GPU the hardest over here. My theory is that there is one bad CUDA core on my second card so it behaves most of the time but eventually crashes when I'm at sustained 100% GPU usage on that card. Would I need to buy the extended EVGA warranty to get an advance RMA?
|
shadowboricua
iCX Member
- Total Posts : 487
- Reward points : 0
- Joined: 2011/04/27 11:38:23
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/30 06:34:15
(permalink)
My GPUs use to get 2025 stable on air..... and now they also only get 2025 stable on water, even with temps cut in half. The only thing that improved was memory from +400 to +500.
|
Bobmitch
CLASSIFIED ULTRA Member
- Total Posts : 7811
- Reward points : 0
- Joined: 2007/05/07 09:36:29
- Status: offline
- Ribbons : 49
Re: 1080 TI FE Successful stable overclocks
2017/04/30 06:46:20
(permalink)
Did some initial playing around. +160 on core and +400 on memory thus far. Getting 2050 pretty stable with just fan curve in Afterburner.
MSI MAG X670-E Tomahawk; Ryzen 7 7800X3D; Asus TUF RTX 4070 TI OC; Seasonic Vertex GX-1000 PSU; 32 GB Corsair Vengeance DDR5-6000 CL30 RGB; Corsair iCue Link H150i LCD 360MM AIO; 2-Western Digital Black 4 TB SN850X NVMe; Creative SoundBlaster Z; Lian Li Lancool III; Corsair K70 RGB Pro MX Speed Silver Keyboard; Razer Viper 8K Mouse
Heatware: https://www.heatware.com/u/72402 Affiliate code: 1L2RV0BNQ6 Associate Code: UD82LJP3Y1FIQPR
|
Category 5
Superclocked Member
- Total Posts : 148
- Reward points : 0
- Joined: 2017/03/02 10:20:25
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/30 07:04:35
(permalink)
Disheartening that water isn't improving much. Are you adding voltage? Did you try each card by itself? If that's SLI the slower card is running the show.
2050 on air is excellent. BTW
|
yoitsmegabe
New Member
- Total Posts : 33
- Reward points : 0
- Joined: 2016/07/18 16:56:19
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/30 14:35:39
(permalink)
Damn, did I just get a dud? My evga founders crashes at 1965 my max temp is 75@ 88% fan speed. I had an asus founders that hit 2025 but stayed at 1987 b/c of thermals. I returned that one for a msi gaming x thinking it would be even better but it turned out to be defective. The EVGA was all micro center had in stock.
|
Category 5
Superclocked Member
- Total Posts : 148
- Reward points : 0
- Joined: 2017/03/02 10:20:25
- Status: offline
- Ribbons : 0
Re: 1080 TI FE Successful stable overclocks
2017/04/30 14:49:30
(permalink)
I don't think it's a dud. These just vary a lot more than the 1080s. Your not the only one stuck under 2000
|