TR Forums

Waco · Wed Jul 08, 2015 2:21 pm

Have you monitored VRM temps before/after? I'm half tempted to do this mod on both my 780s if the VRM and memory temps are stable!

drfish · Wed Jul 08, 2015 4:47 pm

I've only been looking at the core temp with GPU-Z. I picked the Corsair over the NZXT option because I was worried about the same thing and this one was designed to cool everything. I don't have temps though.

Waco · Thu Jul 09, 2015 2:18 pm

Silly me, I didn't realize you swapped the whole bracket for the Corsair thing.

Nevermind, I'd imagine the temps are perfectly fine.

drfish · Tue Oct 20, 2015 7:01 am

So a question for people more versed in GPU overclocking than me... Last night I disabled the overclock on my 780 because after months of seeing artifacting in low GPU usage scenarios I wanted to make sure it wasn't a permanent problem with the chip or something. Lo and behold with everything set to default the artifacting went away and the voltage and clockspeed went UP.

A little more history, I've been running it with boost at 1150Mhz, power target set to 106% and with +37mv - and when I play a demanding game it runs at 1.2v and everything is fine. However, if I play a less than demanding game, like the dabbling in Planetbase I was doing over the weekend, it only runs a ~0.9375v maybe around 700Mhz (still hitting 100fps) and it has all sorts of problems with artifacting (but never crashing)... Funny thing is that with everything set back to default it runs at ~835Mhz but now it is using ~1.025v and the artifacting is gone. What goes on here? Why is it using less voltage and causing problems when I'm explicitly allowing it to use more?

Forgive me for not taking down the exact voltage and clock speed numbers.

Atradeimos · Tue Oct 20, 2015 7:55 pm

drfish wrote:
Why is it using less voltage and causing problems when I'm explicitly allowing it to use more?

What utility are you using to OC? Are you changing the memory clocks at all?

biffzinker · Tue Oct 20, 2015 11:30 pm

Sounds like the overclock is interfering with the default clock/voltage/power target table programmed in the bios-(epprom on the card) that causes the lower voltage/clock your seeing. I think someone posted a thread here or I read about it on another forum. Started being noticed with Kepler, and I think Maxwell has the same issue.

Edit 1:TechPowerUp has bios tweakers for Kepler/Maxwell that may allow you to change the clock/voltage/power target table, but then you risk having to flash the card.
http://www.techpowerup.com/downloads/Utilities/BIOS_Modding/

Edit 2: Forgot I was going to include a image of the program

drfish · Wed Oct 21, 2015 6:17 pm

So far this seems to be fixed if I do not add the extra 37mv. Memory is not OCed. Going to try and go past 150Mhz after a couple more days of no problems.

Jigar · Thu Oct 22, 2015 3:19 am

Can you post your 3D mark scores ? I want to see how much boost you are getting.

drfish · Thu Oct 22, 2015 4:43 am

With overclock: http://www.3dmark.com/3dm/8991903?
No overclock: http://www.3dmark.com/3dm/8992077?

I'll post a followup if I decide to try and take it further or OC the memory...

Jigar · Thu Oct 22, 2015 12:49 pm

Something isn't right with the scores here - OCed GTX 780 should slaughter my HD 7970...

Here are my OC results - http://www.3dmark.com/fs/6287286

Comparison for better understanding - http://www.3dmark.com/compare/fs/6284362/fs/6287286

EDIT: GPU @ 1150 MHZ - RAM @ 1700 MHZ

drfish · Thu Oct 22, 2015 1:03 pm

They were run hastily this morning, G-Sync was on, vsync was application controlled I think unless the driver forced it on again like is has been. I'll have to double check. I just used the default settings of the demo...?

Jigar · Fri Oct 23, 2015 12:24 am

drfish wrote:
They were run hastily this morning, G-Sync was on, vsync was application controlled I think unless the driver forced it on again like is has been. I'll have to double check. I just used the default settings of the demo...?

Same here, i am using the default settings of the demo - i do think your card might be bandwidth starved, can you check if your card is on pcie x16 gen 2X ? Also Gsync and Vsync need to be off...

drfish · Fri Oct 23, 2015 2:30 am

OCed results with G-Sync and vsync forced off: http://www.3dmark.com/3dm/9001998?

Card is plugged into an PCIe3 x16 slot.

Jigar · Fri Oct 23, 2015 3:35 am

I am stumped, you obviously have a beefy card and your CPU is already showing its grunt, why are your scores so low.

EDIT: I will check what is normally expected 3Dmark score by GTX 780.

EDIT 2: Looks like i was correct, your scores should be almost double - http://www.overclock.net/t/1506758/this ... 0k-gtx-780

Review scores - http://hexus.net/tech/reviews/graphics/ ... 80/?page=4

drfish · Fri Oct 23, 2015 6:24 am

Huh... I guess I'll have to keep looking into this... :-?

Performance in games is what I would expect at least - will have to double check that too now.

vargis14 · Fri Oct 23, 2015 8:37 am

drfish wrote:
Huh... I guess I'll have to keep looking into this...

Performance in games is what I would expect at least - will have to double check that too now.

Since it is water cooled and you know what you are doing ever think about getting a modded bios or modding your own bios?
The guys on Overclock.net's forums are very helpful if you do not want to mod it yourself you can send them your bios image from GPUZ and they will mod it.
Does your card happen to have a bios switch and more than one bios? Since I do not know what particular card you have.

As for overclocking I like EVGA's Precision X software it is a little more complicated but it has never let me down.

drfish · Fri Oct 23, 2015 8:54 am

I might go the modded BIOS route if I have to wait extra long for Pascal but generally I take a safe and mild stance on overclocking. So far switching to the water cooler has been a nice experiment. It's just a reference EVGA card, very small stock overclock, no toggles for fancy modes like on a Classified or Kingpin version. I'm using Precision X now, might try going to +200-250Mhz this weekend just for kicks, maybe even bump the memory too now that I know the artifacting was from the +37mv...

Really curious about the 3DMark thing now, maybe I'll try running it with my X34 disabled and using my secondary monitor as my primary display... Will report back.

Jigar · Fri Oct 23, 2015 10:00 am

drfish wrote:
I might go the modded BIOS route if I have to wait extra long for Pascal but generally I take a safe and mild stance on overclocking. So far switching to the water cooler has been a nice experiment. It's just a reference EVGA card, very small stock overclock, no toggles for fancy modes like on a Classified or Kingpin version. I'm using Precision X now, might try going to +200-250Mhz this weekend just for kicks, maybe even bump the memory too now that I know the artifacting was from the +37mv...

Really curious about the 3DMark thing now, maybe I'll try running it with my X34 disabled and using my secondary monitor as my primary display... Will report back.

Can you share the screenshot of your bios settings (All of it) ? May be that might help in pointing out the odd tweak out.

EDIT: Just observed one more thing- your signature shows you have 16GB Ram but 3DMark shows your system has 8GB RAM - are you intentionally running single channel setup? I haven't seen dual channel setup make this much difference in 3dMark/Games but maybe you can retry running the bench with dual memory channel setup.

drfish · Sat Oct 24, 2015 6:27 am

8GB is correct, vargis has 16GB in his sig.

I ran the benchmark again, same Nvidia control panel settings but this time on my secondary screen instead and got a much better score: http://www.3dmark.com/3dm/9016521?

Now I'm wondering of the high 100Hz refresh rate was a factor in the calculation? Who knows... I'm pretty sure this is a 3DMark problem and not a problem with my rig. :-?

Edit: Ran it again with secondary monitor disabled, same result as on the X34 with it enabled - something about running on the X34 makes the score lower even when G-Sync and vsync are disabled. Updated Compare result: http://www.3dmark.com/compare/fs/6301708/fs/6287286

Jigar · Mon Oct 26, 2015 3:32 am

9503 score is something in line with what you should expect from GTX 780. If i were you i would check FPS of all games on both the monitor, if X34 is producing low FPS compared to your other monitor, you might have a bad unit.

Jigar · Wed Oct 28, 2015 2:15 am

Found some bad user reviews for X34 - http://community.acer.com/t5/Predator-D ... 0434#M1137

ultima_trev · Wed Oct 28, 2015 6:10 am

You have your GTX 780 OCed to 1100/7000 and can only get ~10800 on GPU score? Something must be very wrong as it should rival a stock GTX 980 at those clocks (>13000). R9 390 and GTX 780 tend to perform the same clock for clock if not a small advantage going to the GTX 780. I can do 13000+ at 1080 core with no OC to memory.

Fri Oct 30, 2015 12:14 am

drfish wrote:
Huh... I guess I'll have to keep looking into this...

Performance in games is what I would expect at least - will have to double check that too now.

http://nvidia.custhelp.com/app/answers/detail/a_id/3798

This is a GeForce Hot Fix driver, version 358.59 that addresses the following issues:

•G-Sync Windowed mode & full-screen performance issue

drfish · Fri Oct 30, 2015 5:24 am

Yep, I've been running that driver since 10/20. :-?

TR Forums

plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Re: plunge Bought a GTX 780

Who is online