Seemingly great OC'ing PCS+ 290 throttling?

Took a turn at GPU overclocking to see what my new PCS+ 290 could do.

Without bumping voltage at all, I got up to 1175 on the core, making stops at 1100, 1125, and 1150 on the way.

One thing I've noticed about this card is that it gets quite a bit louder than my Sapphire Tri-X 290 with both at stock settings (Sapphire:1000Mhz, Powercolor:1040). Seems like Powercolor chose 70 as a target temp, causing the fans to get up to 80% (~3400RPM) at times. The Tri-X, meanwhile, maxes at around 79-80C, but the fans stay at less than 50% (~2400RPM). I'll likely start toying with custom fan profiles as I'm fine with 80C on this GPU and would much prefer a quieter card.

The numbers (running BF4 @1440P, Ultra, 2xAA, 80 FOV):


At 1175, I started to see what looks like throttling. Possibly a coincidence, but right when it hit 72C, the usage dropped to 86%, core speed began to fluctuate...



Is that usually what throttling looks like?

Seemed like I was on a roll, without even bumping up the voltage.

I can't believe I've never used custom fan profiles before! Set up a curve in Trixx with the points: 60% @ 80C and 100% @ 95C.

Card temperature maxed out at 77C and the fans at 46%! Powercolor must have a crazy aggressive fan speed profile built into these cards.

Bumped my core back up to 1175 with the new fan profile curve. BF4 seemed like it was running alright for about 4-5 minutes, then I got some DirectX crash I've never seen before? I wasn't able to get a screen cap, but looking back at the realtemp output I noticed...

-core speed was jumping around a bit from 1099-1175 (some fluctuation is normal, but that seems like a lot)
-temp was at 75C when the crash occurred
-fan speed was at 46% @ crash time, then jumped to 53% (even though the temperature dropped after the crash - maybe it was Trixx that made it crash?)
-VRM1 temp was 89C, but it had been higher in previous tests without issue

There really wasn't too much out of the ordinary. I'm guessing 1175 is just right around (well, a bit over) the max of the card at stock voltage.

I'm happy with that, I suppose. I might toy around with bumping up the voltage since temps aren't out of control.


EDIT: Was looking back at my VRM temps. My Tri-X 290 was getting high 70's on VRM1 (with, presumably a default fan profile similar to the custom one I created for the PCS+). The PCS+ 290 was getting about the same with its default fan profile (which was a little aggressive, in my opinion). When I added the custom fan curve profile, that's when VRM1 temps got up in the 90s. Two more questions:

1) I wonder if Powercolor made their fan profile so aggressive on the PCS+ just to keep the VRM cool? The core certainly doesn't seem to require 80% fan speed (only needs 46% max from my tests).

2) Would the VRM heat issues (90C+) affect my overclock on the card?


I'll take a closer look at the VRMs of both cards and see if there's a difference in the heatsinks. Maybe the PCS+ could benefit from something like this?...

http://www.newegg.com/Product/Product.aspx?Item=N82E16835426042&cm_re=290_vrm-_-35-426-042-_-Product

TL;DR What does throttling look like in 290s and can VRM temps cause it even if your core temps are fine?

my 2 cents from owning previous high end cards, fans on all modern GPUs are built to last much longer then the card, so i wouldnt idle on anything less then 40% fan (you will find its far better then the default 20 or so % and wont have higher idle temps then 50% fan speed) then find out your average load temps and jack the fan up another 10%

So in my case 40% fan 30 degrees idle, then under load it goes to be 60% fan speed. Since my current GPU is low end the temps are low, high end cards have far higher temps due to the power going through them, i would just do what you need to keep the card under 80 degrees on load while trying to keep the fan speed under 80% (noise ect and possible lifetime)

I ran my EVGA 295 for 4 years with 80% fan on load before the card died of a code 42, fans were never in question, legit.

Hope that helps you decide what kind of profile to run, as for the stability issues, i think you will find the problem is more then almost certain in the overclock, since the power is untouched i really doubt its going to OC more then a few percent.

I could probably ramp it up a little bit, but 80% is pretty loud

It doesn't go over 80C, even after an hour or so of gaming (with fan speed staying below 50% on my curve), so I'm not sure why Powercolor was so aggressive with it (well, actually I'm thinking it's the VRM heat issue..)

Anyone know if this will fit on the PCS+ 290 without an aftermarket HSF?

http://www.newegg.com/Product/Product.aspx?Item=N82E16835426042&cm_re=290_vrm-_-35-426-042-_-Product

I'm assuming it just goes under the HSF, but I don't know if it's specifically designed to be used with the G10 adapter thing or if it will go under any HSF.

I emailed Gelid and they said it would work. Guess I'll give it a try

How hot were your vrms running? The xfx double D cards would hit 110℃ vrm without throttling.

I'm assuming you turned up powertune? Stock zero % powertune would cause throttling on my cards right around +50mv and 1150 core, under water.

And if you want to check for throttling open gpu-z sensor page, open heaven or valley, and watch for clock speed dips that do not coincide with scene changes. Windowed mode or dual monitors helps with that. And the dips can be anywhere from like 8-100mhz. I did a lot of powertune testing back when the mantle drivers first came out and powertune was broken.

My VRM1 temp was in the high 90s. From what I've read the past few days, it seems it's a common issue with the PCS+. The Gelid VRM heatsink was only $14, so I figured, might as well give it a try to eliminate it as a possible issue.

I'm at stock voltage (though, I think I also read that the PCS+ has some adaptive voltage feature? maybe that was a diff card, actually...). Either way, the only thing I adjust in Trixx was the core speed.

Turn up power tune if you are overclocking. 90℃ vrm won't cause throttling but 0% power tune easily can.

Gah! Ordered the Gelid VRM kit based on the email I received from their (apparently) inept support team stating it would "work just fine". Didn't even come close to fitting under the heatsink of my Powercolor PCS+ 290. The heatsink over top of the VRM1 area (right of the core) comes in contact with the longer heatsink from the Gelid kit. It's a good 1-2mm too tall. :banghead:

It's ok, I thoroughly enjoy wasting money and time...

GELID TEAM,

I ordered the linked kit for my Powercolor PCS+ 290. I did so solely because you said "it should work fine". It arrived today and I attempted to install it with no luck. The larger VRM heatsink from the kit is too tall and makes contact with the heatsink on the video card's cooler.

Thank you very much for wasting my money and, much more tragically, my time. I will make sure I let others know of the standup job your support team is doing...

<redacted>
From: [email protected] <[email protected]>
Sent: Tuesday, September 16, 2014 11:11 PM
To: <redacted>
Subject: RE: Message from GELID Solutions website : End User

Dear <redacted>

Yes, it should work fine.
GELID TEAM

-------- Original Message --------
Subject: Message from GELID Solutions website : End User
From: <redacted>
Date: Wed, September 17, 2014 12:05 am
To: [email protected]

Name: <redacted>
Phone:
Email: <redacted>
Country: United States
Message: Will this item work with the heatsink/fan on a Powercolor PCS+ R9 290? http://www.gelidsolutions.com/products/index.php?lid=1&cid=13&id=108&tab=1 Thanks!

Click to expand...

On a positive note, I made sure the thermal tape over VRM1 was properly seated when I reinstalled the stock heatsink. Also applied some leftover Coollaboratory Liquid Pro to the chip.

I'll need to test it more extensively to see if it really made a difference, but on a quick BF4 run it looked like the core temp dropped ~5C and the VRM1 temp dropped down to the mid 80s (instead of 90s).

I have the pcs+ and my vrm1. Ever gets close to the 90C mark. I have a custom fan profile set, so fans are at 100% during stress or bench tests. Hottest my vrm1 got was 78C. That was after repeated heaven and fur mark 3d runs to test stability when I was playing with max OC at stock voltage.

I'm able to get 1150mhz core and 1600mhz memory for benching. Some games don't like the memory clock though. So gaming wise, I can squeeze 1150/1550 at stock voltage. Other than that, vrm1 temps never get past 60C usually when gaming, even for extended periods of time.

I should probably mention my fan speed when gaming is usually between 50%-75%.

Also, when did you buy this card? I know for a while there were alot of issues. For instance, my card was part of a batch that needed BIOS flashing. Card was constantly giving black screens, stutters, and straight driver crashes. Contacted customer support. I had to send a picture of all 3 bar codes on the backplate of the card to verify I as an owner. They emailed me the bios. I had to use ATIWINFLASH to flash it on there. Viola. Everything is flawless.

I haven't had any issues with black screens, even when gaming at 1150 on the core. I had heard about those though and will take a look at what bios is on the card.

Your temps for VRM1 seem pretty much right on what mine were before I started using the custom fan curve. From my adventure with the Gelid heatsink, I'm guessing the heatsink on vrm1 just isn't the greatest (there's not much surface area to it compared to the Gelid) so it just requires more air flow. My core temps are fine regardless, so I'm just trying to find the balance between fan speed and VRM temp at this point.

Definatly messed up that they said it will fit, but try and sand it down a little, either the bottom where it mounts or the top, the heatsink that is, to shorten the height. Aluminum sands very easily, especially only 1-2 mm..

ninjacore said:

...."inept support team stating it would "work just fine". Didn't even come close to fitting under the heatsink of my Powercolor PCS+ 290.

Click to expand...

I thought about that. After I removed surface area, though, who knows if I'd even gain anything, temps-wise.

I still have the kit. It was only $15 or something. I'll probably mess with it later if I have more time or just save it for use if I were to WC the card.

I think you might be onto something with the VRM temp. I was on OCUK reading about this card, and they were stating that, the reason the original bios was having problem, was it limited the power sent to the VRM, which caused the card to be unstable, especially under load. Seems like they set a limiter on it, to keep it from heating up, maybe it was unstable at hotter temps? Id love to find out what happens with that new sink on the vrm, with the new bios, and overclocked.

I battled with vrm1 temps for months on this card. I was generally getting same temps as you and sometimes closer to 100C on VRM1. I finally got fed up and ordered the Gelid kit with the NZXT G10 and a Corsair h90. After install VRM temps would top out at around 86-87C which was much more comfortable for me but not that great of an improvement. I then ordered some fujipoly extreme for the VRM under the heatsink and temps dropped to around 75C after gaming for a couple of hours. Huge difference. Core never got above 50C so wasn't really an issue. About a month ago I ordered a Komodo waterblock from swiftech, threw a little fujipoly on VRM's and now they max out around 45C. It really comes down to how much you want to put into it. G10 and H90 didn't help much with noise but it did with temps. the komodo made everything whisper quiet and awesome temps but had to build a new custom loop which costs me another $400.

I've seen powertune cause Throttling like supertrucker said.
I also recently had BF4 throttling on me and i took the checkbox off Disable ULPS (it unchecked when i installed 14.9.1 drivers) and it stopped doing it(most would argue it should have the opposite effect) It could have maybe been the drivers but i tried them before a week ago with the same issues but i also kept rechecking Disable ULPS in Afterburner.

It may have been a glitch with BF4 specifically i'm not sure but it caused BF4 FPS to bog down hard and gpu usage to sit at 0-5% with occasionally spiking back up to 90-100% randomly.

These 290/x's have been plagued with weird issues like this since release. Both my cards are under water with generous amounts of Rad @ 35-40*C under load and Vrms never going over 40*C yet i still have the odd issue with certain things that very much looks like some kind of throttling lol.

Maybe one day we will have a Hard and Fast Do's and Don't list for these cards for OCing that works 100% of the time to get 100% performance from them but i fear they will be End of life by the time we figure them out completely like we did with 7xxx series cards .

ed_biii said:

I battled with vrm1 temps for months on this card. I was generally getting same temps as you and sometimes closer to 100C on VRM1. I finally got fed up and ordered the Gelid kit with the NZXT G10 and a Corsair h90. After install VRM temps would top out at around 86-87C which was much more comfortable for me but not that great of an improvement. I then ordered some fujipoly extreme for the VRM under the heatsink and temps dropped to around 75C after gaming for a couple of hours. Huge difference. Core never got above 50C so wasn't really an issue. About a month ago I ordered a Komodo waterblock from swiftech, threw a little fujipoly on VRM's and now they max out around 45C. It really comes down to how much you want to put into it. G10 and H90 didn't help much with noise but it did with temps. the komodo made everything whisper quiet and awesome temps but had to build a new custom loop which costs me another $400.

Click to expand...

Overall, I'm more than satisfied with the card. I can OC to 1150 and game, so I'm happy. I've never really thought it was worth it to watercool a GPU. You spend 50-100% more on WC (than just the card) and get 10-20% more performance?

I also have no idea if keeping the VRMs cooler would even help my OC.