HG10 N980 & N970 Thread - Post Issues Here

[mattlach]

Mattlach: Ambient is 22,7*C

And ASIC.... I don't know how to check that? How I do to check that in Gigabyte cards? I looked in nvidia Control panel and Nvidia Experience but couldn't find anything.

 

You can check the ASIC quality using GPU-Z and right clicking on the top left of the window and selecting "read ASIC Quality".

 

Higher ASIC numbers usually mean lower voltages can be used to reach the same clocks, so they tend to run quieter and cooler, and can also - depending on who you ask - overclock higher.

[sebastiannielse]

71.5% is mine. Its was a completely reference card, with ref cooler and Everything, Before I fitted the N980. Edited November 16, 2015 by sebastiannielse

[shinne]

I got a question, can the HG10 n980 only be used with a reference card? I might an MSI GeForce GTX 980 Ti GTX 980Ti 6GD5T OC and was hoping I could some how fit the HG10 with a AIO watercooler.

[sebastiannielse]

The PCB must be reference, but the cooler doesn't need to be reference. But I would still recommend getting a card with a reference cooler if you are going to remove the stock cooler anyways, because then you can be sure its a reference card with reference cooler and all Components just at the right Place.

[mattlach]

Also, I wanted to mention this.

 

 

Let's resist the temptation to say "It worked for me, thus it must work for everyone else". That's not the way manufactured products work. There are part to part variations in any manufacturing process. If the tolerances arent set right, you might have some that work, and some that don't.

 

 

By tolerances I mean, taking the screws as an example, they will be X mm long +/- Y mm.

 

Every single part has a tolerance by how much it can deviate from the nominal, because in the grand scheme of things, if you measure accurately enough, everything will deviate a little, so you always have to provide a range of acceptance.

 

This is really super simple when you are dealing with just one component, but it quickly gets really complex when you have a large amount of different components with different complex geometries.

 

In the case of the HG10 bracket each of the following will have min/max type specifications:

• Screw lengths
  
• Standoff lengths
  
• hole counter-bore depth
  
• bracket material thickness
  
• Hydro cooler bracket dimensions and thickness
  
• Thermal pad thickness

 

etc. etc.

 

This is further complicated by the fact that it is mating to products that are out of their control, ie the video card, which will have some tolerances of its own, most notably in PCB thickness, GPU mount height, VRM mosfet heights, etc. etc.

 

Now for each of these parts you have to specify dimensions such that when you add up all the tolerances everything still fits.

 

What if all of your components are at their max specified dimensions? What if all of your parts are at their min specified dimensions? Are there combinations of the above that can lead to worst cases?

 

Now add that all these different components come from different vendors, likely of varying quality. it's difficult enough to do a proper tolerance stack-up when all the parts are within your specified tolerances. Now assume a vendor supplies parts that are out of tolerance? Everyone does takes a sample and measures key dimensions when they receive parts, but samples don't always catch everything, and it is usually impossible to measure everything completely. (well, in Defense contracting and medical products this is usually done, but that is also why they cost so much)

 

I'm not mentioning this to excuse the current situation. Something somewhere obviously broke down, be it in design controls, verification/validation or supplier controls, and as the top level brand, the buck stops with Corsair. Personally, before launching any product, I would like to do a test with at least 60 (well 59, based on the binomial distribution) brackets across all part numbers listed as compatible. If none of them have issue during install, you'd be 95% confident that 95% of the population will work. If even one fails, you stop and don't launch.

 

It's possible pre-launch testing was expedited, as they were already late, and the pressure we were putting on them rushed things.

 

The reason I do mention this - however - is so that you understand that there are going to be differences between every single box containing an HG10. Each one is going to have different HG10 subassembly components and different screw kits, and they will be mounting to different video cards using different Hydro coolers, all of which will have differing dimensions hopefully within their specified tolerances, but possibly outside of them.

 

It is rare for a product to hit the market and be all bad. Most products are usually tested at least to some extent prior to launch. It is - however - not as rare, especially on parts with many matching complex dimensions to have tolerance stack up issues where certain combinations of HG10/Screws/board/Hydro Cooler work and otehrs don't.

 

So, for those of you who have made these work, that is great! Just don't assume that this means that every box out there is going to work for everyone. It many not be that you have better installing skills than everyone else, and they have just messed it up. it is absolutely certain that there is box to box variation. Manufacturing - like everything else in the universe - has inherent variation.

 

When you make this kind of conclusion based on a single observation (yours) the binomial distribution tells us:

 

Welcome to Minitab, press F1 for help.

Test and CI for One Proportion 

                       95% Lower
Sample  X  N  Sample p      Bound
1       1  1  1.000000   0.050000

 

This means, at the 95% confidence level, you can only say that 5% of the units out there work based on this one observation :p

 

Your mileage may vary, as they say.

 

Now rather than be all negative, I'd like to support Corsair in addressing this issue and making it work for all of us. I have been a Corsair customer for a long time, and I have come to trust their products more than most of their competition. If you are in the business of engineering products long enough, eventually you'll have a bad launch. it is inevitable. I have every bit of confidence that Corsair can and will fix this, hopefully in a timely manner!

[mattlach]

Pretty sure something is wrong with your numbers. People with custom loops see load temps of ~37-40C on the GTX 980 Ti. There is no way you can achieve lower than that with an H100i, no matter how it is mounted. I suspect you took a screen cap after the Valley loop had stopped and the GPU immediately dropped to the water temp (because it was no longer under load) as you were switching to the Corsair Link page.

 

Unless your ambient termps are about 10C, then it would make sense.

 

 

There is something that is not lining up here.

 

The reason I say this is that HardOCP just tested the MSI sea Hawk 980 Ti.

 

As we all know, it uses a mounting similar to the HG10, and a corsair H50 based AIO (but with slightly beefier fans)

 

In their test, which I believe is performed at 20C controlled ambient, they measured the sea Hawk with an idle temp of 23C (33% fan), during full load at stock clocks, the fans stayed at 33%, and temps rose to 36C at max.

 

In max overclock (+87mv, 1567Mhz in game clock) they never saw temps rise above 42C, again with the fan speed staying at 33%. Raising the fan speed to 100% didn't improve their overclock.

 

I have very high confidence in HardOCP's reviews, more so than any online PC hardware reviewer.

 

It seems to suggest that your custom loop figures are either wrong, or from a rather low end loop.

 

here is the article for your review.

 

It seems to suggest that if we can get a good mate to the cooler, we should see exceptional temps.

[xixou]

I doubt they will test 59 of them (lack of time and boards).

[mattlach]

I doubt they will test 59 of them (lack of time and boards).

 

Probably true.

 

My background is in the medical field where the risks of selling bad product (and the cost of the end product) are a little bit higher :p

[savagepagan]

I read on these forums that using the AMD mounting bracket was advised over the Intel bracket. When will they be in stock? Edited November 16, 2015 by savagepagan

[virdog]

There is something that is not lining up here.

 

The reason I say this is that HardOCP just tested the MSI sea Hawk 980 Ti.

 

As we all know, it uses a mounting similar to the HG10, and a corsair H50 based AIO (but with slightly beefier fans)

 

In their test, which I believe is performed at 20C controlled ambient, they measured the sea Hawk with an idle temp of 23C (33% fan), during full load at stock clocks, the fans stayed at 33%, and temps rose to 36C at max.

 

In max overclock (+87mv, 1567Mhz in game clock) they never saw temps rise above 42C, again with the fan speed staying at 33%. Raising the fan speed to 100% didn't improve their overclock.

 

I have very high confidence in HardOCP's reviews, more so than any online PC hardware reviewer.

 

It seems to suggest that your custom loop figures are either wrong, or from a rather low end loop.

 

here is the article for your review.

 

It seems to suggest that if we can get a good mate to the cooler, we should see exceptional temps.

 

42c overclocked? I think I need to try the AMD mount. My 60c seems way too high!

[sebastiannielse]

mattlach: That review seems to use only gaming to measure the temperature, not a benchmarking software. Since a game does not put full load at the card always, the temps will of course be lower.

 

Also their ambient seem to be lower, my ambient is more near 23*C.

I did a second test to see if I could get the temps lower, and I actually got the temps lower now.

By running unigine for a hour, I did a "burn-in" of the card so the TIM gets to settle propely (as you know, Artic Silver 5 needs a longer session with higher temps to "settle" propely), and now the temps during unigine stays around 35-42.

 

The HG10 N980 is VERY effective cooling the VRM and VRAM too. My card does not have any VRM/VRAM probe, but I can feel that the bracket itself gets hot to the touch when the card is under load. Thats a strong suggestion that solutions without VRM/VRAM cooling such as those "fits-all-card" products with just air blowing at the VRM and no heatsink, and no cooling of VRAM at all, that will mount to any NVIDIA or AMD card, will shorten the life of your card.

[railven]

42c overclocked? I think I need to try the AMD mount. My 60c seems way too high!

 

The temps for the "success" installs are still a tad too high for my linking/expectations.

 

For example here is my setup using the EVGA Hybrid Kit (which is silent, minus the stock fan being a tad too loud but I've yet to swap it):

 

http://i.imgur.com/2JQ7ZH2.png

http://i.imgur.com/FjFC9vf.png

 

That was FireStrike Extreme with +300 on core and mem on a ref Zotac 980 Ti ASIC score something like 72%.

 

There is no way a 60C load score is satisfactory. I was getting that with a custom fan curve on the ref cooler. Granted a little louder.

 

Hopefully Corsair figures it out because I feel there is way more potential in the product/setup of an AIO cooler than current N980 users sans custom mods are experiencing.

[virdog]

I ordered some new paste and will give the cooler mounting another shot.

 

I remember that I would get 80c under load so 60c aint bad.

[sebastiannielse]

I think 60c is very bad, even if you get 80c with stock cooler.

80c then you are literally TOASTING the silicon into a bit of coal over time... (come on! Look at the rubbish nvidia stock cooler. The surface area of the stock cooler is so small so you can literally laught at it, Nvidia really need to hire a new designer for their cooling)

70c is like "You are atleast TRYING to cool the card, but come on!"

60c is more like "Yeah, a little bit of cooling but not great!".

50c is like "Nice cooling dude, but it can be improved."

40c is like "Finally! Now the card will survive very long time."

<35 in load is like "You are a cooling champion!".

 

After tinkering a little with the settings in Corsair link with fixed fans instead of temperature controlled fans, I have managed to achieve a maximum temp of 40*C in unigine. Should experiment a little more, connect the pull fans of the h100i GTX to the Corsair Commander mini instead, and make the pull fans run a bit faster than the push fans, to create a lower pressure inside the radiator, thus lowering the temperature more.

Edited November 17, 2015 by sebastiannielse

[TheDudeLasse]

I think 60c is very bad, even if you get 80c with stock cooler.

80c then you are literally TOASTING the silicon into a bit of coal over time... (come on! Look at the rubbish nvidia stock cooler. The surface area of the stock cooler is so small so you can literally laught at it, Nvidia really need to hire a new designer for their cooling)

70c is like "You are atleast TRYING to cool the card, but come on!"

60c is more like "Yeah, a little bit of cooling but not great!".

50c is like "Nice cooling dude, but it can be improved."

40c is like "Finally! Now the card will survive very long time."

<35 in load is like "You are a cooling champion!".

 

After tinkering a little with the settings in Corsair link with fixed fans instead of temperature controlled fans, I have managed to achieve a maximum temp of 40*C in unigine. Should experiment a little more, connect the pull fans of the h100i GTX to the Corsair Commander mini instead, and make the pull fans run a bit faster than the push fans, to create a lower pressure inside the radiator, thus lowering the temperature more.

 

Settle down mate. GPUs can handle 80c like a walk in the park. You'll have upgraded your graphics cards 2-3 times before the silicon dies on you. Lifespan extension is the least interesting factor about water cooling.

[xixou]

70 degrees is good enough to avoid freq throotling. everyting below 60 is useless.

[skkane]

Not quite. Maxwell scales really well with low temps. You can push your oveclocks considerably if you can keep it under 40.

[IceBone]

My friend in the UK has taken delivery of the N970. Now I just need to buy the H50 and I can install it first thing next year. Excitement!

[mattlach]

70 degrees is good enough to avoid freq throotling. everyting below 60 is useless.

 

I wouldn't go quite that far. In some cases you can get better overclocks at lower temps. Not in all though.

 

Once I get mine installed such that I have low temperatures (My expectations are at least to keep up with the Sea Hawk as I'll be using larger radiators) I plan on turning the fans to 100% and seeing what my max overclock is.

 

After that I will be stepping the fans down and bit by bit allowing the temperatures to rise until I start seeing instability.

 

Once I hit instability, I am going to create a fan profile that keeps the GPU's ~5C below the instability level, and then do a long term stability test to confirm that this is actually good.

 

This way I will get the best of both worlds. High speed and max ocerclock when I need it (but only as high speed as is required to get me to ~5C below instability level) and fans spinning all the way down to 0 when I don't need them (This is why I selected the Noctua fans as they have similar static pressure and airflow characteristics as Corsairs beefy fans bundled with the H110i GTX at 100%, but as opposed to those fans, they have a minimum speed of ~430rpm (500rpm in specs, but testing shows 430rpm) and if the PWM signal goes down to 0 the fans stop. That, and at full speed they are ~10dB quieter than the Corsair fans from the H110i GTX.

 

Surprisingly the Noctua fans can do this while rated at 0.18A and 12V, whereas the default corsair fans that came with my H110i GTX are rated at 0.7A each.

 

The downside with these fans - however - is their price. I scoffed at their $28 per fan price tag at first, but you really do get what you pay for.

[Corsair Dustin]

There is something that is not lining up here.

 

The reason I say this is that HardOCP just tested the MSI sea Hawk 980 Ti.

 

As we all know, it uses a mounting similar to the HG10, and a corsair H50 based AIO (but with slightly beefier fans)

 

In their test, which I believe is performed at 20C controlled ambient, they measured the sea Hawk with an idle temp of 23C (33% fan), during full load at stock clocks, the fans stayed at 33%, and temps rose to 36C at max.

 

In max overclock (+87mv, 1567Mhz in game clock) they never saw temps rise above 42C, again with the fan speed staying at 33%. Raising the fan speed to 100% didn't improve their overclock.

 

I have very high confidence in HardOCP's reviews, more so than any online PC hardware reviewer.

 

It seems to suggest that your custom loop figures are either wrong, or from a rather low end loop.

 

here is the article for your review.

 

It seems to suggest that if we can get a good mate to the cooler, we should see exceptional temps.

 

Actually, MSI's implementation is fairly different from ours and doesn't incur the complexities of having to support so many Corsair coolers. That's why the Seahawk (or Hydro GFX as we prefer) works so well.

 

That being said, I personally still lean towards the HG10-N980 for the ability to go up to a 240mm or 280mm rad, though I have the privilege of being able to test potential fixes internally as well as the experience to tweak fitment of the bracket.

 

We're currently investigating fixes to get everyone's N980s up to speed. I have an unofficial - and I must stress this is unofficial - but I have an unofficial fix that we've tried internally and has worked fairly well. If you're using an Asetek-built cooler (H50, H55, H75, H80i GT, H90, H100i GTX, H105, H110, H110i GTX) you can place #6 steel washers between the cooler's bracket and the HG10's standoffs and it should ameliorate bowing, improve temps, and prevent standoffs from breaking. I recommend using double-sided tape to keep the washers in place. This isn't the kind of fix we can reasonably expect our end users to do, and it's not a perfect one.

 

That said, this is an unofficial fix. But if you have an HG10-N980 and you're having trouble with it, this is an option.

 

In the meantime, we're continuing to work pretty aggressively on getting this problem solved.

[mattlach]

Actually, MSI's implementation is fairly different from ours and doesn't incur the complexities of having to support so many Corsair coolers. That's why the Seahawk (or Hydro GFX as we prefer) works so well.

 

Yeah, I looked at the pictures of it open, and it appears that since they can decide on the cooler themselves, they can use only Asetek-type coolers, and thus have the space to mount it using the 4 holes surrounding the GPU.

 

I'd imagine this is similar to Sebastiannelse's mod. but since they likely have some good fixturing in their processes, they were probably able to align it better than a modder could.

 

That being said, I personally still lean towards the HG10-N980 for the ability to go up to a 240mm or 280mm rad, though I have the privilege of being able to test potential fixes internally as well as the experience to tweak fitment of the bracket.

 

We're currently investigating fixes to get everyone's N980s up to speed. I have an unofficial - and I must stress this is unofficial - but I have an unofficial fix that we've tried internally and has worked fairly well. If you're using an Asetek-built cooler (H50, H55, H75, H80i GT, H90, H100i GTX, H105, H110, H110i GTX) you can place #6 steel washers between the cooler's bracket and the HG10's standoffs and it should ameliorate bowing, improve temps, and prevent standoffs from breaking. I recommend using double-sided tape to keep the washers in place. This isn't the kind of fix we can reasonably expect our end users to do, and it's not a perfect one.

 

That said, this is an unofficial fix. But if you have an HG10-N980 and you're having trouble with it, this is an option.

 

In the meantime, we're continuing to work pretty aggressively on getting this problem solved.

 

Thank you very much for keeping us up to speed with your progress!

 

I am still waiting for the official solution, but do you have a ballpark guesstimate of when that might happen, and if it will involve returning the bracket or not at this point? I'm a pretty patient person, but I - too - am really starting to itch to get this project done. If I can have an understanding of how much time it might take, I can decide if I want to proceed with the washers, or wait it out.

[jossrik]

This may be off topic, but I can't seem to find a good answer anywhere. I want the HG10 N980, but can't find it anywhere for sale (North America), are they retooling it? I read about issues with installs all over the place. I only just heard about it a couple weeks ago, but would love to try it out, but even finding product pages is near impossible, and when I do find it they say it's discontinued. Is there going to be an alternate? So many questions.

[Makalaure]

There is something that is not lining up here.

 

The reason I say this is that HardOCP just tested the MSI sea Hawk 980 Ti.

 

As we all know, it uses a mounting similar to the HG10, and a corsair H50 based AIO (but with slightly beefier fans)

 

In their test, which I believe is performed at 20C controlled ambient, they measured the sea Hawk with an idle temp of 23C (33% fan), during full load at stock clocks, the fans stayed at 33%, and temps rose to 36C at max.

 

In max overclock (+87mv, 1567Mhz in game clock) they never saw temps rise above 42C, again with the fan speed staying at 33%. Raising the fan speed to 100% didn't improve their overclock.

 

I have very high confidence in HardOCP's reviews, more so than any online PC hardware reviewer.

 

It seems to suggest that your custom loop figures are either wrong, or from a rather low end loop.

 

here is the article for your review.

 

It seems to suggest that if we can get a good mate to the cooler, we should see exceptional temps.

 

 

You can see from their screen caps that they measured GPU temps when it was only under 63% load, hence the low temps. If you look at their GPU-Z screen caps, you can see they are not testing the cards in a repeatable/consistent manner, so their results are not comparable to anything. Also, the fan they have set to 33% is the blower for the VRM/VRAM, so it will not affect GPU temps whatsoever, they are correct. You need to plug the radiator fan supplied with the Sea Hawk in to your motherboard and set the fan profile on there, which GPU-Z will never report, and Afterburner can not control (if you look at their pics of it installed on their test bed you can see the fan cable inserted in to a motherboard header if you require verification). All in all, a terrible way of reviewing: unless you want to bias your reviews, of course. I'm pretty sure I can get temps of 25C playing minesweeper on my Hybrid. Technically that would be whilst playing a game.

 

That review seems fairly decent when it comes to reviewing the FPS of specific scenes/sections of games, but that's pretty much it. I would completely disregard their temperatures and their overclocks.

 

There are at least 2 other reviews out there (here and here) that put the sea hawk max temp at ~50C. Which is still pretty good considering the overclocking potential.

Edited November 18, 2015 by Makalaure

[Makalaure]

70 degrees is good enough to avoid freq throotling. everyting below 60 is useless.

 

Actually it has been confirmed several times that GPU Boost 2.0 starts reducing the boost at ~65C. I am not aware of any thresholds below that, but I suppose there could be.

Edited November 18, 2015 by Makalaure

[anti-duck]

TTL got a maximum temperature of 56° C in a very extended run of Heaven with a heavily overclocked Sea Hawk. Not too bad considering the Sea Hawk only has a H55, but still a little bit off from the competition.