Yeah, i've changed thermal pads and paste and ripped the warranty label. They use ALOT of paste, and I mean ALOT. Before changing my card was hitting about 55°c on the core at 320w with stock bios. Now i'm hitting about 48°c at 370w with the MSI 4080 Super Expert bios. I've reduced the fan speed to 1000rpm, instead of the 1400rpm i was using before. (2x360 rad)

 

That much difference would be worth it for me. I'm at 69c core 85c hotspot now at 370w but that's on 54% fan speed on air. If I can drop 5-8c that would be nice. MX6 or PK3 should be able to do that.

 

New psu corsair rm1000x v3 same coil whine.
Got a new msi rtx 4080 super suprim coming in a few days.

 

I can be honest here. On the stock BIOS at 320w my card is as good as quiet with coil whine but at 400+w it has quite the squeak lol.

I thought about grabbing a Alphacool Eiswolf for the card but then I remembered I can't actually fit another 360 in my case..

Is there like, a 240 or 280 AIO solution like the Eiswolf for the 4080? The Inno3D uses the "reference" PCB btw. Same as PNY and some other models.

 

I can be honest here. On the stock BIOS at 320w my card is as good as quiet with coil whine but at 400+w it has quite the squeak lol.

I thought about grabbing a Alphacool Eiswolf for the card but then I remembered I can't actually fit another 360 in my case..

Is there like, a 240 or 280 AIO solution like the Eiswolf for the 4080? The Inno3D uses the "reference" PCB btw. Same as PNY and some other models.

If you run higher power and voltages throgh the vrm its normal to hear buzzing sound, but i am talking with official bios, if there is coil whine on it, that card prob last max 3-4 years because already too high for it.

 

Just picked up a new Gigabyte 4080 Super Gaming OC before the prices seemingly went up over night. I would have liked the Master, but since release, haven't seen them in stock at release price. Quick impressions - its a really nice card. Seems well built. Zero coil whine so far. PCB is small. 3000/1500 very easy. Temps are very good for a stock cooler. The cooler is huge compared to the pcb. My old x299 platform not doing me any favors in benches. Some glitching between 3d Mark, AB(latest beta and nvidia driver), and GPUz software. Moving voltage slider decreases read voltage to .950mv instead of increasing, as read by AB. Stock was reading 1.00mv. Spiking to 1.095mv during bench in GPUz.. 3080/1500 and it starts to get perfcap due to power. Topping at 355w. Temps are good.

 

I repasted it but for the Inno3D the sticker is on the backplate and another one from backplate.wrapped over to the PCB. Neither of them have to be broken to remove the cooler. I repasted without damaging either of them.

I did figure out it's not paste or fan speed / airflow limiting my cooler. The block just isn't any more efficient. At 355-360w constant load temps are basically identical between 38% fan speed and 60% fan speed. Not even 1c difference. Only at the full 100% does it drop a bit but less then I'd expect. About 6c only.

So, is it still within reason to run a peak of 74.9c core with 87.5c hotspot and 60.0c VRAM? It sounds pretty reasonable to me tbh at constant 360w load (3045 @ 1.100v).

 

my sticker broke after getting wet due to a leak from the water block

 

I was trying some undervolting + Overclocking with the curve and while my card loves to just stretch it's legs at the full 1.100v with 370w it really doesn't wanna undervolt OC well at all.

I tried to push something like 2925 @ 1.025v but in RT it is so unstable it will crash within minutes. 2865 @ 0.985v (which is what my 4070 Ti ran fine) it's not even loading into Cyberpunk with PT enabled.

So yeah, can't just blindly trust the curve. +195 at max voltage is stable but +195 at 1.025 is not for example. I can probably pull off about 2910-2925 at 1.035v ish but that isn't enough of an undervolt to be worth it. I wanna make a low power profile that uses like max 270-280w for hot summers and games that don't need the full 3060-3045Mhz. I will probably end out somewhere in the low 2800's with 0.980-1.000v ish I guess.

EDIT: so, I got a stable profile set up. 2850 @ 0.990 with 12500 (+1000) VRAM. It's sitting around 262w in Cyberpunk with Path Tracing with temps at 61-62 core 74c hotspot 58c VRAM at very low fan speeds (about 34%) and thus is very very quiet and efficient. FPS is only about 4 (89 vs 93 in the same static scene) lower than 3045 +500 VRAM @ 1.100 at 360w. Percentage wise that's about 5% mainly because I can push VRAM higher to +1000 (or more, haven't tested any higher yet) without dropping core clocks due to higher power consumption but it drops power and temps by like 30% so yeah. Way more efficient.

She's definitely a keeper with two exceptions. If I ever find a MSRP or used ASUS Strix or factory AIO card I'll scoop that up for sure.

 

+195 at max voltage is stable

Nice! Would you mind to highscore-benchmark the card with TSX and SW, upload the scores on 3dmark.com and post the links to your results here?

3DMark.com search

3DMark.com search

www.3dmark.com

3DMark.com search

3DMark.com search

www.3dmark.com

 

Sure, as you wish. Speed Way is not CPU or RAM bound at all though, and Time Spy Extreme barely. A quick run should be simple: push the fans to 100 %, clock up voltage, GPU and VRAM, maybe open a window in the room and bench it, that is what I did (ShirKhan @3DMark). My card cannot run +195 MHz stable in games. 180 in some of them, only 165 everywhere. That's why I'd love to see your scores.

 

Got the new card.
Here is old vs new coil whine.
New

Old

To me the coil whine on old vs new sounds exactly the same.

Funny thing, boost clock old was 2880mhz the new 2790mhz, almost 100mhz

 

Actually, if you do Ctrl F in MSI Afterburner for curve editor and switch from voltage tab to temperature tab you'll see the 4080S does NOT drop boost bins at 70 but at 83. It's a flat line. Mine doesn't drop boost bins at all even at insane temps 81c core 96c hotspot still did not drop from 3060 @ 1.100 at all with a 450w BIOS.

Not that I run that daily but for benches.. it's rather nice.

Daily is at 2850 @ 0.990v with +1000 memory now drawing about 260w in Cyberpunk with PT and in lighter stuff like Division 2 or World of Tanks it barely breaks 200w. 55c core 66c hotspot at 34% fan speed. I like.

 

Actually, if you do Ctrl F in MSI Afterburner for curve editor and switch from voltage tab to temperature tab you'll see the 4080S does NOT drop boost bins at 70 but at 83. It's a flat line. Mine doesn't drop boost bins at all even at insane temps 81c core 96c hotspot still did not drop from 3060 @ 1.100 at all with a 450w BIOS.

Not that I run that daily but for benches.. it's rather nice.

Daily is at 2850 @ 0.990v with +1000 memory now drawing about 260w in Cyberpunk with PT and in lighter stuff like Division 2 or World of Tanks it barely breaks 200w. 55c core 66c hotspot at 34% fan speed. I like.

I got the same number on my undervolt.
Set 2865mhz, drops to 2850mhz at 0.990v.
The hardess pass by far was 3dmark speedway.

 

What effective clocks are considered good bins for a 4080 super on air on timespy extreme with 420w PL?

 

My undervolt 0.990v at 2865mhz is a 135mhz overclock.

Thanks, that I could see in the video. Very nice. Yet, the numbers don't seem to match, explanation below.

What is your stock boost clock?

2820, 60 MHz lower than yours. So, my card should need +195 MHz @990 mV to reach 2865. But as shown above, +255 are needed. 60 MHz missing, that's odd.

Next what is strange: Your SW score of 7547 with this great OCUV setting.

Here around +200 pts are scored with only +120 Mhz @975 mV. In Speed Way this is a notable difference. A higher score with a significantly lower GPU frequency ... maybe a matter of too many background apps in Windows on your side?

Link

Could you upload the run in the video and link to it? Maybe the frequency data shed some light on this.

 

Thanks, that I could see in the video. Very nice. Yet, the numbers don't seem to match, explanation below.


2820, 60 MHz lower than yours. So, my card should need +195 MHz @990 mV to reach 2865. But as shown above, +255 are needed. 60 MHz missing, that's odd.

Next what is strange: Your SW score of 7547 with this great OCUV setting.

View attachment 2649853

Here around +200 pts are scored - with only +120 Mhz @975 mV. In Speed Way that is a notable difference. A higher score with a significantly lower GPU frequency ... maybe a matter of too many background apps in Windows on your side?

View attachment 2649861
Link

Could you upload the run in the video and link to it? Maybe the frequency data shed some light on this.

The only thing i can think of why the score is lower is that i run pcie Gen 3.

Sure i can do that tomorow and even a run with gen 4.

 

Curious what type of minimums you guys are getting @ 1080p with a 5800X3D, High settings, in CS2. I get some bad minimum dips when starting on the T side of Nuke like mid to high 100s, the CT side i'm in the low 200s for my mins, in a 10 vs 10. When less players are in the server the minimums are noticeably higher.

 

nilssohn
Here is the score running pcie gen 4.

I scored 7 628 in Speed Way

Intel Core i9-13900KF Processor, NVIDIA GeForce RTX 4080 SUPER x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

Running the fans at 100% and got a even better result, 2880mhz (drops to 2865mhz) at 0.990v.

I scored 7 673 in Speed Way

Intel Core i9-13900KF Processor, NVIDIA GeForce RTX 4080 SUPER x 1, 32768 MB, 64-bit Windows 11}

www.3dmark.com

Got my block today, hoppfully i get even better result

 

Got my block today, hoppfully i get even better result

Before you install it, maybe you wanna check the card's potential in SW on air first for comparison. 1100 mV, no powerlimit, max GPU clock, 100 % fans. Please clock the VRAM to its limit, Speed Way score is memory dependant.

Highscore so far:

I scored 8 238 in Speed Way

Intel Core i5-13600K Processor, NVIDIA GeForce RTX 4080 SUPER x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

 

well I tested the limits of my card windforce v2 4080S. i couldnt hold 3000MHz for 5 secs. 2985MHz only for 10+ sec and crash. it can do 2970MHz right now.
and I put mems to 12800MHz (25.6Gbps) which all 4080S cards should do, I also tried 26Gbps but no difference in perf.

edit: power used was msot 360W, this is 385W bios. So i dont have to flash 400+W bioses as its not even close to 385W.

 

I got around 8260 in Speedway. The 4080 Super is really an amazing GPU for 4K gaming.

@robalm how is the quality of the EKWB block? I thought their stuff seemed pretty chintzy in the past. It’s also really expensive at like $280.

 

It's mutch better then the alphacool block, that one was so poor that i send it back.
I whould say the ekwb block is great, a bit expensive for sure and i hate that you have to cut tye termalpads yourself.

 

My block also arrived

 

If I ever decide to watercool the card I'm 100% going Alphacool for the block. Had many good experiences with their blocks.

 

I did that, and it was trash
It's like they have zero quality control.
The ekwb block has no scratches, while the alfhacool is full of scratches.

https://www.overclock.net/threads/alphacool-waterblock.1809827/

 

foli10

Please follow up if you mount it.
I have not the time to for it yet.

 

Are you going to use different pads? I always thought the EKWB pads were a little chintzy.

If I order one, I may as well go all out and get some nice Gelid pads too.

 

Sorry I replied on page earlier instead of just posting it.

RE: Alphacool and EK.

From personal experience, Alphacool is still trash I don't think they even check quality of their products before they ship.
As far is EK? I used EK once in 2019 and never again, but my nephew who got their block recently can confirm that what EK was always terrible at and that was their nickle plating prematurely wearing out, especially pealing is still the same years later and dealing with returns! that is another headache.

He send his Quantum Vector RTX 4090 RGB Nickle plexi block for RMA due to nickel turned black and came off clogging his loop, Its been 2 weeks, he still waiting, they might look nice and shiny, but not for long, frankly I'm surprised EK is still in business.

I used Aquacomputer and Heatkiller blocks, Heatkiller more so on one of my Strix 3090 and even on much earlier, older cards e.g Titans Black, Titan RTX and never ever single issue, in fact, after almost 2 years taking one of the blocks of Titan Black after using distilled with Mayhems Biocide, block still looked like new.

 

Heatkiller or Optimus... can't go wrong with either.
Limited compatability tho. Only 90 series cards for optimus and only TUF/STRIX and FE for 4080.

 

If you are happy with "just a block" and nothing fancy you can't beat Bykski or Barrow.

 

Since these are the ones available locally I use them too, never had an issue, even with the CPU blocks, they perform fairly well for the price. Can't really complain.

 

Anyone who has water-cooled for a long time knows by now which brands are to be most likely best in their line of products with lowest failure rate of any kind and easiest RMA process compare to other brands.

In my 2 decades of water-cooling, although I still do air-cool specific systems, but here is my list of blocks, fittings and coolants brands that have performed absolutely superb with 1-2% small issue(s) that were rectified by company(s) within timely fashion and these products in my experience are "buy once" and forget it, unless you upgrading and block(s) are no longer compatible.

GPU and CPU blocks:

Heatkiller
Aquacomputer
Bykski

Fittings always Bitspowers. I have tried Aquacomputer, alphacool.....nahhh....Bitspowers tolerance and quality is by far superior, expensive, but in my opinion worth it.
Liquids Mayhems Biocida with distilled water has been awesome and Aquacomputer Double protect Ultra.

 

I absolutely loved my bitspower summit m pro block for the looks but the performance just wasn't there. On the 5900X I swapped it for a TechN that was far superior. For fluid I always use Mayhems X1 premix in a 5 liter can and for tubing Mayhems Ultra Clear. I had EK fittings, junk. The y splitter especially felt super loose and leaked all the time at the junction o ring..

 

I'm using a 90 degree adapter I bought on Amazon for $20 with my RTX 4080 Ventus 3x. I need it to be able to close the sidepanel. hopefully it doesn't melt.

Amazon.com

 

Hello!

First, sorry for my bad english...

I have a 4080 ichill frostbite with D5 pump and 1080 rad. Temp water doesn't exceed 28°C in game but the GPU is around 60°C . Do you think the waterblock is incorrectly mount?
What do you recommend that I do?

Thank you in advance

 

Hello!

First, sorry for my bad english...

I have a 4080 ichill frostbite with D5 pump and 1080 rad. Temp water doesn't exceed 28°C in game but the GPU is around 60°C . Do you think the waterblock is incorrectly mount?
What do you recommend that I do?

Thank you in advance

What's the hotspot temperature? Check in HWINFO64. That would tell us way more about the quality of the mount.

As soon as I see single D5 + 1080 rad I'm like.. flow rate? My D5 could barely handle a 280 + 420 + CPU + GPU block + 10/13mm tubing + lots of restrictive fittings. Flow rate even at 100% PWM was.. borderline... And if I ran the pump at anything under 80% PWM it would make temps shoot up a lot. How is the loop built fitting wise, lots of 90's, small tubing? CPU block also in the loop?