*Official Benchmark Thread* - Post it here or it didn't happen :D

[jaybee83]

1 hour ago, cylix said:

Oh man think is the best looking 4090 out there and like always hard to get😂

haha i figured this would be right up your alley in all white!

[Mr. Fox]

3 hours ago, Etern4l said:

 

Ha. I'm liking the idea of liquid cooling more and more. The main problem with the external radiator grid I've found so far is aesthetics. Didn't anyone come up with a nice chassis for this, for those who don't have the time/skills/tools to make one?

I really like how my MO-RA3 360 looks. It comes in a nice free-standing housing and has feet to keep it stable. I added the 1/4 inch aluminum wing plates on each side to mount the 240mm reservoirs and pumps. I have an inexpensive ATX PSU mounted on the underside of the desk to power the fans and pumps and a Lamptron hub to control the fans and pump speeds. https://shop.watercool.de/MO-RA3-Series

 

 

 

[Etern4l]

21 minutes ago, Mr. Fox said:

I really like how my MO-RA3 360 looks. It comes in a nice free-standing housing and has feet to keep it stable. I added the 1/4 inch aluminum wing plates on each side to mount the 240mm reservoirs and pumps. I have an inexpensive ATX PSU mounted on the underside of the desk to power the fans and pumps and a Lamptron hub to control the fans and pump speeds. https://shop.watercool.de/MO-RA3-Series

 

 

 

 

Pretty cool, thanks for sharing. Can't really fault this kind of look:

 

With 18 fans this will cool down anything. Nice job with the reservoir extension too, the whole setup has a mildly mad scientist vibe to it. I like 🙂

 

Edit: BTW For those less mechanically equipped I assume there are some external reservoir stands too?

Overall this stuff basically solves multi GPU cooling without limited and dubious top-bottom in-case radiator setups as per LianLI D001 (or something like that).

[Mr. Fox]

18 minutes ago, Etern4l said:

BTW For those less mechanically equipped I assume there are some external reservoir stands too?

There might be but I have never seen any. I purchased the aluminum plates on Amazon already the right width and length so all I had to do was drill four holes in the plates. Two holes to align with the threaded inserts on the radiator housing and then two holes to bolt the reservoir brackets to the plate. I left the aluminum raw and scruffy looking. Had I wanted to spend more time making it fancy I would have wet sanded and polished the plates to a mirror finish. I chose to go with the gritty industrial look as a statement, but it is more of a lame excuse for my laziness.

 

Side note: the reason for the excessive tubing length is I need the flexibility of bypassing the radiator while still having use of the pumps by simply removing and reconnecting two QDC fittings when using the chiller for benching. The extra tubing length is necessary to accommodate that, even though it detracts from the appearance.

 

As a side benefit, the extra tubing length also facilitates moving the radiator or computer around on my desktop without one dragging the other along with it or having to disconnect anything.

[Etern4l]

17 minutes ago, Mr. Fox said:

There might be but I have never seen any. I purchased the aluminum plates on Amazon already the right width and length so all I had to do was drill four holes in the plates. Two holes to align with the threaded inserts on the radiator housing and then two holes to bolt the reservoir brackets to the plate. I left the aluminum raw and scruffy looking. Had I wanted to spend more time making it fancy I would have wet sanded and polished the plates to a mirror finish. I chose to go with the gritty industrial look as a statement, but it is more of a lame excuse for my laziness.

 

I presume that the external reservoir setup makes the maintenance that much easier, or does it not matter?

[Mr. Fox]

16 minutes ago, Etern4l said:

 

I presume that the external reservoir setup makes the maintenance that much easier, or does it not matter?

Absolutely... 150%. Anything that you can move outside of a computer case makes it easier to work on, as space constraints no longer exist. Even with an absurdly gigantic case like my Level 20 XT and the massive tower that @Papusan has, a case eventually presents physical limits of some kind. That is also why I prefer an open bench versus a case. I really miss having everything mounted on an open bench for that reason. I do not miss the added difficulty of keeping it clean in this Arizona dust bowl, but everything is easier to work on when nothing can get in your way or limit access. I can't bring myself to sell the Praxis Wetbench because I can envision eventually going back to it at some point, in spite of the dust problem that caused me to stop using it.

[tps3443]

1 hour ago, Etern4l said:

 

I presume that the external reservoir setup makes the maintenance that much easier, or does it not matter?

I have seen some beautiful Mora3 setups before, so nice in fact it really makes me want to set aside the funds and do it all for nothing more than nice looks. But they are expensive. They sell aluminum and copper variants too with lots of accessories. 
 

I run a Pora3. The Pora3 is the “Poor man’s Mora” 😎. It’s a solid copper 1,080x45mm thick radiator. Holds (18) 120mm fans. And it’s only $122 dollars. I actually have two of them.  But one is plenty overkill. 

Okay, I’m kidding. It’s not really called a “Pora” it’s really a Alphacool Nexxxos 1080, and they sell enclosures with feet as well. It’s pretty cool. After hooking it up, I never looked back. Real God like cooling. 

 

 

 

[johnksss]

To also add to this thread....there may be a 4090 Kingpin as someone found and reported the sku's for it. It was not said who was making it or where, just that there is a sku for it.

[electrosoft]

6 minutes ago, johnksss said:

To also add to this thread....there may be a 4090 Kingpin as someone found and reported the sku's for it. It was not said who was making it or where, just that there is a sku for it.

 

That would rock. Even more reason to wait and see if we can break free of the cookie cutter PCBs that seem to dominate most of the models outside of VRMs.

 

 

[Reciever]

14 minutes ago, johnksss said:

To also add to this thread....there may be a 4090 Kingpin as someone found and reported the sku's for it. It was not said who was making it or where, just that there is a sku for it.

I believe GamersNexus did report that EVGA did have 4090's but they were in development for such purposes. 

 

I would imagine they would be akin to ES samples from Intel unless KingPin is moving to another AIB? 

[johnksss]

10 minutes ago, electrosoft said:

 

That would rock. Even more reason to wait and see if we can break free of the cookie cutter PCBs that seem to dominate most of the models outside of VRMs.

 

 

 

2 minutes ago, Reciever said:

I believe GamersNexus did report that EVGA did have 4090's but they were in development for such purposes. 

 

I would imagine they would be akin to ES samples from Intel unless KingPin is moving to another AIB? 

Not sure at this point....You know how the rumor mill works..... Just because someone says it doesn't make it true....

[electrosoft]

More fuel for the 4090ti fire....

 

"However, Kopite is now sure that there won’t be a new TITAN with such specs. What he referred to before as “The Beast” might end up with slightly altered memory specifications as RTX 4090 Ti. The Ti variant is reportedly not canceled, and it may still happen in the future, he adds."

 

 

 

https://videocardz.com/newz/nvidia-titan-ada-is-reportedly-not-planned

 

 

 

 

[tps3443]

Anyone play A Plague Tale: Requiem yet? 
 

Boy is it power hungry!! Peaking at over 625 watts. 
 

 

This 3090 smashes it though!! 
 

With this kind of power usage, any standard 3090 would just clock flop on you. It would probably tank to 1,700Mhz range.
 

 

[Etern4l]

10 minutes ago, tps3443 said:

Anyone play A Plague Tale: Requiem yet? 
 

Boy is it power hungry!! Peaking at over 625 watts. 
 

 

This 3090 smashes it though!! 
 

With this kind of power usage, any standard 3090 would just clock flop on you. It would probably tank to 1,700Mhz range.
 

 

 

40C 🤣

[Reciever]

Just now, Etern4l said:

 

40C 🤣

I think mine ran around 54c but I just have memory clocked up everything else is stock. 2700x bottlenecked it I think so probably more apt to have it on the 10850k

[Etern4l]

11 minutes ago, Reciever said:

I think mine ran around 54c but I just have memory clocked up everything else is stock. 2700x bottlenecked it I think so probably more apt to have it on the 10850k

 

Also great, I mean either is fine in practice. The recommendation I've heard in the context of mining was to keep the GPU under 70C for any long term use.

 

If we were to draw precise comparisons, we would need a few more pieces of data here: ambient temp, noise level (dBA) and ideally extra electricity required to achieve the given temperature (e.g. someone could be using a 500W cooler/aircon unit).

[Mr. Fox]

22 minutes ago, Etern4l said:

40C 🤣

 

21 minutes ago, Reciever said:

I think mine ran around 54c but I just have memory clocked up everything else is stock. 2700x bottlenecked it I think so probably more apt to have it on the 10850k

4 minutes ago, Etern4l said:

Also great, I mean either is fine in practice. The recommendation I've heard in the context of mining was to keep the GPU under 70C for any long term use.

 

If we were to draw precise comparisons, we would need a few more pieces of data here: ambient temp, noise level (dBA) and ideally extra electricity required to achieve the given temperature (e.g. someone could be using a 500W cooler/aircon unit).

Chiller + radiator keeps the temps from going low enough for condensation, but keeps the GPU cold enough to avoid room temperature thermal throttling that degrades performance even though the GPU is not hot enough to need to be throttled. You start losing clock speeds rapidly and performance begins to decline around 45-48°C and progressively worsens as the temperatures rise.

 

It is unfortunate that Turing and Ampere technology began this aggressive psychotic behavior. There is no reason for the GPU core clocks to move down from maximum boost clock until core temps start pushing past ~75°C.

[tps3443]

42 minutes ago, Etern4l said:

 

40C 🤣

Water is locked to just 70F.

 

22 minutes ago, Mr. Fox said:

 

Chiller + radiator keeps the temps from going low enough for condensation, but keeps the GPU cold enough to avoid room temperature thermal throttling that degrades performance even though the GPU is not hot enough to need to be throttled. You start losing clock speeds rapidly and performance begins to decline around 45-48°C and progressively worsens as the temperatures rise.

 

It is unfortunate that Turing and Ampere technology began this aggressive psychotic behavior. There is no reason for the GPU core clocks to move down from maximum boost clock until core temps start pushing past ~75°C.

I use a switch for my radiator fans to turn them off. Condensation can form really bad. But I would have to set my water temp to about 53-55F to reach condensation.  I usually keep it at 70F liquid temp though.

[Reciever]

2 minutes ago, Etern4l said:

 

Also great, I mean either is fine in practice. The recommendation I've heard in the context of mining was to keep the GPU under 70C for any long term use.

 

If we were to draw precise comparisons, we would need a few more pieces of data here: ambient temp, noise level (dBA) and ideally extra electricity required to achieve the given temperature (e.g. someone could be using a 500W cooler/aircon unit).

Mine stayed around 56c or so but as you mentioned ambients play a role, I also have 2x twin window fans so I my room is negative pressure with the occasional AC from the vent and the system itself is near the window.

 

For the context of mining the GPU die (well for ethereum at least) was that memory was far and away more important to pay attention to. The not very long term minded just ran it as is and killed their cards in 6 months running 110c on the memory. I kept mine around 85-90c when it was in full time use. Tuned to 300w, for gaming I tuned it to 350w as a quick fix because my laptop started giving me issues

 

 

[tps3443]

10 minutes ago, Reciever said:

Mine stayed around 56c or so but as you mentioned ambients play a role, I also have 2x twin window fans so I my room is negative pressure with the occasional AC from the vent and the system itself is near the window.

 

For the context of mining the GPU die (well for ethereum at least) was that memory was far and away more important to pay attention to. The not very long term minded just ran it as is and killed their cards in 6 months running 110c on the memory. I kept mine around 85-90c when it was in full time use. Tuned to 300w, for gaming I tuned it to 350w as a quick fix because my laptop started giving me issues

 

 

If I restore defaults, and limit power to 500 watts. It runs about 30c. Same 1.1V. No under volting or anything. 

 

[Etern4l]

41 minutes ago, Mr. Fox said:

 

You start losing clock speeds rapidly and performance begins to decline around 45-48°C and progressively worsens as the temperatures rise.

 

It is unfortunate that Turing and Ampere technology began this aggressive psychotic behavior. There is no reason for the GPU core clocks to move down from maximum boost clock until core temps start pushing past ~75°C.

 

That's quite unexpected, obviously not something one would be able to observe with air cooling. How much % performance loss are we talking about between say 40C and 65C? @tps3443could you run a quick experiment to roughly quantify this (assuming that @Mr. Fox's temporary 3070 is not WCed)? Could just run a GPU benchmark of choice twice: with 40C temp target and 65C temp target (assuming you can control this via cooling only), everything else being equal.

 

19 minutes ago, Reciever said:

For the context of mining the GPU die (well for ethereum at least) was that memory was far and away more important to pay attention to. The not very long term minded just ran it as is and killed their cards in 6 months running 110c on the memory. I kept mine around 85-90c when it was in full time use. Tuned to 300w, for gaming I tuned it to 350w as a quick fix because my laptop started giving me issues

 

 

 

Good point about the memory temps, I guess the assumption is that memory has higher temperature tolerance. 

The 3090 suffers from a specific issue - I don't remember exactly, but it is a combination of chip types and insufficient memory cooling due to 2-sided card design. I recall that on the 4090 all mem chips are on one primary side of the PCB, and mem temps are much lower.

 

That said, while Ethereum mining was memory bandwidth bound, it would still load the cores to decent temps so the recommendation was probably based on some emprical results in terms of the ratio of failed cards operating in certain conditions.

 

[tps3443]

FUN FUN FUN!!!! Memory at 23GBPS 😃

 

Great performance out of the RTX3090. I think 4K 60+ may be possible on one of these last Gen AMPERE GPU’s. I had all of my dip switches enabled on the back of my GPU. Cut down on power usage DRASTICALLY!! 

[Reciever]

19 minutes ago, Etern4l said:

 

That's quite unexpected, obviously not something one would be able to observe with air cooling. How much % performance loss are we talking about between say 40C and 65C? @tps3443could you run a quick experiment to roughly quantify this (assuming that @Mr. Fox's temporary 3070 is not WCed)? Could just run a GPU benchmark of choice twice: with 40C temp target and 65C temp target (assuming you can control this via cooling only), everything else being equal.

 

 

Good point about the memory temps, I guess the assumption is that memory has higher temperature tolerance. 

The 3090 suffers from a specific issue - I don't remember exactly, but it is a combination of chip types and insufficient memory cooling due to 2-sided card design. I recall that on the 4090 all mem chips are on one primary side of the PCB, and mem temps are much lower.

 

That said, while Ethereum mining was memory bandwidth bound, it would still load the cores to decent temps so the recommendation was probably based on some emprical results in terms of the ratio of failed cards operating in certain conditions.

 

There were some AIB's that used glorified plastic (MSI) on their backplate, then others that had a backplate but no pads. it was a crapshoot imho.

[tps3443]

4 minutes ago, Reciever said:

There were some AIB's that used glorified plastic (MSI) on their backplate, then others that had a backplate but no pads. it was a crapshoot imho.

I know that the 3090FE’s with even just their 350-400 watt power limits. The memory TJMax which is a sensor inside one of the memory modules would run north of 100C-110C. People were re-doing the pads and getting that down substantially to the 80’s which was perfectly fine. Watercooling any of them helps drastically. 
 

Apparently this was fixed a while back though. I think any 3090 now a days is good to go. 

[Reciever]

49 minutes ago, tps3443 said:

I know that the 3090FE’s with even just their 350-400 watt power limits. The memory TJMax which is a sensor inside one of the memory modules would run north of 100C-110C. People were re-doing the pads and getting that down substantially to the 80’s which was perfectly fine. Watercooling any of them helps drastically. 
 

Apparently this was fixed a while back though. I think any 3090 now a days is good to go. 

Yeah I believe that is accurate as well. I knew going in that I would make modifications to the GPU as I bought it for mining and later to be re-used for gaming or other such purposes so its a long term investment kind of thing.