- last edited on by
ROGBot
If you happen to be as foolish as I am and give in to the devil on your shoulder to upgrade a pair of the same cards from pascal then I’ll be offering up some food for thought. Whether it’s worth it is of course always subjective. Some will submit that it’s worthless because nothing will use it or their GTX 460 works just fine. That’s not what this is about.
The obvious is this is not a bargain basement venture. These cards are expensive thanks to crypto miners. I paid just over half as much for my 1080Tis that are outrageously priced now too. A lot of price gouging going on with these cards at some sources going well over MSRP. For the record here these were purchased from the egg for $1399 each, the lowest I have seen. Some butt fleas are $1800+. *The premium is already bad enough, don’t need serious gouging going on with it. *These Strix cards are $200 more than the Nvidia FE that can be bought *directly from Nvidia for $1199. *So needless to say the price point is enough to turn back the squeamish.*
Firstly these puppies DRAW A LOT of POWER! *I haven’t graphed it yet but my 1500W (not kVA) online dual conversion UPS for the first time heated up, fans went to jet engine levels*and I looked The display and saw 1200 watts output! *This is of course the entire load with CPU lightly loaded. In the coming weeks I’ll attempt to retrieve a few of my UPS logs to see what’s what.*
With power comes heat. If you intend to run even one of these on air make sure you have beyond excellent air flow. The ENORMOUS heat sinks may pull the heat off the GPU but it just dumps it in your case. Enough so that it raised the temp of my coolant by 4C and it was still climbing before I was like why are my fans ramping up when these aren’t under water yet. So for now when I’m loading them the case door is open. With 4 slot spacing the top card starves for air and becomes too hot to touch the back plate. I could have made a grilled cheese sandwich on it. The specs say 2.7 slots but they may as well just concede to 3 slots because the PCIE slot that is available with 2 slots is completely inaccessible. *By far the largest hestsinks on a GPU I’ve ever seen. I’ll post up pics in an update comparing to the Strix 1080Ti O11G.
Power limit is very real and you will hit it before you hit what these cards are capable of. *I’ve gotten better results with lower Vcore so far just because thats keeping me out of power limit. My guess it this is to keep the heat at bay. *On a cold start I’ve gotten 2.1GHz without throttling but it’s short lived. An honest 2050 though. Which where I am right now on air is +150 core and +300 VRAM, no more than 50% on the Vcore and power limit to max.**
With the cons out of the way these cards provide some impressive results. I’ve bested all my marks and the 4 spot on Realbench. I’ll be posting up more later with some results once I get them under water. *Right now the biggest take away is if you want to try 2 of these make sure you have a power supply big enough. Like 1500W minimum as 1200 won’t cut it. *And if you are thinking about doing it on air.......DON’T!!! Biggest problem is with 4 slot spacing (only a handful of cards can offer more) there is just not enough air getting to the top card even in open air.*
One more point to make for this post, more posts to follow, these cards are saturating Gen 3.0 PCIEX8 bandwidth which up to pascal just didn’t happen. It’s not by a lot but if you don’t have a 16X slot for each card *expect to lose 5% on high workloads like 4K and higher. Doesnt matter for 1080 and about 3% on 2K.*
Look at these honking things. Don’t mind the temp bypass.
***
The obvious is this is not a bargain basement venture. These cards are expensive thanks to crypto miners. I paid just over half as much for my 1080Tis that are outrageously priced now too. A lot of price gouging going on with these cards at some sources going well over MSRP. For the record here these were purchased from the egg for $1399 each, the lowest I have seen. Some butt fleas are $1800+. *The premium is already bad enough, don’t need serious gouging going on with it. *These Strix cards are $200 more than the Nvidia FE that can be bought *directly from Nvidia for $1199. *So needless to say the price point is enough to turn back the squeamish.*
Firstly these puppies DRAW A LOT of POWER! *I haven’t graphed it yet but my 1500W (not kVA) online dual conversion UPS for the first time heated up, fans went to jet engine levels*and I looked The display and saw 1200 watts output! *This is of course the entire load with CPU lightly loaded. In the coming weeks I’ll attempt to retrieve a few of my UPS logs to see what’s what.*
With power comes heat. If you intend to run even one of these on air make sure you have beyond excellent air flow. The ENORMOUS heat sinks may pull the heat off the GPU but it just dumps it in your case. Enough so that it raised the temp of my coolant by 4C and it was still climbing before I was like why are my fans ramping up when these aren’t under water yet. So for now when I’m loading them the case door is open. With 4 slot spacing the top card starves for air and becomes too hot to touch the back plate. I could have made a grilled cheese sandwich on it. The specs say 2.7 slots but they may as well just concede to 3 slots because the PCIE slot that is available with 2 slots is completely inaccessible. *By far the largest hestsinks on a GPU I’ve ever seen. I’ll post up pics in an update comparing to the Strix 1080Ti O11G.
Power limit is very real and you will hit it before you hit what these cards are capable of. *I’ve gotten better results with lower Vcore so far just because thats keeping me out of power limit. My guess it this is to keep the heat at bay. *On a cold start I’ve gotten 2.1GHz without throttling but it’s short lived. An honest 2050 though. Which where I am right now on air is +150 core and +300 VRAM, no more than 50% on the Vcore and power limit to max.**
With the cons out of the way these cards provide some impressive results. I’ve bested all my marks and the 4 spot on Realbench. I’ll be posting up more later with some results once I get them under water. *Right now the biggest take away is if you want to try 2 of these make sure you have a power supply big enough. Like 1500W minimum as 1200 won’t cut it. *And if you are thinking about doing it on air.......DON’T!!! Biggest problem is with 4 slot spacing (only a handful of cards can offer more) there is just not enough air getting to the top card even in open air.*
One more point to make for this post, more posts to follow, these cards are saturating Gen 3.0 PCIEX8 bandwidth which up to pascal just didn’t happen. It’s not by a lot but if you don’t have a 16X slot for each card *expect to lose 5% on high workloads like 4K and higher. Doesnt matter for 1080 and about 3% on 2K.*
Look at these honking things. Don’t mind the temp bypass.
***
Well *block installation got stalled. Seems performance PCs warehouse staff picked the wrong PN and sent me 1080Ti blocks. They sent me a shipping label to return which I already shipped and they already shipped out the correct blocks so tomorrow.....
jelome1989 wrote:
You bought the Barrow block right? Keep us posted, interested on the results. Would probably settle for Barrow or Bitspower. I've stalled my build for 2months+ now. Can't afford to wait til March to watercool my Strix.
Also does the backplate matter? The Barrow block is compatible with the Strix backplate right? Does it come it contact with the PCB?
EDIT: OK just checked my own Strix and it seems it doesn't.
Also just saw this review:
https://www.techpowerup.com/reviews/ASUS/Water_Cooling_the_Asus_GTX_1080_Ti_Strix/8.html
Is that for real? 5c+ drop in VRM temps due to a better backplate?
If the backplate has a ridge and thermal pads to pull the heat from the back it will certainly help to some extent as you are now using the back plate as a heatsink. The Strix 2080Ti regardless or version has a backplate that is simple an aesthetic cover. It’s mounted on standoffs and makes zero contact with the board. I’m tempted to put some thick thermal pads on the back of the VRMs so the stock backplate exhibits some thermal properties.
If you have to make a choice right now and can find them I’d go for the bitspower. I’ve had that block in hand and mounted it. It made good contact with all points. I want *overly impressed with the plexi design and can’t tell you how well thermal dynamics are but it’s a rolls Royce compared with the barrow rusted out VW. Ive not seen any positive reviews on the Byski and seeing how they are more expensive than an EK it just doesnt make sense. They seem to appeal to the Asian market but not in the US.
I’ll certainly keep you posted to tell you Do NOT buy the barrow block!!!!!
The engineering is suffering, in not only the block and how it makes contact with components but also the fact the because the G1/4 terminals are shifted so far to the left one cannot use an Nvlink bridge, can’t mount it. I’ve already contacted the seller Performance PCs and how they react to this after having sent the complete wrong one already which they made good on will make or break my reviews and feed back from them. The bits power block did not have the short comings this one has. I would have stuck with the bitspower if I could source another. Looks like I’m just going to have to learn patience. I’m going to research the Byski blocks and look very closely at the blocks cold plate.
In the images below you can see that only the chip, memory and VRMs on the outboard side made contact. VRMs toward the output side not only has gaps in the plate by design but also only made partial contact. None of the chokes did althoough thicker thermal pad may have resolved that. The instructions are non existent, not that I need them as I’ve done so many I can do these in my sleep, they show a smaller card and different block and only 4 screws. At least the bitspower shows all the correct hardware and you have to figure it out. The VRMs closer to the output side are cooled passively by contact with the support frame by factory intsallation as is VRAM. I have to say I wasn’t expecting a perfect finish but burrs that were not cleaned up in the block where holes were drilled and tapped for the Plexi top left me bleeding. Not that anything ever looks as good as the marketing but this misses by a fairly large margin. Parts that are chromed are done so very poorly and the cheap wreaks from within. I conceded on aesthetics but will not on build and performance. Here’s a few shots, I’ll snap more before I send these back or to the junk yard, depending on Performance PCs reaction.
Firstly card layout. The VRMS circled prettyy much get nothing from this block. For whatever resaon the part that is supposed to make contact is cut out not even present on 3 of these VRMs for reasons unknown. Ill show that in a bit. This is actually a beautiful card all nekked! both front and back....
Moving on you can see that the only part of the block that actually makes contact as it should is the chip and the VRAM (although not much of a pressure imprint on the VRAM)
The main VRMS on the outboard side have some of them making contact, not very well, but not all. Same for the chokes, just a few even touch. Both should have solid imprints all the way down. This may or may not be corrected with thicker thermal pads.
Heres the left side where the block is cut out. The spots you see where there is an additional layer of thermal pads there is no metal, its been machined all the way down flush with the rest of the block. On on end there is a tiny little part that I wonder why they even bothered as it wouldnt cover 10% of a single VRM. One choke made contact.
Here is their stock image, note the area circled in red that omits coverage of 3 of the VRMs. One end doesnt even start until its past a VRM then the other has that little bitty piece that covers nothing, Ill upload a better image later.
Nice pre installed gouge from tooling under the plexi. Almost looks like a crack but Im not 100% sure without removing the block. Makes one think whats present where you cant see?
Now for the real deal breaker. No NVlink possible. I looked at the Bitspower block and it appears the G1/4 terminals are far enough to the right that this isnt a problem. Its a huge problem on this block. I looked at other links and most of them are even wider at this point so its not a matter of the ASUS ROG Nvlink
The engineering is suffering, in not only the block and how it makes contact with components but also the fact the because the G1/4 terminals are shifted so far to the left one cannot use an Nvlink bridge, can’t mount it. I’ve already contacted the seller Performance PCs and how they react to this after having sent the complete wrong one already which they made good on will make or break my reviews and feed back from them. The bits power block did not have the short comings this one has. I would have stuck with the bitspower if I could source another. Looks like I’m just going to have to learn patience. I’m going to research the Byski blocks and look very closely at the blocks cold plate.
In the images below you can see that only the chip, memory and VRMs on the outboard side made contact. VRMs toward the output side not only has gaps in the plate by design but also only made partial contact. None of the chokes did althoough thicker thermal pad may have resolved that. The instructions are non existent, not that I need them as I’ve done so many I can do these in my sleep, they show a smaller card and different block and only 4 screws. At least the bitspower shows all the correct hardware and you have to figure it out. The VRMs closer to the output side are cooled passively by contact with the support frame by factory intsallation as is VRAM. I have to say I wasn’t expecting a perfect finish but burrs that were not cleaned up in the block where holes were drilled and tapped for the Plexi top left me bleeding. Not that anything ever looks as good as the marketing but this misses by a fairly large margin. Parts that are chromed are done so very poorly and the cheap wreaks from within. I conceded on aesthetics but will not on build and performance. Here’s a few shots, I’ll snap more before I send these back or to the junk yard, depending on Performance PCs reaction.
Firstly card layout. The VRMS circled prettyy much get nothing from this block. For whatever resaon the part that is supposed to make contact is cut out not even present on 3 of these VRMs for reasons unknown. Ill show that in a bit. This is actually a beautiful card all nekked! both front and back....
Moving on you can see that the only part of the block that actually makes contact as it should is the chip and the VRAM (although not much of a pressure imprint on the VRAM)
The main VRMS on the outboard side have some of them making contact, not very well, but not all. Same for the chokes, just a few even touch. Both should have solid imprints all the way down. This may or may not be corrected with thicker thermal pads.
Heres the left side where the block is cut out. The spots you see where there is an additional layer of thermal pads there is no metal, its been machined all the way down flush with the rest of the block. On on end there is a tiny little part that I wonder why they even bothered as it wouldnt cover 10% of a single VRM. One choke made contact.
Here is their stock image, note the area circled in red that omits coverage of 3 of the VRMs. One end doesnt even start until its past a VRM then the other has that little bitty piece that covers nothing, Ill upload a better image later.
Nice pre installed gouge from tooling under the plexi. Almost looks like a crack but Im not 100% sure without removing the block. Makes one think whats present where you cant see?
Now for the real deal breaker. No NVlink possible. I looked at the Bitspower block and it appears the G1/4 terminals are far enough to the right that this isnt a problem. Its a huge problem on this block. I looked at other links and most of them are even wider at this point so its not a matter of the ASUS ROG Nvlink
Ive fired off an email to EKWB to see if they have any answer I keep hearing soon but that story has been out since the card was released. I threw in some brown nosing tactics in hopes of maybe at least an idea of how much longer Ill be on air. Watercool is always extremely late so I dont expect anything from them until 3080Ti is released. Early adoption which in my case wasnt that early has its pitfalls. I only hope that my trials and tribulations help out others with their decisions. Like I said if there was better availability with the bittspower I would have stuck with that, im shooting myself in the foot about now for returning the one I had to microcenter that they actually had listed very cheap at $119. Just couldnt find another and I wanted to wrap this build up. One big take away is no matter who says that bitspower and EKWB got their designs from barrow is either intimately affiliated with a barrow rep, cashing out or just plain on the crack pipe because Im here to tell you this is the most poorly designed block of any type I've ever had the displeasure of falling prey to.
@Raja
Can you divulge the reason why the two chokes were turned 90 degress? This seems to be the only reason this card needs a specific water block, otherwise it would be on the compatible list with the rest for the EK Vector.
This is making me wish I still lived down south where I had a buddy with a bridgeport milling machine. He would notch out for those chokes in a skinny minute. I cant see it taking much if the support bracket can clear the chokes and still make good contact with the mosfets then it can be that big of a deal even if one just machines the entire choke channel a little wider.
Can you divulge the reason why the two chokes were turned 90 degress? This seems to be the only reason this card needs a specific water block, otherwise it would be on the compatible list with the rest for the EK Vector.
This is making me wish I still lived down south where I had a buddy with a bridgeport milling machine. He would notch out for those chokes in a skinny minute. I cant see it taking much if the support bracket can clear the chokes and still make good contact with the mosfets then it can be that big of a deal even if one just machines the entire choke channel a little wider.
.jpg "Raja"){kind=avatar}
