Alphacool GPX AMD rx 580 install options

[Win_Wiz]

My air-cooled (and overclocked) Aorus RX 580 4GB is very noisy when gaming, so I have ordered NexXxoS GPX AMD RX 580 M04
I know this hybrid block is very restrictive so I am wondering how to install the GPX for best (near silent) operation.

Currently my loop consist of these components:

Danger Den MC-TDX (cooling my I5 2500K - 4.5GHz@1.39V)
Black Ice Pro III 360mm RAD (with 6xArctic F12 PWM fans)
XSPC LAING D5 RESERVOIR TOP V3 (including ALPHACOOL VPP655 D5@2650 RPM)
2x10mm ID 90º Push-In fittings
4x10mm ID straight Push-In fittings
About 1.5meters (or 5 feet) 10,3mm OD/7,5mm ID soft tube

The old radiator manages to cool my cpu passive at low loads, when priming my CPU core temperatures peak @ 62ºC.
I really don't want to increase my pumps RPM, because its practically silent @2650rpm.
So should I install the GPX parallel with my CPU block, to limit the restriction?
Or maybe remove the jet plate insert even though it will increase my GPU temp about 10ºC?
Or will my D5 pump handle CPU and GPU blocks in serial, maybe if I replace the 2x90º fittings with 2x45º fittings?
If I go for serial and remove the insert should I place the GPX before the MC-TDX CPU block?

I know I could test all the above options and stick with what works best, but draining my loop several times is time consuming, so I hope someone with more knowledge than me would be able to advise me.
(I have also posted these questions on overclock.net - I hope that's OK?)

 

[eidairaman1]

Running wc requires mx no matter what, either you get heat soaking or noise, rather have parts cool.

Have you thought about a separate pump and rad?

 

[Win_Wiz]

Running wc requires mx no matter what, either you get heat soaking or noise, rather have parts cool.

Sorry English is'nt my native language, would you please explain what mx means?

Have you thought about a separate pump and rad?

That is to expensive, I prefer to spend my limited budget replacing my old and noisy hard drives with a silent 2TB SSD...

This old rad used to cool an overclocked CPU + NB + GPU in pull configuration using a weaker pump, so I don't believe it's necessary.

 

[John Naylor]

Wow, that makes me feel old .... had three teenagers back then.

The D5 is a decent pump and that block I remember as being on of the least restrictive.

http://martinsliquidlab.petrastech.com/DangerDen-MC-TDX.html

You want the full flow going thru the CPU but if you split the upstream of GPUs upstream and recombine, then you cut the back pressure thru the rads to 28% of a single block with half flow thru each. The CPU needs the low because of the small thermal mass / transfer area. A ful cover water block's large mass and contact area gets by fine with half flow ... I have GFC cards only in parallel, splitting before and recombining. At 1200 rpm on the fans, CPU cores average in low to mid 70s and GPU is at 39C. I'm using a Swiftech 35X2 pump. In gaming fans usually peak at 550 - 650 rm at which they are inaudible.

https://martinsliquidlab.wordpress.com/2012/01/29/swiftech-mcp35x2-pump/6/

On a hybrid, you are only cooling the GPU so it suffers to an extent from the same limitations of CPU block. And that'sI think will be your problem ... small thermal mass and large restriction

I woulda suggested a MSI 1080 Seahawk w/ EK full cover water block, which I have seen used for $450 on ebay and they are more than 50% faster.
https://www.ebay.com/sch/i.html?_nkw=GTX+1080+SEA+HAWK+EK+X

A GTX 1060 air cooled would be quieter and faster and the new 1660 Ti is 33% faster and very quiet (0 at idle / 32 dbA at load) but that's no fun when ya have water.

With the 1080 Seahawk (RK Version), you are looking at about 450 theoretical watts of full load about 270 - 330 watts of which would need to be handled by the Radiator, rest allotted to not everything peaking at once and thermal radiation from blocks, rad shrouds, fittings and tubing into the case. Using the Radiator Size Estimators here...btw, that Estimator is based upon data published on MartinsLiquidlab site (old and new links above) ... he did the testing, i did the math to make the spreadsheets.

https://www.overclock.net/forum/61-water-cooling/1457426-radiator-size-estimator.html

A 45mm thick rad w/ (3) 120mm fans in push would generate 272 watts of cooling w/ fans at 1400 rpm / 329 watts in push / pull. Jump to 1800 rpm on the fans and those numbers would be 343 watts / 415 watts

 

[eidairaman1]

John he was asking about a WB and pump layout for his 580, stick within reasonability.

 

[Win_Wiz]

I don't believe GTX 1060 is faster than my RX 580, and 450$ for a used GFX card is way more than I want to spend. Besides I think Nvidia is to greddy so I prefer to support AMD, competition is important!

I just want to get the best out of what I've got.

Regarding my 360 Rad's cooling capacity, doesn't it depend on what deltaT I'm willing to accept?

Regarding fullcover vs. Hybrid. I have cooled several gfx cards with GPU only blocks and heatzinks on voltage regulators+ RAM, always with great results, so I know it's possible. Just decided I would try something new this time -hence the GPX.

 

[John Naylor]

John he was asking about a WB and pump layout for his 580, stick within reasonability.

That was his 1st question ... there were more

1. If your are lifting a 100 pound weight out of a 20 foot hole with (2) 10 long ropes, one rated at 40 pounds and one at 60 ... worrying about what order you put them is not really relevant when no matter what option you chose, the weight stays in the hole.

2. A hybrid is not going to produce silent operation ... order will not change the sound output.

3. "So should I install the GPX parallel with my CPU block, to limit the restriction? " answered.

4. "Or will my D5 pump handle CPU and GPU blocks in serial, maybe if I replace the 2x90º fittings with 2x45º fittings? " answered (hybrid block will have huge resistance at 10 times the usual flow rate.)

5. The last two questions ... won't matter so didn't answer.

I don't believe GTX 1060 is faster than my RX 580, and 450$ for a used GFX card is way more than I want to spend. Besides I think Nvidia is to greddy so I prefer to support AMD, competition is important!

I just want to get the best out of what I've got.

Regarding my 360 Rad's cooling capacity, doesn't it depend on what deltaT I'm willing to accept?

Regarding fullcover vs. Hybrid. I have cooled several gfx cards with GPU only blocks and heatzinks on voltage regulators+ RAM, always with great results, so I know it's possible. Just decided I would try something new this time -hence the GPX.

1. It's all in how you look at it .... outta the box, the 580 is lightly faster than the 1060 6GB. But the 1060 has much more overclocking headroom than the 580. There's no wrong choice provided you looked at all the factors, some will matter to different people, some will not. if you don't OC, then the 580 may be the better choice ... you said sound matters which would be a significant factor, but WC will eliminate some of that ... tho, the hybrids are often louder than air cooled. If you like the 580, no reason to look elsewhere. Only reason I mentioned it was that budget was an issue and you could easily spend more or get less trying to address the issues you mentioned. It's an alternative to consider, make the choice that works for you.

Anyways it's all out the window today, at least as far as comparing new to old, with the release of the 1660 (23% faster than 580 / 10% faster than the 590.) ... however, if buying used, Im gonna bet that if you bought that 580 in the last week, it likely the more cost effective. A month ago, I prolly wuda said different.

2. Delta T is a thing that used to be the 1st thing folks decided when getting in to WC. The general description associated with Delta T is.

Enthusiast Cooling - Delta T = 10C
Reasonably Good Cooling - Delta T = 15C
You can tell ya fiends "I have water Cooling" - Delta T = 20C

10C is used for the Estimator spreadsheet because .. a) it's what we all want if no budgetary or space limitations and, b) more importantly, it's base 10 do math is easy and can do in ya head. Let me give you an example:

You calculate that you need250 watts. The estimator tells you that your rad will produce 200 watts of cooling @ 10C Delta T. So, your anticipated coolant temp will be:

10C x 250 / 200 = 12.5C ... I don't see anyone getting antsy with that result.

3. In most cases, "good results means "I ran it and nothing broke". Look at the graph here

https://cdn.videocardz.com/1/2016/06/MSI-GTX-1080-GAMING-X-clocks-vs-temp.png

You see that at 55C, the card clocked at @ 1936 .... at about 64C it dropped about 12 Mhz to 1924 .... ...then at about 68-69C, it's able to maintain a stable clock speed.each step is another 12 Mhz. The MSI Seahawk 1080 keeps the GPU about 22C cooler than the air cooled card at 48C. That should keep you from seeing to much throttling.... may cards stay between 39 and 42C will full cover.

On the PCB, the MSI Seahawk hybrid hits 59C at the VRMs which is excellent ... many 3rd party hybrid solutions don't do as well, main problem being they are so rarely tested.

I'd like to see Wiz test the MSi 2080 Ti Gaming X, the MSI Seahawk and the MSI Seahawk EK ... but even then, from reviews I have seen in the past, MSI and others who have similar offerings tend to put the hybrid and full cover blocks on one of the lesser PCB designs. I fear that any gain you might get from the better cooling might be erased by the "lesser" PCB. here's an example how the different PCBs affect performance in TPUs OC test.

MSI Lightning Z - 236.7 fps
MSI Gaming X Trio - 226.6 fps
MSI Duke - 220.5 fps

Is it worth opting for the one with the EK block if they are slapping it on the Ventus PCB ?

best of luck, let me know how it turns out, if i miss any responses feel free to nudge me with a PM.

 

[Win_Wiz]

1)
Ok so you don't believe 2 ropes with a combined rating of 100 pounds will actually lift 100 pounds? I actually think it's possible, but you need to be smart, weight destribution is key in your example.

2)
I don't need my PC to be like 0dB silent when gaming, just pretty quiet.
I'm sure hybrid cooling will beast my RX 580 stock cooling in this regard.

3) Running a low restriction CPU block parallel with a high restriction GPU block, would force the water to mainly flow through the low restriction CPU block = less cooling for my GPU. But I'm wondering if it would cause GPU temperatures to increase compared to my cards stock air cooling.
Sadly I don't know how to calculate this and I'm guessing it's rather complex.

4)
I'm a little confused about your comment "huge resistance at 10 times normal flow rates"
Please explain what you mean by this.

5)
So it doesn't matter if I limit my loops flow by using the jet plate insert to cool my GPU to something like 50 degree c versus removing the insert gaining higher flow but also like 10 degrees higher GPU temperature?
I would imagine cooling to 50 degree would induce more energy into my loop than cooling to 60 degree would??

I forgot to mention that my gigabyte aorus Rx 580 is a 6+2 pin card, so power draw is limited by this.
I believe my overclocked i5 2500k consume about 110 watts a full load, so CPU and GPU consumes about 350 Watts @ 100 usage. My monitor is only 75Hz so I use adaptive sync to limit fps @ 74. So while gaming CPU+GPU should output less than 350 Watts combined.

That was his 1st question ... there were more

1. If your are lifting a 100 pound weight out of a 20 foot hole with (2) 10 long ropes, one rated at 40 pounds and one at 60 ... worrying about what order you put them is not really relevant when no matter what option you chose, the weight stays in the hole.

2. A hybrid is not going to produce silent operation ... order will not change the sound output.

3. "So should I install the GPX parallel with my CPU block, to limit the restriction? " answered.

4. "Or will my D5 pump handle CPU and GPU blocks in serial, maybe if I replace the 2x90º fittings with 2x45º fittings? " answered (hybrid block will have huge resistance at 10 times the usual flow rate.)

5. The last two questions ... won't matter so didn't answer.



1. It's all in how you look at it .... outta the box, the 580 is lightly faster than the 1060 6GB. But the 1060 has much more overclocking headroom than the 580. There's no wrong choice provided you looked at all the factors, some will matter to different people, some will not. if you don't OC, then the 580 may be the better choice ... you said sound matters which would be a significant factor, but WC will eliminate some of that ... tho, the hybrids are often louder than air cooled. If you like the 580, no reason to look elsewhere. Only reason I mentioned it was that budget was an issue and you could easily spend more or get less trying to address the issues you mentioned. It's an alternative to consider, make the choice that works for you.

Anyways it's all out the window today, at least as far as comparing new to old, with the release of the 1660 (23% faster than 580 / 10% faster than the 590.) ... however, if buying used, Im gonna bet that if you bought that 580 in the last week, it likely the more cost effective. A month ago, I prolly wuda said different.

2. Delta T is a thing that used to be the 1st thing folks decided when getting in to WC. The general description associated with Delta T is.

Enthusiast Cooling - Delta T = 10C
Reasonably Good Cooling - Delta T = 15C
You can tell ya fiends "I have water Cooling" - Delta T = 20C

10C is used for the Estimator spreadsheet because .. a) it's what we all want if no budgetary or space limitations and, b) more importantly, it's base 10 do math is easy and can do in ya head. Let me give you an example:

You calculate that you need250 watts. The estimator tells you that your rad will produce 200 watts of cooling @ 10C Delta T. So, your anticipated coolant temp will be:

10C x 250 / 200 = 12.5C ... I don't see anyone getting antsy with that result.

3. In most cases, "good results means "I ran it and nothing broke". Look at the graph here

https://cdn.videocardz.com/1/2016/06/MSI-GTX-1080-GAMING-X-clocks-vs-temp.png

You see that at 55C, the card clocked at @ 1936 .... at about 64C it dropped about 12 Mhz to 1924 .... ...then at about 68-69C, it's able to maintain a stable clock speed.each step is another 12 Mhz. The MSI Seahawk 1080 keeps the GPU about 22C cooler than the air cooled card at 48C. That should keep you from seeing to much throttling.... may cards stay between 39 and 42C will full cover.

On the PCB, the MSI Seahawk hybrid hits 59C at the VRMs which is excellent ... many 3rd party hybrid solutions don't do as well, main problem being they are so rarely tested.

I'd like to see Wiz test the MSi 2080 Ti Gaming X, the MSI Seahawk and the MSI Seahawk EK ... but even then, from reviews I have seen in the past, MSI and others who have similar offerings tend to put the hybrid and full cover blocks on one of the lesser PCB designs. I fear that any gain you might get from the better cooling might be erased by the "lesser" PCB. here's an example how the different PCBs affect performance in TPUs OC test.

MSI Lightning Z - 236.7 fps
MSI Gaming X Trio - 226.6 fps
MSI Duke - 220.5 fps

Is it worth opting for the one with the EK block if they are slapping it on the Ventus PCB ?

best of luck, let me know how it turns out, if i miss any responses feel free to nudge me with a PM.

Thank you, I appreciate you answer.
I bought my RX 580 on sale last year for about 150$ -It was cheaper than the GTX 1060 3GB cards.
I sold the included game bundle for 20$. So my RX 580 only cost me 130$ AND it overclocked really well, memory run default at 1800 but overclocks to 2500 MHz. GPU is stable at 1480MHz without any throtteling. This overclock nets 15000 3d marks, so really close to a stock GTX 1070 with my CPU.
So even though it's noisy I like my cheap Rx 580
Watercooled I think 1500 MHz GPU is possible, but I'm not going to push voltages for 20 MHz extra.

How can hybrid cooling generate more noise?
The stock cooling generates segnificantly more noise trying to keep my GPU @ 70 degree than my 6 arctic fan running full speed through my rad!
If the GPX block really increases my computers total noise I'm going to return it...

 

[kapone32]

Water cooling a 580 is a good idea and if you are thinking adding the Eiswolf block willbe fine as it has it's own pump unit. I had a R5 2600 and a Vega 64 cooled on an Eisbaer 420 AIO and the CPU would idle at 26 C and not go over 52 C under load. The GPU would idle at 25 C and not go above 60 C for the GPU and memory. SO I think you should be good

 

[phill]

Depends on the quality of the cooler on the 580.. My 480 Strix was amazing, 100% load, 40C load temps with voltage tweaks and such Couldn't fault it. With only 40% fan which was nice and quiet, even with 6 of them

Still, with a full cover block, it will definitely give much lower temps and noise will be less as well. I'll use my 7970 Lightenings as an example.
Idle temps of 40C and above (depending on ambient temp and fan setting) and then 80C for gaming load (I used Crossfire and triple screen 1080P setup) when I had the cards put under water, temps were idling about the 25C mark, fully loaded and overclocked, 40C load temp...

Says it all really...

 

[kapone32]

Depends on the quality of the cooler on the 580.. My 480 Strix was amazing, 100% load, 40C load temps with voltage tweaks and such Couldn't fault it. With only 40% fan which was nice and quiet, even with 6 of them

Still, with a full cover block, it will definitely give much lower temps and noise will be less as well. I'll use my 7970 Lightenings as an example.
Idle temps of 40C and above (depending on ambient temp and fan setting) and then 80C for gaming load (I used Crossfire and triple screen 1080P setup) when I had the cards put under water, temps were idling about the 25C mark, fully loaded and overclocked, 40C load temp...

Says it all really...

Exactly once you go water cooling on your GPU it's hard to go back and the monstorsity that the Sapphire Nitro Vega 64 (at least 2.5 slots wide) and the fact that there was no way to control the fans and they were set at 50C before they would even spin up made me look for a water block for the card.

 

[Win_Wiz]

My NexXxoS GPX Hybrid water block arrived yesterday.
I was a little surprised that it doesn't cool the VRM chokes, even though my Aorus RX 580 stock cooler has thermal pads on them...
I removed the jet insert and noticed the thermal paste between the waterblock and the alu cooling profile didn't seem to have perfect contact, so I replaced it with a rather thick layer of Grizzly kryonaut.
After mounting the GPX I baked my graphicscard at 150 C for 20 minutes, and re-tightened the skews while the card was hot!
Without the jet insert the GPX cooler manages to keep my overclocked rx 580 at 40 C running the super position benchmark, and my CPU temperature doesn't seem much affected by the increased thermal workload on my single waterloop. For normal light usage like browsing or watching movies the loop manages to cool my PC at 0 rpm -So that is very nice.
VRM1 is also very cool but VRM2 max at 75 C. I think 75 C is ok but not great.
I decided to watercool my GPU because the stock cooler was very noisy at 2900rpm. With watercooled CPU+GPU and no mechanical hard-drives my PC is now very silent.
Overall I am very pleased with the GPX hybrid cooler.

 

[lsevald]

Nice! I got a NexXxos GPX on my 1080ti, the only block I could find for my Palit card. It keeps the GPU temps in the 30-40°C range full gaming load (dedicated loop, 360rad and Laing D5), while OC'ed to 2012MHz with 1.093Vgpu and 140% power target (350W). I have yet to see any throttling, so I'm really happy with it. I didn't change anything on the water block. I wasn't aware of a jet insert or thermal compound between the block and alu profile. Do all NexXxos GPX's have this? Just wondering if I should take a look The jet insert might be fine since I'm using a dedicated loop for the GPU?

 

[Win_Wiz]

Im pretty sure all nexxxos gpx's have the removable jet insert. That's what's make them so effective and restrictive. Removing it causes GPU core temp to increase about 10 degrees C. For silent operation I prefer to run my d5 at speed setting 2, so I decided to take the hit on temperature to gain flow. I don't have a flow meter but I'm guessing the GPX is 3 times less restrictive without the insert.
Regarding the the less than perfect contact, all gpx's suffer more or less from this problem.

 

[Win_Wiz]

I could not resist tweaking my rig a little bit.
With the nexxos installed my RX 580 is stable at 1550MHz, but needs 1.24V. So I have mounted a 120mm fan on the nexxos above the VRM2 chokes. Now VRM2 temp. peaks at 78C running furmark, but runs below 70C when benchmarking and gaming. The fan also seems to assist cooling the GPU. Even at 1550MHz/1.24V max. GPU temp is 42C. -Without the jet insert
I modified my RX 580 bios with Uber Samsung timings and surprisingly my Rx 580 4GB ram seem stable at 2800MHz!
HWinfo doesn't report any memory errors and I don't see any artifacts but performance seems best at 2400MHz, so I'm guessing AMDs driver automatically increase the timings at high frequencies??
I also sound deadened my pump and cabinet with materials made for sound deadening cars. All my fans are pwm controlled with speedfan, so they run 180rpm at low loads up to 850rpm at full load. The PSU fan runs at 650-1000rpm depending on load. So I only hear a pleasant hum when my PC is working hard

 

[smamo]

I could not resist tweaking my rig a little bit.
With the nexxos installed my RX 580 is stable at 1550MHz, but needs 1.24V. So I have mounted a 120mm fan on the nexxos above the VRM2 chokes. Now VRM2 temp. peaks at 78C running furmark, but runs below 70C when benchmarking and gaming. The fan also seems to assist cooling the GPU. Even at 1550MHz/1.24V max. GPU temp is 42C. -Without the jet insert
I modified my RX 580 bios with Uber Samsung timings and surprisingly my Rx 580 4GB ram seem stable at 2800MHz!
HWinfo doesn't report any memory errors and I don't see any artifacts but performance seems best at 2400MHz, so I'm guessing AMDs driver automatically increase the timings at high frequencies??
I also sound deadened my pump and cabinet with materials made for sound deadening cars. All my fans are pwm controlled with speedfan, so they run 180rpm at low loads up to 850rpm at full load. The PSU fan runs at 650-1000rpm depending on load. So I only hear a pleasant hum when my PC is working hard

I think you are correct with regards to memory timings being automatically changed based on frequency. I notice some very strange behavior when overclocking my RX 580, including BIOS modification to adjust memory straps. I can't explain what is going on, but the drivers seem to be doing something against my will. The best success I have had is setting the memory frequency in the BIOS to be exactly what I want and not messing around in Wattman or any other overclocking software.

 

[Win_Wiz]

Yeah the driver seems to auto tune the timings -and the auto tune actually seems to work pretty well on my card. Timings level 1 are loose timings, level 2 are tight (fast) timings. So my advise is try level 2 and raise the memory speed until you get memory errors in HWinfo. My 4GB (Samsung) ddr5 is error free at 2150MHz with level 2 -But some cards apparently generates errors at anything above 2000MHz with level 2 timings. If your vmem generates errors at less than 2100MHz level 1 or auto might perform better...

 

[eidairaman1]

As memory clocks go up timings get looser for stability.

 

[smamo]

As memory clocks go up timings get looser for stability.

Right, but it doesn't seem to follow the straps I set up in the BIOS. It's almost like there is some fail-safe setting that it reverts to. I also find that overclocking the memory in software does not always apply, even though applications report the new frequency. The only way it is definitely faster is if I set the default clock in the BIOS and restart. I can't tell you exactly what's going on, but I have noticed some very odd behavior with my RX 580 when overclocking the memory.

 

[eidairaman1]

Right, but it doesn't seem to follow the straps I set up in the BIOS. It's almost like there is some fail-safe setting that it reverts to. I also find that overclocking the memory in software does not always apply, even though applications report the new frequency. The only way it is definitely faster is if I set the default clock in the BIOS and restart. I can't tell you exactly what's going on, but I have noticed some very odd behavior with my RX 580 when overclocking the memory.

@MrGenius @NdMk2o1o Ndmight be able to help

 

[NdMk2o1o]

How did you come to 2800 unless you have got the silicon lottery then you'll be hard pressed to reach anywhere near those speeds on a 4gb 580, my 4gb can only oc to around 2100 on the vram with voltage albeit I have elpida vram and not Samsung but even then you're trying to oc from 8gbps to over 11gbps. I think that's why the performance drops off as its not stable at 2800 and the driver is reducing the clocks in internally, such can happen if you oc to high in some cases rather than just crash or artifact it will run at reduced performance.

Rather than oc it to its limit with a bios mod id find your max oc using ab or any other tool you prefer then bios mod it slightly lower for everyday use. For instance I can oc to 1550 core and 2100 vram but have left some headroom with my mod and run 1450/2000 then I can game on that all day long and just bump up the clocks further with ab when I feel like benching etc

Just my 2 cents.