Aurora R9, DIY project

@GTS81 @Anonymous @HanoverB 

I tried H80v2 with ML120PRO as push fan but the result is actually worse than my stock cooler+push-pull fan. 

Once I started PC, I noticed the idle temp is between 35 and 40, while my previous temp is around 30, also the case temp is much higher, sitting above 40 all the time. 

I also re-seated the block once just in case the thermal paste was not applied well but the result is similar. Under stress test, there was no obvious improvement on temp either. 

There might be a few reasons for this:

1. I feel the air flow coming from top is very weak. I even took top cover off when testing, not comparable with my push-pull setup with stock rad with top cover on.  I think Push-pull setup is a must for such thick rad. 

2. The tube blocks the air flow ,causing the case temp higher than stock.  The case temp has big impact on CPU performance as I said in my previous post. There is no good solution for H80v2 tubes since they are quite thick and a bit too long for our case, that is why the stock cooler has short tubes.3

3. Maybe the pump is not seating well ? not sure about this, but based on the paste pattern on the copper plate before reseating, it had good contact. 

I will have a break for a few months. I spent more time with my R9 than my wife recently, starting causing some issues. I will keep this H80v2 since I want to do more test in the future. 

A few more thoughts:

 The liquid temp was stting around 34 degrees constantly with H80v2. No wonder CPU idling temp cannot drop to 30, further indicating one 25mm ML120PRO pushing is not enough for this thick rad. IMO, Two ways might improve : Push-pull setup with 25mm+15mm or using stock 38mm fan as push only.I forgot to test 15mm+25mm last night, which might help a bit with the liquid temp but I highly doubt it will improve temp a lot under stress test, because, in my test, even with liquid temp sitting at 35, CPU still reach over 80 with 8 cores running at 4.7ghz, yielding about the same performance as my previous setup.

Do you know how to put 40mm fan around VRM area of CPU ? Any good way to secure it ? @GTS81 @HanoverB @Anonymous  Thanks

@Anonymous 

I have not tried flipped the fan but that would increase case temp probly. Ideally, cold air should come from bottom and front, then hot air exits from top and back, which would have better air flow in the case since hot air is supposed to rise. I guess we cannot get perfect setup overall in our case, which needs to be decided based on personal preference. 

@coldfish_91 : And that, is the reason why I was away from the community all day today. Between setting up my new TV, installing a hanging cabinet below it, and still keeping the family happy, I had to give up time with you guys here! 

Hope all is well now/ will be well soon. I remember when I messed around with my GPU mod and later cable sleeving, my wife wanted to call the feds on me because she thought I was trying to run some underground Internet crime ring.

@Anonymous mentioned the flipping of the CPU radiator fan(s) because he was curious as he never had to mess around with measly AIO radiators like us. On that note, I got to see some of those so called full tower cases today at local Fry's (Thermaltake 20). It is BIG.

@Anonymous I tried add 40mm like this but the difference on temp is neglectable.  The 40mm fan only has about 4 CFM.

I will see if I can fit a larger fan (70mm-92mm) on top of cooler like "cryorig hybrid liquid kit ". 

VRM fans cannot reduce cpu temps when you have an AIO and thats not the purpose for them.

@coldfish_91 added the VRM fan because @Anonymous answered his earlier question:

Do you know how to put 40mm fan around VRM area of CPU ? Any good way to secure it ? @GTS81 @HanoverB @Dell630i  Thanks

But yeah, it's not gonna change the CPU temp. Maybe in some infinitesimal way by smoothing out the Vcore curve. 

  I see.  Do you think putting a bigger fan on top of top would help CPU reach higher freq without thermal throttling ?  I really want to reach 8 cores 5.0Ghz with current setup. 

@r72019 @Anonymous @GTS81 

@coldfish_91 :

I see.  Do you think putting a bigger fan on top of top would help CPU reach higher freq without thermal throttling ?  I really want to reach 8 cores 5.0Ghz with current setup. 

You can always replace the push ML120 Pro fan with the Noctua IPP fan that @r72019 uses. That is a much burlier fan and closer to static pressure strength like the fan that is shipped with your R9. Or just try putting back the stock fan.

When looking at bigger radiators + stronger fans, you are trying to solve the issue of removing heat from the system. Assuming that you have not looked into optimizing the other end of the problem which is the introduction of heat into the system. How is your CPU tuned in XTU? You did mention 4.8GHz @ 85C. Can you please elaborate more on these?

• Any AVX offset?
• What is the PL1 power limit?
• What is the PL1 duration?
• What Vcore are the 8 cores running at to hit 4.8GHz?
• Are you controlling Vcore manually i.e. fixing the Core voltage?
• What is the power drawn at 4.8GHz?
• Any Vcore offset?

Looking at how well you pulled off all the mods so far, I believe you should be well versed enough with OC-ing. If not, just a reminder that you can do real damage to your CPU with OC. Throttling technology is pretty advanced nowadays but let's just practice some care. Personally I won't feel comfortable going beyond 1.35V Vcore. The 14nm transistors especially the fast switching ones turn on at pretty low voltages so you'd be deep into overdrive by the time you go beyond 1.3V.

@GTS81 Unfortunately, I am not quite familar with CPU/GPU overclocking. This is actually our first PC at home. 

I leave everything by default with only using AWCC to OC CPU.  As shown on XTU, PL1 power limit duration is 28s. No Vcore offset or AVX by default.  Max power drawn for 8 cores at 4.7/4.8ghz is about 185W, after 28s, it will power throttle to about 140W and 8 cores at 4.4ghz.  I used Auto managed voltage in AWCC.  For Vcore, since it is jumping around all the time, I am not sure how to read it while doing test, average value maybe ?   The max voltage I have seen is about 1.42-1.45 on one or two cores when running test with 8 cores at 4.7ghz.  

I tried to use manual voltage but it has to pass the test in AWCC in order to enable it.  Most of time, it will fail so I just use auto-managed voltage for peace of mind. 

My BIOS has already been updated.... The manual voltage option keeps failing to pass the test in AWCC. Will that mean the voltage has been locked up ? @GTS81 

@coldfish_91 :

My BIOS has already been updated.... The manual voltage option keeps failing to pass the test in AWCC. Will that mean the voltage has been locked up ? @GTS81

So you mean sliding the voltage value even a little up or down will not be allowed? If so, then yeah, it's locked up. But I don't expect it to lock for over volting. Still it sucks because we need to undervolt to get the Vcore headroom to reduce the power consumption. That's what my entire previous write up was for.

@coldfish_91  Thanks for the heads-up regarding Plunder Volt.  Will have to do some research as it's the first time I'm hearing about it. Hopefully they didn't roll it into 1.0.20 BIOS update because then I'm borked....

@coldfish_91 :

I use a combination of AWCC OC Control + XTU for overclocking. First I choose an OC profile from AWCC which is guaranteed to be stable. Then I switch over to XTU.

In XTU, I run a baseline small FFT using Prime95 with the settings OCC gave me. Here, you can see my screenshot:

With only OCC default, I'm getting that fast-for-28s-then-slower performance like you. Here are the key values:

1. First 28s

• All cores 4.5GHz for AVX.
• Vcore 1.33V
• Temp 93C
• Package TDP 184W.

2. Sustained

• All cores 4.28GHz for AVX.
• Vcore 1.222V
• Temp 79C
• Package TDP 140W (limit by PL1 setting. Default for my CPU is 95W ).

Under "Advanced Tuning", I started moving the "Core Voltage Offset" down slowly. Start with -0.01 V and move downwards. I stopped at -0.1V. Also, I moved by "Turbo Boost Power Max" to 150W. This way, I get full tap 210W for 28s, and then 150W sustained after. Then I reran Prime95.

First, I don't even have to breach 140W and never had the distinction between PL1 and PL2. Here are the key metrics:

• All cores 4.5GHz AVX
• Vcore 1.219V
• Temp 80C
• Package TDP 138W

Try tuning the offset and let us know how it goes. Just a reminder that if you go too far, the system will crash, and you will start from scratch. And the way the R8 crashes, it goes through a series of reboot/ black screen/ blue screen. If you find yourself seeing more than 2 blue screens while attempting boot, hit the F2 key to get into BIOS and restore optimized defaults. You lose all OC, even the AWCC OCC but that gets you back on stable footing.

As PSA to all, especially @amstel78 and @r72019 who does some OC, we have a new potential problem in our hands in the name of Plunder Volt. The security patch coming from Team Blue is supposedly meant to "lock voltage at BIOS". Whatever that means. I'm not downloading that latest BIOS until I see some reports from OC forums on the impact of this.

@amstel78 Hopefully they didn't roll it into 1.0.20 BIOS update because then I'm borked....

You can always roll it back, from Dell drivers and support, prior versions, then 1.0.19.  I'm still on 1.0.18, with no present intention of taking a future update.