[COMPLETED] 2200G with external brick on a budget

Flumper Dinkle

Trash Compacter
Feb 26, 2018
48
70
Here is an update on this build.

Over the last couple of weeks I replaced my monitor with a LG 25UM58 21:9 ultrawide. The GPU output resolution increased from 1920x1080 to 2560x1080 but with around 6% loss in display area (see Displaywars link).

Though conventional wisdom would suggest at least a 29" display for ultrawides I decided to try the smaller screen due to limited desk space. Though initially I found it difficult to work with the smaller sized print it only took a couple of days to become accustomed to it. It is quite a nice display and productivity stuff, like working with spreadsheets, is much easier to do.

Trying out Team Fortress 2 on this display caused temperatures to climb more than I expected to 85'C. I guess the GPU is working a lot harder than before on account of the 30% increase in resolution. It also did not bode well as TF2 is not the most graphic intensive game I think the GPU can handle. In addition I got a blue screen crash with GPU error suggesting that, despite all the Prime95 stress testing, the overclock was not stable.

In order to tame temperatures and give the GPU more headroom I succumbed and lowered the CPU to 3.3GHz, which under LLC Mode 8, caused the maximum VCore to top out at 1.24V. The GPU/SOC voltage is now reduced to 1.1875V and frequency back at 1500MHz.

I've also been progressively tightening memory timings and found that the HyperX's were stable at 3033@CL14. However I reverted back to the sweet spot of 3200MHz and am now testing timings at CL15.

I purchased a 4010 exhaust fan for just $1.50 from Aliexpress but it had an incredibly annoying whining noisy, so I put it away. Better 40mm fans may be needed. The 4010 did not improve the CPU and System temps much but I think it could be useful to circulate air to other parts of the case (HDD and PSU) and minimize hotspots.

The intake 6010 fans work very well but I found diminishing returns when the RPMs climbed more than around 3800 when gaming. So I capped it at a constant 3800 when in a 70 to 88 degree range. This provides a steady constant low noise which is much less annoying compared to the ramping up and slowing down noise these fans make. I also setup a steep climb to 100% (6000 RPM) beyond 88, but this temperature is hard to touch now. Overall gaming is a lot more comfortable on the fan noise front.

I'll be buying a couple of better quality 2 or 3pin 4010s to help with exhaust. I'll also need a Molex D type to fan splitter to power it up. Will take a couple more weeks I guess.
When you were initially stress testing using Prime95, did you also run any GPU stress tests simultaneously? If not that's maybe how you never got a bsod back then but you are now. However if your new settings are proving stable for your needs then problem solved.
If you're concerned about fan noise I'd really recommend the Noctua 40x10 fans. I'm now running 3 at full speed and they're very quiet
 

Shahmatt

SFF Lingo Aficionado
Original poster
Sep 6, 2017
101
53
When you were initially stress testing using Prime95, did you also run any GPU stress tests simultaneously? If not that's maybe how you never got a bsod back then but you are now. However if your new settings are proving stable for your needs then problem solved.
If you're concerned about fan noise I'd really recommend the Noctua 40x10 fans. I'm now running 3 at full speed and they're very quiet

Thanks for your reply.

Ah no. I did not stress the GPU with Prime95. I was more concerned about CPU temps. But what I did do was play TF2 a few times and also run Passmark to check stability. It took a 3rd or 4th hour of play to bring out the instability. Funny how these things work.

Yes the Noctua 4010s are very quiet but they do cost a bit. The 4010 cheap fan I purchased was actually just to test effectiveness of exhaust before investing proper - and they do have some impact, but not very much. I've also been quite impressed by the AVC fans that are working as intake. They were only $17 for the pair, but 6000RPMs move a heck of a lot of air, and what noise is made is I'm sure because of the filters and not motor noise.

While typing this out I've just had another idea of how cooling might be improved overall, but I need to open up the casing to check a few things. Will do a test tonight and see what happens.
 

Flumper Dinkle

Trash Compacter
Feb 26, 2018
48
70
Thanks for your reply.

Ah no. I did not stress the GPU with Prime95. I was more concerned about CPU temps. But what I did do was play TF2 a few times and also run Passmark to check stability. It took a 3rd or 4th hour of play to bring out the instability. Funny how these things work.

Yes the Noctua 4010s are very quiet but they do cost a bit. The 4010 cheap fan I purchased was actually just to test effectiveness of exhaust before investing proper - and they do have some impact, but not very much. I've also been quite impressed by the AVC fans that are working as intake. They were only $17 for the pair, but 6000RPMs move a heck of a lot of air, and what noise is made is I'm sure because of the filters and not motor noise.

While typing this out I've just had another idea of how cooling might be improved overall, but I need to open up the casing to check a few things. Will do a test tonight and see what happens.
If you were wanting to do simultaneous stressing then for reference I was running Prime95 and Furmark simultaneously. Furmark seems to have issues displaying iGPU temps and load percentages so ignore its numbers but it definitely puts 100% load on your GPU.

Yes the noctuas are super expensive and it sounds like you have suitable fans in place anyway. At least if noise does become a concern you have options (removing filters or changing fans).
 

Shahmatt

SFF Lingo Aficionado
Original poster
Sep 6, 2017
101
53
If you were wanting to do simultaneous stressing then for reference I was running Prime95 and Furmark simultaneously. Furmark seems to have issues displaying iGPU temps and load percentages so ignore its numbers but it definitely puts 100% load on your GPU.

Yes the noctuas are super expensive and it sounds like you have suitable fans in place anyway. At least if noise does become a concern you have options (removing filters or changing fans).

So what I wanted to try was check the effectiveness of a push-pull config. After playing around a bit with temps I ended up removing the filters and switching both fans to exhaust. I stuck the filters on to the other end and did some more testing.

Because the filters were not in the way noise was much diminished. Also now, after having got more familiar with o/cing and voltages, it seemed more apparent to me that exhaust was more effective.

So I stuck with this configuration and restored CPU to 3.5GHz, and temps were now well controlled. After a few more hours I've now pushed CPU to 3725 (LLC mode 3) with VCore topping out at 1.32. Also the GPU I pushed to 1525 @ 1.2125V.

Prime95 yields top temps of 93'C. Stressing GPU on AIDA64 yields 73'C. Noise is very low with these fans and for about an hour's gameplay it was well controlled.

I feel I've wasted my time a bit stressing on dust filtration, I'll be trying to cover up the tiny holes as much as I can with filters. But in the mean time I guess this is a good result.
 

Flumper Dinkle

Trash Compacter
Feb 26, 2018
48
70
So what I wanted to try was check the effectiveness of a push-pull config. After playing around a bit with temps I ended up removing the filters and switching both fans to exhaust. I stuck the filters on to the other end and did some more testing.

Because the filters were not in the way noise was much diminished. Also now, after having got more familiar with o/cing and voltages, it seemed more apparent to me that exhaust was more effective.

So I stuck with this configuration and restored CPU to 3.5GHz, and temps were now well controlled. After a few more hours I've now pushed CPU to 3725 (LLC mode 3) with VCore topping out at 1.32. Also the GPU I pushed to 1525 @ 1.2125V.

Prime95 yields top temps of 93'C. Stressing GPU on AIDA64 yields 73'C. Noise is very low with these fans and for about an hour's gameplay it was well controlled.

I feel I've wasted my time a bit stressing on dust filtration, I'll be trying to cover up the tiny holes as much as I can with filters. But in the mean time I guess this is a good result.
Can't be time wasted if you've learned something from it! Now remember with all those fans on exhaust any holes in the case will essentially be acting as intake and that's where dust may get in if you're worried about controlling dust.
Personally I don't worry about dust and give my PCs a quick blast with an air can every couple months or so. This seems to keep them clean enough and if I'm really worried I go in about it with a cotton swab and alcohol.

Good to hear your temps are under control and with a good OC too! I'm still amazed by what the 2200g can do especially considering the cost.
 
  • Like
Reactions: Shahmatt

Shahmatt

SFF Lingo Aficionado
Original poster
Sep 6, 2017
101
53
Can't be time wasted if you've learned something from it! Now remember with all those fans on exhaust any holes in the case will essentially be acting as intake and that's where dust may get in if you're worried about controlling dust.
Personally I don't worry about dust and give my PCs a quick blast with an air can every couple months or so. This seems to keep them clean enough and if I'm really worried I go in about it with a cotton swab and alcohol.

Good to hear your temps are under control and with a good OC too! I'm still amazed by what the 2200g can do especially considering the cost.

Yes it's great value. For a while I was tempted with buying the 2400G. But in one video review I saw it was effectively demonstrated that, after overclocking both chips, the 2200G catches up to the 2400G and matches its performance.

That reviewer showed that both the Vega 8 and 11 were bottlenecked (by cache size or something) and the additional Vega CUs could not produce better frame rates. The reviewer wondered if even a Vega 6 could have produced the same overall performance. Something worth pondering over I feel.
 
  • Like
Reactions: Flumper Dinkle