What's new
  • Please do not post any links until you have 3 posts as they will automatically be rejected to prevent SPAM. Many words are also blocked due to being used in SPAM Messages. Thanks!

Completed I need a 2020 treat build

sswilson

Moderator
Staff member
Joined
Dec 9, 2006
Messages
21,367
Location
Moncton NB

On the off chance that anybody is paying attention to the small details.... the wiring config of that lighting controller is NFG.... RGB connectors must be connected in order starting at #1 in the lower RH corner (which is where they were until I moved them for better cable management... :) ).
 

Lysrin

Well-known member
Joined
Mar 10, 2014
Messages
4,683
Location
Nova Scotia
Looks great @sswilson ! I hadn't seen that case before. Really like the smaller full ATX cases. Great choice and great looking build!
 

sswilson

Moderator
Staff member
Joined
Dec 9, 2006
Messages
21,367
Location
Moncton NB
Looks great @sswilson ! I hadn't seen that case before. Really like the smaller full ATX cases. Great choice and great looking build!

Thanks. I wasn't planning on doing any kind of WC on this build and the Vid card was relatively short so it was a perfect chance to look at something small(ish) in a mid tower case.

Still working out how I want to approach temps/performance on this. I suspect the games/resolutions I want to play on it won't be excessively taxing so I'm second guessing the OC I've done and saved both a 4300 all-core OC and a stock (undervolted by .15V) profile to pair with an OC'd card, and an undervolted one.

I'm leaning towards the stock/undervolted profile as it's giving me substantial temp and noise benefits with minimal performance hits in Timespy, and the only real downside being 15(ish)% performance hit in heavy multithreaded workloads like cinebench.


I guess it comes down to "how good is good enough"?
 

Lysrin

Well-known member
Joined
Mar 10, 2014
Messages
4,683
Location
Nova Scotia
Thanks. I wasn't planning on doing any kind of WC on this build and the Vid card was relatively short so it was a perfect chance to look at something small(ish) in a mid tower case.

Still working out how I want to approach temps/performance on this. I suspect the games/resolutions I want to play on it won't be excessively taxing so I'm second guessing the OC I've done and saved both a 4300 all-core OC and a stock (undervolted by .15V) profile to pair with an OC'd card, and an undervolted one.

I'm leaning towards the stock/undervolted profile as it's giving me substantial temp and noise benefits with minimal performance hits in Timespy, and the only real downside being 15(ish)% performance hit in heavy multithreaded workloads like cinebench.


I guess it comes down to "how good is good enough"?
I came to a similar decision with overclocking my 8700K. Aside from the entertainment and learning from the act of overclocking itself, I didn't see much real-world use difference for games I played so why add the extra heat and noise as you said?

I ended up just doing XMP and whatever settings went with that and let it be.
 

sswilson

Moderator
Staff member
Joined
Dec 9, 2006
Messages
21,367
Location
Moncton NB
Heh... well my 8700K is under water so it's keeping the OC. :)

Ryzen is just weird WRT performance. There's apparently a lot of stuff going on behind the scenes related to power and it isn't something you can easily see by looking at the numbers...

Case in point... I had been running "stock" with a -.15V offset and was seeing pretty much 4200 boost clocks across the board during a cinebench run but got a final score of somewhere around 10850. I realised that I'd left PBO (AMD's built in boost clock OC) at auto so I'm currently doing another run with PBO disabled and the offset voltage removed (so default vcore).

The cinebench run at those settings was confusing.... it was running @ 4000 across all cores (vice 4200 on the previous run), but had much higher temps, and came in @ 12128 which was a significant improvement over the previous run.

There's obviously some kind of performance tuning going on WRT the Vcore offset + PBO that results in higher sustained boost clocks but somehow lower performance related to those clocks.

Definitely something to play with though.... kinda like the good old days of tweaking secondary memory timings one click at a time and then running a quick superpi to check the results it had. :)
 

sswilson

Moderator
Staff member
Joined
Dec 9, 2006
Messages
21,367
Location
Moncton NB
If you are staying in stock cooler, have you tried much with pbo off? I know it will be a minor performance drop, but I think they generally suck back quite a bit more power with PBO.

Also, I can’t remember if it was PBO on or off discussions, but I seem to remember that peak performance was generally around an offset of -0.075 to -0.08...obviously this will be chip dependant, but it may be worth a shot.

One would assume that this would be chip dependent, but your numbers were eerily close to what I ultimately ended up with. :)

PBO = Off / CPU Freq = Auto / (4024 on cinebench load) / Offset = -.0875 / GPU undervolted via the AMD control software / cpu temps don't break 70C, and aren't maxing out the stock cooler.


3 - 4% difference between stock with undervolted cpu + gpu over a fixed clock 4300 Prime stable OC + OC'd gpu. I'd call that a fair trade-off for much lower temps and noise levels.

Screenshot 2020-12-14 162718.png

The cinebench numbers are a bit more of a penalty for not OC'ing but they're still sub 10% from a prime stable OC on a heavily multi-threaded load. 13377 was prime stable 4300, 12977 was the motherboard's one button OC, 12170 was my attempt to raise the offset to .1, and 12370 was the -.0875 offset I finally settled on.
 

Bond007

Well-known member
Joined
Jun 24, 2009
Messages
6,271
Location
Nova Scotia
Nice. I am sure you will keep playing overtime, but it sounds like a solid middle ground to me.
 

sswilson

Moderator
Staff member
Joined
Dec 9, 2006
Messages
21,367
Location
Moncton NB
Interestingly, during the GFX timespy runs the cpu clock is 4300+ so I'm getting decent boost clocks during a typical gaming workload. (The CPU test is somewhere around 4150 - 4225 ).
 

sswilson

Moderator
Staff member
Joined
Dec 9, 2006
Messages
21,367
Location
Moncton NB
System temp testing.jpg

Case / System temperature testing under way. So far so good with (under load) the SP fans running around 1K RPM up front, and the lian-li PWM running around 1200. Been running about 1/2 hr so far and system temp readings are max 50C w/ a raise of 5-8C. (I'll probably run it 2 -3 hrs to see what it ends up at).

SG should be happy to note that I've synced the keyboard and front fans to a fixed colour.... :)

One other note on the case.... it's been mentioned in many of the reviews, but I didn't glom onto the significance of it being mentioned almost globally.... There is zero tint on this glass side, and it's almost freakish with the brain needing to check for the screws to convince itself that the side is on. :)
 

Latest posts

Twitter

Top