Building a gaming PC

+
I'm likely getting a Meshify C for the case replacement (full size version).

If going ATX then strongly recommend the Fractal R6. Comes normal with no glass side. (same, i dont get the purpose of glass panels as i REALLY dont get the idea of putting your PC on the desk. What a waste of real estate. Put it under desk).
The case is one of the quietest ones on market and probably one of the most configurable with everything - radiators or air.
Also Fractal has superb build quality and great support - like noctua. No threaded screws or having to push in backplane to fit cards. Also theyll answer your emails in 24 hrs or less.

heres my build - https://pcpartpicker.com/b/TKkdnQ
Lo & behold it was the featured build of the week with not a single LED / RGB nonsense. (I dont even plug the power LED in). Its been upgraded soemewhat since 2018 - 2080ti, replaced the 3tb black HDD with a 12tb Ultrastar and upped the 1tb 970 with a 2tb model. (I had bought a 4tb qvo to replace the 3tb HDD but sold it before installing.)

Note that evey hole / filter is taped up (even pci cover grilles with clear tape). Air in via front filter, air out via PSU, GPU vent and rear fan. Front to back.
A ton of static pressure and totally silent.

If i could change one thing it would be a screwed in rear fan grille instead of having to use the dremel. But that option is very rare (only seen it on an old Lian Li case i had). Also its being very picky.
And front 4x usb3 instead of 2x usb 2 & 2x usb3. Its a simple mod though. (They now also come with usb C)
 
Last edited:
If going ATX then strongly recommend the Fractal R6.

It's no doubt a good case. It has a few features I dislike or don't need though. I'm not keen on a front hinged door for a variety of reasons. Noise isn't a huge concern since I use headphones. I re-tuned my 8700k OC the other day and the 2k RPM Noctua fans aren't absurdly loud at full RPM either. Noise being of minor concern is the main reason I was looking for a performance focused case.

I'm still not certain on whether I want a small ATX vs a larger option either. The initial preference was small because larger tends to mean more stuff I don't need. High amounts of dedicated space for storage, more IO, additional fan mount locations, sound proofing, obsolete things like 5.25 drive bays, etc. would just end up as excess and unnecessary.

The above is why I had been looking at the Meshify C. It's basically a smaller version of the R6 with a mesh front and fewer bells and whistles.
 
Although it's a midi case, perhaps this one is too big for your taste as from a quick search and comparison it's larger -- but I've really liked what I have read about Phanteks P600. When I finally buy a new rig, I'll probably pick this case (I have an ancient one and want to finally replace it). Excellent cable management, good temperatures, and a front that can be adapted for air flow or cooling. But even with the front fully on, the air flow is actually pretty darn good according to tests I read some months back.

This may be a late reply but.... That case is kind of a big boy. Bigger then I was originally looking to get. Regardless, I kind of said screw it and one is on the way. It's not exactly a perfect match for ITX but looked to be accommodating for anything I'd conceivably pick up in the future.

Unfortunately, the case pickup/search inspired me to launch a larger assault on the money tree. I also have a board, CPU and alternative CPU cooler on the way. It's probably not the best timing with new stuff on the horizon but then I considered the fact there is always new stuff on the horizon. One can never have too many toys :).
 

Guest 2091327

Guest
This may be a late reply but.... That case is kind of a big boy. Bigger then I was originally looking to get. Regardless, I kind of said screw it and one is on the way. It's not exactly a perfect match for ITX but looked to be accommodating for anything I'd conceivably pick up in the future.

Unfortunately, the case pickup/search inspired me to launch a larger assault on the money tree. I also have a board, CPU and alternative CPU cooler on the way. It's probably not the best timing with new stuff on the horizon but then I considered the fact there is always new stuff on the horizon. One can never have too many toys :).

Nice! Congrats with a new rig. It's always a fun time to unpack it all and start to carefully assemble it :)

I'm still on the fence, and wonder if I should wait until new hardware is out or not. Sadly there haven't been any good offers, so for now the decision has been made for me to keep sitting on the fence.

Please report back about your experience with the case, though, as I'm probably buying that one as well. It sounded really good from all the reviews I read and saw. It's actually smaller than the one I currently have -- but then that one has its own eco system :D
 
Please report back about your experience with the case, though, as I'm probably buying that one as well. It sounded really good from all the reviews I read and saw. It's actually smaller than the one I currently have -- but then that one has its own eco system :D

Will do. There are bigger cases but it's not exactly a lightweight either.

In the end I said screw it in my search for the perfect unicorn of a mid-sized, smallish form factor case able to support ATX. Instead I opted to go up in size for compatibility and simplicity. Plus, I had been considering taking the plunge into the custom loop world. It's unlikely to happen in the short-term given the long list of considerations but it tends to require larger sized cases. At least, it's easier to ensure compatibility and keep choices with this direction flexible. It's an open option long-term.

The p600s stood out because cases in the same weight class appear to come with a higher price tag and/or fewer features. At a certain point it's reduced down to a square/rectangular box you put components into. In this regard little things like hinged doors instead of thumb screws, an on demand, modular choice between airflow focus vs noise reduction, etc. are appealing. I couldn't justify paying $180-200 for a big metal box without this stuff when there are big metal boxes for ~$140 with it.

Yes, reviews seemed to paint it in a positive light. I couldn't find anything negative on the case. It's pretty rare for this to be true, regardless of quality.

It's getting a 3900x in it. So fun times there. Evidently it doesn't play by the same rules as most typical consumer chips previous to it. This appears to hold true for Ryzen 2 3000 series chips in general from the info out there (OCing/tuning them isn't roll your face across the keyboard simple as it is with Intel chips). Playing/messing around with the new platform is part of the fun though. It's part of the justification for the choice.
 
Have fun with 3900X :) It surely doesn't work by the same rules as Intel chips.

The plan may just end up be running a stockish config with memory at XMP rated speeds with the IF set appropriately.

I believe straight up manual OC's are off the table for 3900x because everything I've seen says they come with one better and one worse chiplet. It ends up resulting in both operating within the confines of the worse chiplet. Per CCD ocing may be a way around this problem. So it's likely to be explored.

The other option appears to be the PBO + let the chip do it's thing. PBO sounds an awful lot like the multi-core enhancement type concepts on Intel boards. I believe it was reported to not be working initially. If true, hopefully they have ironed it out a bit more. Although, I believe the chip can manage this stuff itself independently of the board. So it's possible to tell the board to give moar powah and have the chip be like, nuh uh.

The more important aspect is how the chips appear to be a tad smarter under the hood. More specifically, the way they throttle frequencies based on thermals. In other words, configuring the chip to run faster may be counter-productive unless you can keep temperatures down in the process.

Toss in the other arbitrary values you can play around with/tweak and fun times. Fun times potentially translating to very little or no performance increase because the chip is probably smarter then the puny little carbon unit configuring it.

Compare this to an Intel chip and... Set frequency -> tweak LLC (ideally a middle range value, depends on board though) -> raise power limits -> find stable voltage for frequency -> enable XMP -> ensure stability of XMP -> done if temps are acceptable. Substitute a lot more tuning if you're not lazy and do memory manually and some minor tuning if you want to go adaptive and/or offset mode or something. <- Boring :).
 
I love GIANT desktop cabinets!
Extra empty space = more room for air circulation = easier & better cooling.
 
The plan may just end up be running a stockish config with memory at XMP rated speeds with the IF set appropriately.

I believe straight up manual OC's are off the table for 3900x because everything I've seen says they come with one better and one worse chiplet. It ends up resulting in both operating within the confines of the worse chiplet. Per CCD ocing may be a way around this problem. So it's likely to be explored.

The whole point of CPU overclocking is getting obsolete, the smaller nodes get, especially due to how frequency boost works. RAM overclocking though is useful so yeah, that's exactly what I'm doing. I don't touch CPU overclocking, just making sure cooling is good, but I set RAM XMP, since I got 3600 MHz / CL16 RAM for it.
 
People are, of course, free to do what they want, but I'll make and argument again for no overclocking whatsoever. If a system's hardware is both compatible and configured properly, the gains from overclocking will be fairly minimal. If people are noticing crazy levels of improvement, that's more indicative of a configuration issue and/or OC values being extremely high.

In the end, OC will increase performance, but can also muck heavily with system timing, needlessly cook hardware (shortening its lifespan), and generally provide gains that may be next-to-imperceptible if compared to that same system configured more efficiently at default clock speeds and voltage.

I know people love to see the big numbers in terms of FPS, but also keep the following in mind:

1.) Your FPS is always limited to your monitor's effective refresh rate. If your monitor is 120 Hz, then you will never, ever see more than 120 FPS. The video card may render 263 FPS, but it can only ever draw 120. The rest of that "performance" is wasted: dropped frames. So, in the example above, that PC is baking its GPU, and displaying less than 50% of its performance output. (Yes, having extra frames rendered can noticeably improve response times and such -- especially if running with Vsync disabled -- but I'd recommend always using a frame cap to ensure you're rendering only ~24 frames more than your refresh rate. In my experience, anything greater than that gives severely diminished returns.)

2.) Physics engines tend to dislike overclocking. Lots of jitters, FPS hitching, or crashing issues can often be tied to physics engines getting all wonky. Many (most?) physics engines will be based on data associated with specific frames. They try to "keep a rhythm" with game processes. One of the big downsides of OC is that it can process things too quickly -- delivering data before the engine is ready for it, then the 3rd-party physics processing (like PhysX) gets all confused and blows up. Best way to test this is to try it with a Bethesda game. Especially Skyrim's physics are prone to all sorts of craziness if the system timing is out of sync with the engine. Great way to identify a good balance is to get Skyrim working well with the OC.

3.) Placebo effect. I have several friends that swear by overclocking, and shake their heads at me for not "taking advantage of my system". :cool: But, these are also the same people that want to know how in the world I get games like TW3 to run sooo smoothly...or how I manage to play Elite: Dangerous with almost no hitching or loading jitters ever. I then explain that my TW3 is running at 48 FPS, capped. They stare blankly at me. I show them that I have Elite locked to 60 at 1080p, and my GPU is normally around 40% in-game. They frown. End result is: when people just sit down and play the game, not looking at the configuration first, the way feels immediately stands out. If they look at the numbers...all of a sudden it starts to feel like there's something "wrong". The true end result is what the gameplay experience is like, not what the numbers say. Less can very much be more.

4.) System stability, especially voltage. A lot of systems have PSUs that are not really all that well suited to overclocking. I'd recommend people invest in a good quality PSU with at least a few hundred watts over what their hardware requires. Both the power overhead and the consistency of the voltage will tend to have a big effect on how the system actually performs. I remember, when I did overclock stuff heavily in the 90's and early 2000's, a bunch of instability problems I was having boiled down to weird voltage fluctuations at inopportune moments.
 
The whole point of CPU overclocking is getting obsolete, the smaller nodes get, especially due to how frequency boost works. RAM overclocking though is useful so yeah, that's exactly what I'm doing. I don't touch CPU overclocking, just making sure cooling is good, but I set RAM XMP, since I got 3600 MHz / CL16 RAM for it.

There is your reason for why in the end it's likely running a minor adjusted stock config with ram at rated speeds. This has less to do with OCing and more with the way Ryzen 2 works though. Apparently, AMD has set the hardware up where it's essentially maximized out of the box. Personally, I see it as a good thing. You can set it and go. Within reason of course.

One of the things I've noticed with Intel chips over the years is this is not the case. There is a lot of headroom available. In fact, the system with an 8700k I have is undervolted. At stock the chip requests 1.3v for the default 4.7ghz boost frequency. It's setup to run with 1.225v load voltage at a max boost of 4.9ghz. In a sense this is a good thing because it can be tinkered around with. On the other hand, it's clearly nowhere near optimal out of the box.

Another good example would be XMP. This Asus board jacks up the IO and SA voltages under XMP when left at auto. It's a common problem with Asus boards. The Asus XMP implementation shoots for maximum compatibility. This means more heat. The system/memory are stable just fine with considerably lower values for these two voltages. Quite frankly, it's target for compatibility at auto is excessive.

I mention the XMP tidbit because in some circumstances the components can go off the rails when allowed to make the decision, so to speak. Most people assume whatever the components do out of the box is "safe". They assume features like XMP are "stock". It's not always true. It's more pronounced with new platforms, varying board manufacturers, busted bioses, and the proverbial growing pains.

People are, of course, free to do what they want, but I'll make and argument again for no overclocking whatsoever. If a system's hardware is both compatible and configured properly, the gains from overclocking will be fairly minimal. If people are noticing crazy levels of improvement, that's more indicative of a configuration issue and/or OC values being extremely high.

In the end, OC will increase performance, but can also muck heavily with system timing, needlessly cook hardware (shortening its lifespan), and generally provide gains that may be next-to-imperceptible if compared to that same system configured more efficiently at default clock speeds and voltage.

See above. Yes, a great number of people push their components to the brink, jack values up and slap "overclocked" on it. They find a guide on the internet and copy settings because everything on the internet is true. Those people are... doing it wrong. The point of manually tuning isn't to get bigger numbers. It's to optimize the hardware. Based upon the specific hardware and component capabilities.

I know people love to see the big numbers in terms of FPS, but also keep the following in mind:

1.) Your FPS is always limited to your monitor's effective refresh rate. If your monitor is 120 Hz, then you will never, ever see more than 120 FPS. The video card may render 263 FPS, but it can only ever draw 120. The rest of that "performance" is wasted: dropped frames. So, in the example above, that PC is baking its GPU, and displaying less than 50% of its performance output. (Yes, having extra frames rendered can noticeably improve response times and such -- especially if running with Vsync disabled -- but I'd recommend always using a frame cap to ensure you're rendering only ~24 frames more than your refresh rate. In my experience, anything greater than that gives severely diminished returns.)

2.) Physics engines tend to dislike overclocking. Lots of jitters, FPS hitching, or crashing issues can often be tied to physics engines getting all wonky. Many (most?) physics engines will be based on data associated with specific frames. They try to "keep a rhythm" with game processes. One of the big downsides of OC is that it can process things too quickly -- delivering data before the engine is ready for it, then the 3rd-party physics processing (like PhysX) gets all confused and blows up. Best way to test this is to try it with a Bethesda game. Especially Skyrim's physics are prone to all sorts of craziness if the system timing is out of sync with the engine. Great way to identify a good balance is to get Skyrim working well with the OC.

The above is why I never hopped on the adaptive sync bandwagon. Great, you can have your game flopping around at 150-120fps on your 120hz display and let the adaptive sync save the day. Or... you could artificially lock it around 120 FPS and have it sit at ~120 FPS damn near the entire time. Not only will it likely run smoother but the hardware will run more efficiently with the latter.

3.) Placebo effect. I have several friends that swear by overclocking, and shake their heads at me for not "taking advantage of my system". :cool: But, these are also the same people that want to know how in the world I get games like TW3 to run sooo smoothly...or how I manage to play Elite: Dangerous with almost no hitching or loading jitters ever. I then explain that my TW3 is running at 48 FPS, capped. They stare blankly at me. I show them that I have Elite locked to 60 at 1080p, and my GPU is normally around 40% in-game. They frown. End result is: when people just sit down and play the game, not looking at the configuration first, the way feels immediately stands out. If they look at the numbers...all of a sudden it starts to feel like there's something "wrong". The true end result is what the gameplay experience is like, not what the numbers say. Less can very much be more.

In fairness, if it feels like something is wrong, crashes are occurring, etc. then odds are something is wrong. Temps being out of whack and hardware behaving irrationally is the definition of unstable. It's what you get when you play with numbers without knowing what they do and/or don't confirm the numbers you have selected work as intended.

4.) System stability, especially voltage. A lot of systems have PSUs that are not really all that well suited to overclocking. I'd recommend people invest in a good quality PSU with at least a few hundred watts over what their hardware requires. Both the power overhead and the consistency of the voltage will tend to have a big effect on how the system actually performs. I remember, when I did overclock stuff heavily in the 90's and early 2000's, a bunch of instability problems I was having boiled down to weird voltage fluctuations at inopportune moments.

Also in fairness, if your system has a sub-par PSU (there are a lot of them out there) then it's a clear sign playing with numbers is going to be limited, assuming it's even on the table. And again, part of the entire process is supposed to be ensuring what has been set functions properly, without a hitch, and without instability.
 
Even RAM OC'ing is largely pointless after 3000MHz or so. I pushed my 3200 kit to 4133 and it did improve performance in CPU demanding titles to some degree, but it's not something you would detect if you weren't looking into the performance parameters.

It is somewhat useful to learn to optimize the timings and frequency of your memory kit manually tho. XMP profiles *can* be unstable and even if they aren't, they don't exactly come even close to maxing the potential performance of your memory.

But it is a time consuming process with a lot of trial and error so def not for everyone.
 
Last edited:
Even RAM OC'ing is largely pointless after 3000MHz or so. I pushed my 3200 kit to 4133 and it did improve performance in CPU demanding titles to some degree, but it's not something you would detect if you weren't looking into the performance parameters.
RAM OC matters when you have 120hz+ monitor to push fps count or in unoptimized games like PUBG, provided that your CPU is powerful enough to process all this data and doesn't deal with freezing. Freezing is mostly happening due to lack of cpu power. 4133mhz RAM won't make i5 9400f (who buys this shit???) suddenly a monster gaming processor. Just cuts frametime spikes from 6-7 to 5-6.
 
But it is a time consuming process with a lot of trial and error so def not for everyone.

I was planning to go with the Ryzen memory calculator. Memory OCing is a pain in the ass.... As silly as it sounds I'd trust a free independent tool over XMP related features.

Not with Zen 2. It differs from Intel in this regard. Optimal RAM frequency for it is around 3600 MHz.

It's the number I'm planning to go with. I preemptively picked up some CL16 3600mhz Ripjaws for it. It's not the best memory but it's not the most expensive either. They're not Neo's though, for whatever that is worth....

I was under the impression the critical aspect was ensuring RAM and the infinity fabric are in sync. The 3600-3800mhz or so range is the ideal setup I've seen recommended. 3600 seems to be the default value floating around because most memory controllers with the CPU's are likely able to handle it. Any higher and you get into the proverbial silicon lottery. For both the RAM and CPU.
 
Not with Zen 2. It differs from Intel in this regard. Optimal RAM frequency for it is around 3600 MHz.
Yeah, memory controller is weaker on Ryzen. But the difference between 3000 and 3600 ram is still quite small. The difference is only really noticeable when looking at performance parameters.

And afaik, going past 3800 on ryzen actually starts to have a negative effect on performance, because of infinity fabric.
 
Yeah, memory controller is weaker on Ryzen. But the difference between 3000 and 3600 ram is still quite small. The difference is only really noticeable when looking at performance parameters.

And afaik, going past 3800 on ryzen actually starts to have a negative effect on performance, because of infinity fabric.

I'm not sure it's necessarily weaker. It's just... different. In hindsight the earlier comment about the IMC probably isn't the case. Based on further reading and playing around with it the 3600 recommendation is probably more based on the infinity fabric. In other words, you could and can shoot higher for the memory frequency but getting it synced up with the IF is ideal. The IF is the bottleneck, however. Getting it much beyond 1800mhz is challenging.

I suppose it makes sense... If you take the little chip and spread it out across multiple chiplets, with an interconnect between them, the interconnect limitations are a big deal. Since RAM frequency is tied to this interconnect, so to speak, it's also limited in how far it can be pushed. You could probably increase RAM frequency much higher if you didn't care about keeping it in sync with the IF.

On an unrelated note.... I should have spent a bit more on the board and went with the X570 Taichi. Ended up with the Auros Elite because all the info seemed to point to nothing being worth the jump in price until the $300 range. The performance is fine but good god the bios layout is a non-intutive clusterfuck :).
 
X570 Taichi is quite good. But I have one annoying bug with it, that Asrock so far refuse to fix. Some LEDs (Dr. Debug) stay on after resume form suspend.
 
X570 Taichi is quite good. But I have one annoying bug with it, that Asrock so far refuse to fix. Some LEDs (Dr. Debug) stay on after resume form suspend.

Yeah, I saw a commentary indicating it's one of the better boards in it's price range. A few claimed it's arguably the best because it's overbuilt for the price range.

I think for the lower end it was between the Auros Elite and one of the Asus boards. With everything in between being quite similar from a performance standpoint. It's implied you have boards in the $250 range to fill out a spot at the price range. Instead of offering anything highly useful for general use they take an existing board, largely copy it, change the coloring scheme and add a few extras. If those features are desired I suppose it's worth it. They were not for me.

It's not a huge deal for a bios to be non-intuitive and formatted strangely. Eventually you figure out the layout. As long as it functions correctly anyway. Regardless, Gigabyte should probably go back to the drawing board with their bios layouts.

In terms of the 3900x.... I'm pleased with it so far. It's probably a bit of a side grade for gaming compared to an 8700k but it's superior for damn near everything else.

Tuning it appears to consist of doing pretty much nothing. It's effectively running stock at the moment with RAM setup to run it's rated 3600mhz profile. It would appear there isn't much else to be done. I haven't tried Ryzen Master and have been exclusively fiddling with tuning via bios though.

PBO is enabled but I'm not sure it's even relevant. I recall reading a write-up claiming the concept is fine but the chip has it's own built in limitations. The long and short of it is by the time the limits PBO raises are met chances are high, at least for the higher Ryzen 2 SKU's, the chip limitations are being met. Thus, it doesn't do much.

Please report back about your experience with the case, though, as I'm probably buying that one as well. It sounded really good from all the reviews I read and saw. It's actually smaller than the one I currently have -- but then that one has its own eco system :D

Just to report back... I'm extremely pleased with the case. I'd go so far as to say it's probably the best I've ever slapped components into. The performance is equally pleasing. It does come with quite a few accessories but these are a bonus due to the modularity they allow. It does a lot of little things well.

In terms of specifics.... Cut-outs for cable routing are some of the best I've seen. The thing basically cable manages itself. One area I liked here was how the built-in tie down points are "open". So instead of struggling to fish straps through tiny little holes you can halfway tie them down and loop them around the tie down points. I've had experiences with other cases where the tie down points for straps were closed off loops. It can get annoying :). There are even individual fan route holes perfectly positioned beside where you install front fans.

Stuff like including plenty of various screws in a dedicated plastic box with seperators isn't major. It is something though. It comes with a lot of spare screws to boot. This came in the accessory box with spinny boy HD cages, a anti-droop GPU bracket and a vertical GPU mount (didn't use any of these but they're there).

The magnetic hinged doors concept is pretty helpful too. They are removable, although the back door is a pain to take off without removing the top panel (held in by 2 thumb screws). You can get it off but it requires opening it up first, as the back top part of the door is underneath the lip on the top panel. Any concerns about the doors closing via magnets were dispelled quickly. The magnets on the doors are pretty damn strong. There is no chance they will inadvertantly open. My only complaint here is getting the doors lined up with both hinge pins to put them back on the case. Although, this could just be a me problem.

All in all it was a good recommendation. So thanks for mentioning it.
 
Top Bottom