Small form factor gaming - build log and support forum for new builders

[AusWolf]

I'm starting this thread as a build log, but if anyone wants to join with their systems, or just to seek advice, feel free.

I've been a PC gamer since 1998, and have been building my own computers since 2004. In the last 7 or 8 years, I've been specialising in small form factor builds. I find packing more punch into a small case more satisfying than slamming high-end components into a large case with plenty of airflow and calling it a day. I've experimented with various generations of Core i3s, i7s, Ryzens, and now I'm back to a Rocket Lake Core i7-based system. My experiences have been vastly different from that of reviewers' who often test new CPUs with top-end cooling and unrestricted power limits, so I thought I'd share some of it. Where to begin...

Life with my Ryzen 9 5950X and Radeon RX 5700 XT was nice until the chip shortage hit that compelled me to sell them to make some money. I have to note that I was running the system in an Aerocool Aero One Mini case with a Corsair H100i 240 mm AIO liquid cooler. Under stock settings with my Asus TUF Gaming B550M WiFi motherboard, the CPU consumed 130 Watts, and ran between 70 and 80 °C under full load. All-core turbo was around 3.6-3.8 GHz. If I enabled the "Asus Optimizer" in BIOS (a feature that disables power limits and sets single-core turbo speeds and voltages as all-core - on AMD platforms at least), the Ryzen 9 needed 180 Watts and came close to thermal throttling with clock speeds around 4.4-4.6 GHz. Naturally, I didn't use this option, as the CPU was overkill for gaming even with stock settings.

Then I sold most of the hardware, retired the case and cooler into the wardrobe, and moved the system into a lot smaller (micro-ATX slim) Aerocool CS-101 office case with a Ryzen 3 3100 with its stock cooler and a low profile GeForce GTX 1650. The concept was, and still is a relatively portable, thin and tiny gaming rig with a sleeper design, which worked out just fine with the above mentioned CPU+GPU combination. CPU maxing out at 72 °C, GPU at 75-76 isn't too bad from a case with virtually no airflow. The Ryzen 3 3100 also has the weird attribute of running at 3.8-3.9 GHz at all times independent of workload, while only needing around 50 W maximum, which I think is a great thing.

Then I got the idea of upgrading into my head. I didn't want to do anything overkill (again), so I decided to stick to the 65 Watt TDP category, and bought a Ryzen 5 3600. First I tried it with the stock cooler, but it came very close to thermal throttling under stock settings, so I quickly bought a be quiet! Shadow Rock LP, which is basically the biggest cooler I can fit into my case. This new cooler managed to keep the 3600 under 90 °C, though weirdly enough, it remained cold to the touch. First I thought contact between the CPU and the cooler was wrong, so I reseated the cooler a couple times, but saw no improvement. I use Arctic MX-4 thermal paste, by the way. Luckily, a friend of mine needed an upgrade for his (bigger than mine) PC, and bought the 3600 off of me.

Then I thought, maybe modern Ryzen's increased heat density, offset CPU die and 88 W power target was the reason I couldn't cool the 3600, so I took a bold move and ordered a motherboard with a Core i7-11700. No "K" suffix here; I've never regarded overclocking in much value. Besides, when you're building SFF, lower TDP (and heat output) is always the way to go. I never even took the stock cooler out of the box, as I remember having trouble with it cooling the Core i7-7700 (again, no "K" designation) I had before the 5950X. At first try, the CPU shot up to 95 °C with 180 Watts of power consumption during the 225 W stock PL2 (secondary power limit), but nested in the lower 60s once Tau (PL2 duration) expired. That gave me the idea of playing with the power limits. Here are some results, including the original findings:

• 65 W PL1: 2.8 GHz all-core boost, low-60s temp (silent fan profile), Cinebench R23 around 9000 points.
• 80 W PL1: 3-3.1 GHz all-core, 65-68 °C (silent fan profile), Cinebench R23 around 9500-9600 points.
• 100 W PL1: 3.3-3.4 GHz all-core, 67-68 °C (turbo fan profile), Cinebench R23 scores 10444 points. The Ryzen 5 3600 nearly throttled even at 88 W PPT with the same cooler.

I know these scores aren't stellar from a modern 8-core CPU, but the goal was never to break records, but to make it work as part of a tiny system, and it does! Before I conclude my home-built random Ryzen 5 3600 vs Core i7-11700 cooling battle, I have to note some more things about my experiences with the i7:

• Single-core performance is unaffected by the modified power limits, as the CPU needs around 50 Watts to maintain its 4.8-4.9 GHz turbo in single-threaded workloads.
• The temperatures mentioned above were recorded with the memory controller set to Gear 2. Oddly enough, the Auto BIOS setting resulted in about 10 °C higher package temp (and no performance gain) for some reason.
• As I mentioned before, the be quiet! cooler ran cold, but let the Ryzen 5 nearly overheat under stock settings, but runs fairly warm while keeping the i7 cool, indicating that heat transfer is much better with the Intel CPU.

My conclusion: Building a small form factor gaming PC that doesn't overheat isn't impossible. One needs to be educated about the possibilities, though. AMD's Ryzen 3000 and 5000 processors are excellent products, but anything more powerful than a Ryzen 3 should stay in a normal PC with at least a tower cooler, as heat transfer is measurably worse than on Intel CPUs. On the other side of things, despite Ryzen's power efficiency, I would still recommend Intel for small form factor gaming rigs.

The future: As the little i7 nicely settled into the small office case, it's time to use it for the intended purpose: gaming. Of course performance won't be affected with the GTX 1650, but I'm planning to run some tests for thermals and clock speeds under various PL settings. As for the further future, I might be looking at AMD's Ryzen 5000G APU line to see if the larger central die they use have any better heat dissipation characteristics than their CPUs. Or I might just sell my Ryzen 3 CPU + motherboard combo and call it a day. I might also move the new system into the Aero One Mini case and increase its power limits even further during a graphics card upgrade (when that happens is a big question mark at the moment). We'll see.

As a bonus, here's a picture of the finished build (details in my profile) with a CPU box for size comparison:

 

[tabascosauz]

Good stuff! That Aerocool case is certainly smaller than it looks at 14L. Nice and slim.

I'm still intrigued as to the "cold cooler" phenomenon, I've not seen it on any of my coolers when I was running the 3700X, which behaves very similarly to the 3600 if not a little hotter. During the appropriate stress tests, the cooler would always be warm to the touch after reaching thermal equilibrium. NH-D9L, NH-U9S, NH-C14S, even the Dark Rock Pro 4. Feel like there was something else going on with that 3600 setup there.

As to Ryzen and SFF, (shameless plug lol) my sig thread has my entire journey from Haswell to Matisse, Renoir and Vermeer - I felt the same way at first as Ryzen is definitely a culture shock moving away from Intel, but it just takes time to get used to it. Whether you prefer AMD or Intel, it's inevitable in the SFF space that if you use a cooler that's small for the intended CPU or a case without much airflow, lots of tweaking will be necessary.

With all mobile and desktop Skylake derivatives from Skylake to Comet Lake, Intel usually has a fair bit of undervolting headroom. If you want to get the most out of your chip, might be worth using BIOS settings or Throttlestop to figure out how much Vcore(/Vcache) you can shave off for free performance or lower temps.

 

[Why_Me]

I'm starting this thread as a build log, but if anyone wants to join with their systems, or just to seek advice, feel free.

I've been a PC gamer since 1998, and have been building my own computers since 2004. In the last 7 or 8 years, I've been specialising in small form factor builds. I find packing more punch into a small case more satisfying than slamming high-end components into a large case with plenty of airflow and calling it a day. I've experimented with various generations of Core i3s, i7s, Ryzens, and now I'm back to a Rocket Lake Core i7-based system. My experiences have been vastly different from that of reviewers' who often test new CPUs with top-end cooling and unrestricted power limits, so I thought I'd share some of it. Where to begin...

Life with my Ryzen 9 5950X and Radeon RX 5700 XT was nice until the chip shortage hit that compelled me to sell them to make some money. I have to note that I was running the system in an Aerocool Aero One Mini case with a Corsair H100i 240 mm AIO liquid cooler. Under stock settings with my Asus TUF Gaming B550M WiFi motherboard, the CPU consumed 130 Watts, and ran between 70 and 80 °C under full load. All-core turbo was around 3.6-3.8 GHz. If I enabled the "Asus Optimizer" in BIOS (a feature that disables power limits and sets single-core turbo speeds and voltages as all-core - on AMD platforms at least), the Ryzen 9 needed 180 Watts and came close to thermal throttling with clock speeds around 4.4-4.6 GHz. Naturally, I didn't use this option, as the CPU was overkill for gaming even with stock settings.

Then I sold most of the hardware, retired the case and cooler into the wardrobe, and moved the system into a lot smaller (micro-ATX slim) Aerocool CS-101 office case with a Ryzen 3 3100 with its stock cooler and a low profile GeForce GTX 1650. The concept was, and still is a relatively portable, thin and tiny gaming rig with a sleeper design, which worked out just fine with the above mentioned CPU+GPU combination. CPU maxing out at 72 °C, GPU at 75-76 isn't too bad from a case with virtually no airflow. The Ryzen 3 3100 also has the weird attribute of running at 3.8-3.9 GHz at all times independent of workload, while only needing around 50 W maximum, which I think is a great thing.

Then I got the idea of upgrading into my head. I didn't want to do anything overkill (again), so I decided to stick to the 65 Watt TDP category, and bought a Ryzen 5 3600. First I tried it with the stock cooler, but it came very close to thermal throttling under stock settings, so I quickly bought a be quiet! Shadow Rock LP, which is basically the biggest cooler I can fit into my case. This new cooler managed to keep the 3600 under 90 °C, though weirdly enough, it remained cold to the touch. First I thought contact between the CPU and the cooler was wrong, so I reseated the cooler a couple times, but saw no improvement. I use Arctic MX-4 thermal paste, by the way. Luckily, a friend of mine needed an upgrade for his (bigger than mine) PC, and bought the 3600 off of me.

Then I thought, maybe modern Ryzen's increased heat density, offset CPU die and 88 W power target was the reason I couldn't cool the 3600, so I took a bold move and ordered a motherboard with a Core i7-11700. No "K" suffix here; I've never regarded overclocking in much value. Besides, when you're building SFF, lower TDP (and heat output) is always the way to go. I never even took the stock cooler out of the box, as I remember having trouble with it cooling the Core i7-7700 (again, no "K" designation) I had before the 5950X. At first try, the CPU shot up to 95 °C with 180 Watts of power consumption during the 225 W stock PL2 (secondary power limit), but nested in the lower 60s once Tau (PL2 duration) expired. That gave me the idea of playing with the power limits. Here are some results, including the original findings:

• 65 W PL1: 2.8 GHz all-core boost, low-60s temp (silent fan profile), Cinebench R23 around 9000 points.
• 80 W PL1: 3-3.1 GHz all-core, 65-68 °C (silent fan profile), Cinebench R23 around 9500-9600 points.
• 100 W PL1: 3.3-3.4 GHz all-core, 67-68 °C (turbo fan profile), Cinebench R23 scores 10444 points. The Ryzen 5 3600 nearly throttled even at 88 W PPT with the same cooler.

I know these scores aren't stellar from a modern 8-core CPU, but the goal was never to break records, but to make it work as part of a tiny system, and it does! Before I conclude my home-built random Ryzen 5 3600 vs Core i7-11700 cooling battle, I have to note some more things about my experiences with the i7:

• Single-core performance is unaffected by the modified power limits, as the CPU needs around 50 Watts to maintain its 4.8-4.9 GHz turbo in single-threaded workloads.
• The temperatures mentioned above were recorded with the memory controller set to Gear 2. Oddly enough, the Auto BIOS setting resulted in about 10 °C higher package temp (and no performance gain) for some reason.
• As I mentioned before, the be quiet! cooler ran cold, but let the Ryzen 5 nearly overheat under stock settings, but runs fairly warm while keeping the i7 cool, indicating that heat transfer is much better with the Intel CPU.

My conclusion: Building a small form factor gaming PC that doesn't overheat isn't impossible. One needs to be educated about the possibilities, though. AMD's Ryzen 3000 and 5000 processors are excellent products, but anything more powerful than a Ryzen 3 should stay in a normal PC with at least a tower cooler, as heat transfer is measurably worse than on Intel CPUs. On the other side of things, despite Ryzen's power efficiency, I would still recommend Intel for small form factor gaming rigs.

The future: As the little i7 nicely settled into the small office case, it's time to use it for the intended purpose: gaming. Of course performance won't be affected with the GTX 1650, but I'm planning to run some tests for thermals and clock speeds under various PL settings. As for the further future, I might be looking at AMD's Ryzen 5000G APU line to see if the larger central die they use have any better heat dissipation characteristics than their CPUs. Or I might just sell my Ryzen 3 CPU + motherboard combo and call it a day. I might also move the new system into the Aero One Mini case and increase its power limits even further during a graphics card upgrade (when that happens is a big question mark at the moment). We'll see.

As a bonus, here's a picture of the finished build (details in my profile) with a CPU box for size comparison:
View attachment 202851

Nice. btw what board did you go with?

 

[AusWolf]

Nice. btw what board did you go with?

Thanks. With Ryzen, I use(d) an Asus TUF Gaming B550M-Plus Wifi, which is an excellent board, so I decided to change only one number to B560M-Plus Wifi.

Edit: This has just the same quality, I only have one gripe against it: my Corsair MP600 Core SSD doesn't fit into the PCI-e 4.0 m.2 slot because the PCI-e x16 retention clip is in the way for the heatsink.

Good stuff! That Aerocool case is certainly smaller than it looks at 14L. Nice and slim.

Thanks. It is really small. I can easily pick it up and carry around the house with one hand.

I'm still intrigued as to the "cold cooler" phenomenon, I've not seen it on any of my coolers when I was running the 3700X, which behaves very similarly to the 3600 if not a little hotter. During the appropriate stress tests, the cooler would always be warm to the touch after reaching thermal equilibrium. NH-D9L, NH-U9S, NH-C14S, even the Dark Rock Pro 4. Feel like there was something else going on with that 3600 setup there.

It is definitely weird. I never had this problem with any CPU. The 5950X was hot, but so was the liquid in the H100i after prolonged use (I used it with very relaxed fan/pump curves). The 3100 behaves fine too. I thought selling the 3600 to a friend (I gave it to him for a test first) would show that maybe the processor is faulty, but he's totally fine using it with a noname tower cooler.

As to Ryzen and SFF, (shameless plug lol) my sig thread has my entire journey from Haswell to Matisse, Renoir and Vermeer - I felt the same way at first as Ryzen is definitely a culture shock moving away from Intel, but it just takes time to get used to it. Whether you prefer AMD or Intel, it's inevitable in the SFF space that if you use a cooler that's small for the intended CPU or a case without much airflow, lots of tweaking will be necessary.

I'll definitely check it out. Building PCs is fine, but if you want to have fun doing it, SFF is the way to go.

To be honest, my biggest culture shock with Matisse was having to use chipset drivers like it was 2013 again. Thank heavens that AMD got rid of that crap with Vermeer. The next one is the high reported idle clocks and power consumption. Every Ryzen 3/5000 CPU I've used reports asking for around 20-22 Watts in idle with the Ryzen balanced (Windows balanced for Vermeer) profile. It's more than any Core i processor I've seen (and I've seen quite a few).

With all mobile and desktop Skylake derivatives from Skylake to Comet Lake, Intel usually has a fair bit of undervolting headroom. If you want to get the most out of your chip, might be worth using BIOS settings or Throttlestop to figure out how much Vcore(/Vcache) you can shave off for free performance or lower temps.

That's good to know. I might try that next.

 

[tabascosauz]

With Ryzen, I use(d) an Asus TUF Gaming B550M-Plus Wifi, which is an excellent board, so I decided to change only one number to B560M-Plus Wifi.

Agreed, both excellent boards. It's hard away to get away from that Asus BIOS.

The next one is the high reported idle clocks and power consumption. Every Ryzen 3/5000 CPU I've used reports asking for around 20-22 Watts in idle with the Ryzen balanced (Windows balanced for Vermeer) profile. It's more than any Core i processor I've seen (and I've seen quite a few).

The only idle clocks on Ryzen you can really trust are either Effective Clock in HWInfo, or Ryzen Master. Just how the CPU works, I'm afraid, hardware has gotten to the point where software is just too stupid or unwilling to keep up and AMD doesn't feel like disclosing all of its proprietary reporting data.

It sounds like you've been only using chiplet Ryzens - unfortunately the idle package power is an inherent symptom of the design. 1CCD is a little better than 2CCD but the floor is still around 15W. By contrast, the modern AM4 APUs (Renoir and Cezanne) idle almost exclusively at sub-10W. More like 6W on mine usually, and that's with static VDDCR_GFX from a hefty iGPU overclock (+19.7% core) and Infinity Fabric above 2000MHz.

About the point on undervolting, I'm still not sure exactly it works on Rocket Lake. On a lot of other platforms it's just a negative voltage offset, but iirc RKL advertised precise V-F curve tweaking - kinda like Core Optimizer but less abstract? Not sure if the TUF BIOS has it, I know the real ROG boards might have it and Intel XTU definitely has it.

 

[AusWolf]

The only idle clocks on Ryzen you can really trust are either Effective Clock in HWInfo, or Ryzen Master. Just how the CPU works, I'm afraid, hardware has gotten to the point where software is just too stupid or unwilling to keep up and AMD doesn't feel like disclosing all of its proprietary reporting data.

That is my complaint against Ryzen at this point (especially with Matisse): there's just too much software involved. On Intel, you do everything in the BIOS. There's no chipset driver, there's no Ryzen Master, not any of that overcomplicated bollocks. I always say simple is best.

It sounds like you've been only using chiplet Ryzens - unfortunately the idle package power is an inherent symptom of the design. 1CCD is a little better than 2CCD but the floor is still around 15W. By contrast, the modern AM4 APUs (Renoir and Cezanne) idle almost exclusively at sub-10W. More like 6W on mine usually, and that's with static VDDCR_GFX from a hefty iGPU overclock (+19.7% core) and Infinity Fabric above 2000MHz.

True. My first Ryzen was the 3100, then I upgraded to the 5950X, then I decided that I don't need it, so I downgraded to a 3600 which didn't work very well in the thin case, and I still have the 3100. I would love to have a 5700GE to experiment with, but AMD decided not to release it for DIY for some reason.

About the point on undervolting, I'm still not sure exactly it works on Rocket Lake. On a lot of other platforms it's just a negative voltage offset, but iirc RKL advertised precise V-F curve tweaking - kinda like Core Optimizer but less abstract? Not sure if the TUF BIOS has it, I know the real ROG boards might have it and Intel XTU definitely has it.

I read somewhere that undervolting Rocket Lake is as easy as setting a -100 mV offset in the BIOS. I'm not sure if it's true, and I'm a little scared to try. The 11700 already has pretty low voltages, barely exceeding 1 V under full load. Playing with power targets seems simpler and safer to me.

Edit: I've just run a 10-minute Cinebench R23 loop (my usual stability test) with a -100 mV offset. It works. It didn't affect thermals (steady 71 °C package temp), as the extra power headroom went into more boost. Now with 3.4-3.5 GHz, it scored 10889 points, which is a 445-point improvement over the same 100 W power limit with no voltage offset. Not bad.

Edit 2: All the effort with the increased power limits and negative voltage offset doesn't really show in real life. In Cyberpunk 2077 that uses the CPU between 30-50% depending on the scene, core clocks only change from 4.1 to 4.3 GHz average. With this in mind, I guess I'll just leave PL1 at 65 W for now. There's no need to heat up the case for no reason.

 

[Valantar]

Great build log That's a pretty interesting case - very old school in many ways (an ODD bay seems so quaint in 2021 ), but still impressively small for what it can fit. Definitely has some deficiencies in terms of airflow and cooling compared to more modern designs, but clearly not much of an issue given your hardware Also looks like there would be plenty of potential for modding it to fit longer GPUs, possibly more intake airflow, etc.

I'm still curious about your 3600 thermals - to me, it sounds like something was off there. Maybe a poor solder joint between the die and IHS? Maybe that cooler just has a weird pressure pattern that aligns very poorly with the offset die? Maybe the IHS on your CPU was sub-par in terms of flatness, leading to a contact issue? Given that I could run my 5800X under a 212 Evo (including its kind of janky clip-on AM4 mount) with zero issues and very decent thermals (open bench, but still), I'm surprised your 3600 was that hot. It's great that you're seeing good results with the 11700 though - undervolting and setting a lower PL1 seems like a good way to go. If I were you (and looking for tangible improvements in gaming and similar non-nT workloads), I would probably experiment with increasing PL1 (within reason/the capabilities of your cooling) while undervolting - that should allow for more sustained boosting on low(er) threaded workloads. If you're seeing 30-50% load un CP2077, that's likely a high load on 3-4 cores, which probably means you're power limited at that point. Then again, in games like that you're most likely GPU limited anyhow, and might as well save your case and GPU from the added heat load of pushing the CPU further. Depends on the game though!

 

[The red spirit]

an ODD bay seems so quaint in 2021

My brand new Cooler Master Silencio S400 came with ODD bay and I put DVD drive there. Then again, there's no RGB, no side window, no USB-C... And the main reason why I got it over Define C, is that ODD bay. It's so stupid that cases don't have basic functionality these days. Most pc users wouldn't be able to put UHD Blu-Ray drive or fan controller if they wanted to. If I was rick one day, I would definitely want to have a case, where I could fit a dedicated fan controller with dedicated knobs for each fan (meaning that it couldn't have doors).

 

[Valantar]

My brand new Cooler Master Silencio S400 came with ODD bay and I put DVD drive there. Then again, there's no RGB, no side window, no USB-C... And the main reason why I got it over Define C, is that ODD bay. It's so stupid that cases don't have basic functionality these days. Most pc users wouldn't be able to put UHD Blu-Ray drive or fan controller if they wanted to. If I was rick one day, I would definitely want to have a case, where I could fit a dedicated fan controller with dedicated knobs for each fan (meaning that it couldn't have doors).

It's still a very rare thing to see these days, and more so in an SFF case where space is at a premium. The reason is pretty simple: ODD bays aren't basic functionality any more, but has grown to be a niche feature. The vast majority of DIY PCs built the past few years don't have any kind of optical drive, and many PC enthusiasts today have never even really used one at all. It's just rather obsolete for most use cases. Nothing is widely distributed on discs these days (save for bundled driver discs which are generally unnecessary due to how many drivers are built into Windows). The last games I bought on disc all mandated activation through Steam and downloading from there, making the disk utterly useless, which just served to underscore the lack of need for an ODD in my case. Software these days is all downloaded. And so on. They clearly still have their uses - I have both an USB DVD writer and an USB BD drive lying around in case I have a use for them, but outside of ripping the occasional movie backup neither has seen use for a long, long time. I think I stopped installing ODDs in my desktops around 2014 or so when I got my Define R4, but by then the ODD in my old case had been broken for years and was never missed. And that's the reality for the vast majority of users. So while I understand the frustration of people who use them regularly with the lack of ODD support in current cases, their removal has been a huge benefit for case designs overall, as front 5 ¼" bays place a lot of restrictions on possible layouts. IMO it's reasonable for users with what is now a niche need to either seek out niche cases or use external drives. If anything, there should be more external 5¼" adapters available.

As for fan control, I've long since accepted that this is far better left to the PC to handle. I used to be really enamored with those front bay controllers full of knobs and with old-school LCD displays reminiscent of a late 90s stereo, but software will always be better able to monitor the relevant data and adjust fan speeds to match, which also leaves me to focus on what I'm using the PC for rather than focusing on peripheral operations - which would be a severe distraction in pretty much any usage scenario. I'm really happy with my current Aquacomputer Quadro, which does fantastic fan control for a low pric. The tactile and haptic elements of PC usage are really important to me, but I also don't believe in holding on to manual control of tasks that the uni-tasking human consciousness is poorly suited to just because something about it feels nice. I suppose the ideal would be a smart interface with analog, motorized knobs that worked with a good fan control software/hardware combo (e.g. Aquasuite), but I shudder to think what something like that would cost. I suppose one could make a Monogram Console do this, but... yeah, I'm not paying $400 for fan control XD

 

[The red spirit]

It's still a very rare thing to see these days, and more so in an SFF case where space is at a premium. The reason is pretty simple: ODD bays aren't basic functionality any more, but has grown to be a niche feature. The vast majority of DIY PCs built the past few years don't have any kind of optical drive, and many PC enthusiasts today have never even really used one at all. It's just rather obsolete for most use cases. Nothing is widely distributed on discs these days (save for bundled driver discs which are generally unnecessary due to how many drivers are built into Windows). The last games I bought on disc all mandated activation through Steam and downloading from there, making the disk utterly useless, which just served to underscore the lack of need for an ODD in my case. Software these days is all downloaded. And so on. They clearly still have their uses - I have both an USB DVD writer and an USB BD drive lying around in case I have a use for them, but outside of ripping the occasional movie backup neither has seen use for a long, long time. I think I stopped installing ODDs in my desktops around 2014 or so when I got my Define R4, but by then the ODD in my old case had been broken for years and was never missed. And that's the reality for the vast majority of users. So while I understand the frustration of people who use them regularly with the lack of ODD support in current cases, their removal has been a huge benefit for case designs overall, as front 5 ¼" bays place a lot of restrictions on possible layouts. IMO it's reasonable for users with what is now a niche need to either seek out niche cases or use external drives. If anything, there should be more external 5¼" adapters available.

Well for me DVD drive is just convenient, because there's always some old files left on random CD or DVD, maybe some music, maybe something else. I would say that SFF marked should really appreciate ODD bays as part of their selling point is compact and powerful system for media playback and pretty much any new movies is still legally on Blu-ray or UHD Blu-Ray. Some old movies are still on DVD. If you are into anime legally, you would be very screwed without some disc reader or even a VHS player (sometimes a Laserdisc). There's just so much stuff on optical media, even if it's old, it's far from being useless. Oh and old games from pre-Steam era. Either you pirate them or you hunt down a disc. Sometimes GOG is helpful, but it's only rarely useful. And the newest game that I have on DVD is Simcity, for some reason it was released on DVD and that's the whole game, not just a code to download it. EA managed to fit it all there.

As for fan control, I've long since accepted that this is far better left to the PC to handle. I used to be really enamored with those front bay controllers full of knobs and with old-school LCD displays reminiscent of a late 90s stereo, but software will always be better able to monitor the relevant data and adjust fan speeds to match, which also leaves me to focus on what I'm using the PC for rather than focusing on peripheral operations - which would be a severe distraction in pretty much any usage scenario. I'm really happy with my current Aquacomputer Quadro, which does fantastic fan control for a low price. The tactile and haptic elements of PC usage are really important to me, but I also don't believe in holding on to manual control of tasks that the uni-tasking human consciousness is poorly suited to just because something about it feels nice. I suppose the ideal would be a smart interface with analog, motorized knobs that worked with a good fan control software/hardware combo (e.g. Aquasuite), but I shudder to think what something like that would cost. I suppose one could make a Monogram Console do this, but... yeah, I'm not paying $400 for fan control XD

I meant it for setting once and forgetting, because auto control sucks most of the time. Seriously, if there was a graphics card with manually adjustable fan speed knob I would really want it, because a card under any circumstances shouldn't exceed 1400 rpm and I would love to have something like that, when the only solution is to either mod card's vBIOS or set up that stuff in AMD control panel after every start up. Graphics cards aside, I still haven't seen a single UEFI, which would let me to set fan speed to one specific value and don't auto adjust fan speed.

 

[tabascosauz]

Graphics cards aside, I still haven't seen a single UEFI, which would let me to set fan speed to one specific value and don't auto adjust fan speed.

You can easily just make a custom fan curve that's flat or near-flat, on any vendor's current BIOSes......I run a nearly flat 50-65% fan curve on my 5900X on the Asus board with 3.6sec ramp-up and 2.8sec ramp-down delaya, and I run literally a flat 70%ish fan curve on the two 80mm case fans in my SFF with the Gigabyte board. No audible ramp-up or down at all.

They won't let you have your desired rpm above a certain temp threshold for safety reasons, but you can just move that threshold up to like 80-90C so it doesn't ramp up until then.

BIOS fan control isn't nearly as dumb as you make it out to be, especially on Asus more than the others. You just gotta take some time to explore the tools at your disposal.

 

[The red spirit]

You can easily just make a custom fan curve that's flat or near-flat, on any vendor's current BIOSes......I run a nearly flat 50-65% fan curve on my 5900X on the Asus board with 3.6sec ramp-up and 2.8sec ramp-down delay, and I run literally a flat 70%ish fan curve on the two 80mm case fans in my SFF with the Gigabyte board. No audible ramp-up or down at all.

They won't let you have your desired rpm above a certain temp threshold for safety reasons, but you can just move that threshold up to like 80-90C so it doesn't ramp up until then.

BIOS fan control isn't nearly as dumb as you make it out to be, especially on Asus more than the others. You just gotta take some time to explore the tools at your disposal.

Well, that's a workaround. I can set it to max speed and then curve is disables, so why I can't just set fans to one specific value at all times without touching that curve? Maybe UEFI fan control isn't so stupid, but it sure is frustrating and clunky. If I could pick between hardware and software fan controller, I would always pick a hardware controller.

 

[AusWolf]

We're moving!

Well, I don't mean "we" as myself and my girlfriend. I mean my PC components into another case.

It all started with me randomly browsing eBay for graphics cards. I didn't intend to buy, I was just looking. I remember RTX 2060/2070/2080 cards selling between 800 and 1000 GBP a month back, and I was surprised to see that their price dropped to around 400-450 for the 2060 and 550-600 for the 2070. It's still quite high, and I still didn't intend to buy. Then I found an EVGA 2070 Black in immaculate condition for £570 with the seller accepting offers. I made an offer of £520 just because why not, which the seller accepted. So there I was, in need of a new case.

The most typical user would have sorted my need out with an ATX tower and called it a day. But I'm not like that. My criteria were such:
1. It had to be micro-ATX compatible,
2. It had to be as small as possible,
3. It had to be quirky, but not too extravagant,
4. It had to have OK airflow to support some play with the CPU PL values in case I run into a bottleneck with the 2070.

With these criteria, my choice fell on the Corsair 280X. Interestingly, the standard version sells around £100-110, while the RGB one costs £160. In my opinion, it really isn't worth paying 1.5x for a couple of RGB fans and a controller that probably has a proprietary control to make sure you can't get away without Corsair's iCue software (which isn't bad, but I didn't want it this time), so I decided for the standard one. It arrived a couple days ago which gave me ample time to set it up during the weekend. My initial thoughts about the case:

• It looks great, and considering its compact size, it's quite easy to install a PC in it.
• Paintwork isn't the best quality everywhere. I moved its fans around, and even just unscrewing a preinstalled screw made the paint around the hole chip off a bit.
• The motherboard standoff / screw combination Corsair supplied is quite crap. I managed to break the thread on 2 out of 8 standoffs at first try without even applying force.
• It's the first case I've seen with no HDD LED on it. When I was connecting the cables to the motherboard, I kept looking for the HDD LED one for about 5 minutes.
• Airflow is alright for non-overclocked systems if you occupy all 4 fan spaces (2 front intake, 2 top exhaust), but overclocking might need you to remove the dust filters at least. Both the 2070 and the 11700 locked to 65 W with the be quiet! Shadow Rock LP top out at 70-72 °C with minimal noise (the 2070 can be heard at full blast, but oh well).
• It has option for a 280 mm AIO on the top, which made me wonder about future upgrades.

All in all, I'm happy with both the EVGA 2070 and the Corsair 280X. The small Aerocool case will now get the Ryzen 3 system back and move to the living room as a movie watching PC. I'm only wondering which graphics card I should use: keep the 1650 and enjoy 4K with the option to game in the living room if I ever decide to do so, or sell it and use my GT 710 and let 4K 30 Hz or 1080p suffice.

Finally some pictures of the build in its current state:

 

[AusWolf]

Curiosity bites me again...

I had a Corsair H100i Platinum locked away from my last build. Since mounting it is no longer an issue with the new case, I decided why not. No pictures this time, just imagine the 240 mm radiator on the top of the case with the fans on it working as exhaust in a push configuration - pretty standard stuff. At first I was thinking about upgrading to a 280 mm AIO, but 1. the resale value of used liquid coolers isn't the best, and then 2. I wanted to try if this one fits in the case first, as most of you probably know that having fan mounts doesn't always mean you can actually use that size of rad and fan in every situation.

As it turned out, the 120 mm fans give just enough clearance for the CPU power cable and the million cables coming from the AIO pump to go through the back compartment of the case, so fitting a 280 mm rad with fans on the top of the case would have been quite problematic. There's plenty of space in the front, but I didn't want to use a radiator as intake, as I wouldn't want to heat up the graphics card and other components unnecessarily. Maybe if I upgrade to a 280 mm one day, I'll mount it there, but not for now.

As for the results:

• I disabled the power limits on the i7-11700, so it runs at 4.8-4.9 GHz single-core and a stable 4.4 GHz all-core.
• In Cinebench R23 (in a 10-minute loop), the single-thread score hasn't changed (1500), but the multi-thread one jumped from around 9,000 to 14,000 - a 55% increase.
• During that Cinebench loop, power consumption peaked at 45 W single-thread and 175 W multi-thread, the latter of which is a 2.7x increase over the stock 65 W power target.
• Temperatures peaked at 55 °C in the single-thread, and 82 °C in the multi-thread loop, which is about the same as it was with the small air cooler and stock power limits.

Conclusion:

I have to mention, I don't like working with the hundred million cables the Corsair AIO has. I also don't like the fact that the pump has a single PWM line connecting to the motherboard, and gets power from SATA. This might be a reason to upgrade to another AIO with less RGB bling and simpler connections in the future. I'm already eyeing the EVGA CLC.

As for the cooling performance - it's okay. I mean, 175 W is not an insignificant power draw from a locked CPU working at factory speeds, but the Corsair AIO manages to keep it around 80 °C, which isn't great, but isn't bad, either. This might also convince me to upgrade in the future - I'll just have to mount a 280 mm rad as intake, which I'm a bit reluctant to do. Opinions, anyone?

 

[Valantar]

I'm already eyeing the EVGA CLC

Just a note: if you're going to upgrade your AIO, go Arctic or EK - they are a clear step above the rest in terms of performance. Personally I would go Arctic, as their fans are far superior to EK's Vardar fans - just that much quieter.

Otherwise, interesting to see how a 240mm AIO handles the "65W" 11700 without power limits. 80°C is perfectly acceptable - the CPU can handle another 30° without being even remotely close to taking damage, so that's a perfectly fine result.

 

[AusWolf]

Just a note: if you're going to upgrade your AIO, go Arctic or EK - they are a clear step above the rest in terms of performance. Personally I would go Arctic, as their fans are far superior to EK's Vardar fans - just that much quieter.

Thanks for the tip. EK AIOs aren't really available in the UK, but I'll keep an eye out. The Arctic is on my "I might buy it at some point" list, next to the EVGA.

Otherwise, interesting to see how a 240mm AIO handles the "65W" 11700 without power limits. 80°C is perfectly acceptable - the CPU can handle another 30° without being even remotely close to taking damage, so that's a perfectly fine result.

It's also interesting to see the "65 W" 11700 sipping 175 W at factory speeds. I wouldn't really complain as the 5950X consumed just as much and ran even hotter when I enabled the Asus Optimiser on the TUF B550M (even though it has double the cores, but never mind).

Acceptable temperatures are interesting too. The 11700 has a Tjmax of 100 °C, but mine cooled by air started thermal throttling at around 90-95. I'm happy with 80-82 as long as it doesn't go any higher in the famously hot British summer.

 

[ThrashZone]

Hi,
I'd look at the warranty first before jumping on a ek aio believe it's only 3 years verses many others at 5 years.

 

[Valantar]

Hi,
I'd look at the warranty first before jumping on a ek aio believe it's only 3 years verses many others at 5 years.

EK's web shop lists them with a 5-year international warranty, though I guess warranty terms can vary depending on your location.

Thanks for the tip. EK AIOs aren't really available in the UK, but I'll keep an eye out. The Arctic is on my "I might buy it at some point" list, next to the EVGA.

They're really new though, so it might take some time for them to become available. They also don't have the most mainstream network of distributors, which is probably a hindrance for this kind of product. But in that case, Arctic will be the clear winner in terms of both cooling capacity and noise. I just installed a set of Arctic P14s in my Meshlicious, and they are likely the best fans I've ever used, Noctua included.

Thanks for the tip. EK AIOs aren't really available in the UK, but I'll keep an eye out. The Arctic is on my "I might buy it at some point" list, next to the EVGA.


It's also interesting to see the "65 W" 11700 sipping 175 W at factory speeds. I wouldn't really complain as the 5950X consumed just as much and ran even hotter when I enabled the Asus Optimiser on the TUF B550M (even though it has double the cores, but never mind).

Acceptable temperatures are interesting too. The 11700 has a Tjmax of 100 °C, but mine cooled by air started thermal throttling at around 90-95. I'm happy with 80-82 as long as it doesn't go any higher in the famously hot British summer.

Is a hot British summer defined as one where the rain mostly evaporates? Joking aside, not throttling with a CPU at 175W is nothing to scoff at regardless of the cooler - given how concentrated the heat in a CPU is and how relatively inefficient transfer is with an IHS in between, CPUs aren't easy to cool regardless of heat output. My 5800X can get pretty close to its throttling point under torture loads, but then the Aquanaut CPU block I'm using is ... meh at best (well, quite frankly it's sub-par, but there aren't many options for DDC pump+block combos!). And I don't tend to run torture workloads anyhow It sitting in the high 60s while gaming, in the same loop as a 275W GPU with just a single 280mm rad? I'm pretty happy with that.

 

[AusWolf]

Is a hot British summer defined as one where the rain mostly evaporates? Joking aside, not throttling with a CPU at 175W is nothing to scoff at regardless of the cooler - given how concentrated the heat in a CPU is and how relatively inefficient transfer is with an IHS in between, CPUs aren't easy to cool regardless of heat output. My 5800X can get pretty close to its throttling point under torture loads, but then the Aquanaut CPU block I'm using is ... meh at best (well, quite frankly it's sub-par, but there aren't many options for DDC pump+block combos!). And I don't tend to run torture workloads anyhow It sitting in the high 60s while gaming, in the same loop as a 275W GPU with just a single 280mm rad? I'm pretty happy with that.

That's really impressive considering the concentrated heat I've experienced with modern Ryzen. I wouldn't dare using a single 280 rad with both a CPU and GPU.

Actually, I'm not even sure if an upgrade from 240 to 280 is worth the investment, especially if I have to use it as intake and sacrifice cool air blowing towards the GPU.

 

[Valantar]

That's really impressive considering the concentrated heat I've experienced with modern Ryzen. I wouldn't dare using a single 280 rad with both a CPU and GPU.

Actually, I'm not even sure if an upgrade from 240 to 280 is worth the investment, especially if I have to use it as intake and sacrifice cool air blowing towards the GPU.

Yeah, as I said, I'm pretty happy with it overall. It's kind of liberating taking a step away from the oft-repeated truisms found in a lot of water cooling discussion, like "water temps should be close to ambient" and similar ideas with dubious claims to wisdom. I've seen my liquid hit ... I think 42 or 43, which I see no problem with. I mean, the hotter the liquid, the more efficiently the radiator will dissipate the heat, so as long as component temps are fine I really see no issue at all. I didn't clean my loop whatsoever for three years, just replaced the liquid once, and it was near spotless, so even if this new loop runs a bit hotter than that (that was with 240+120mm rads in a bigger case) I'm not worried about growth or anything. (I only use premix coolant btw.) I might undervolt my new GPU when I get it just for some better efficiency and lower heat load dumped into the room, and I'll be experimenting with curve optimizer for my CPU as well, but so far I've been happy with its cooling performance (including a ridiculous early summer heatwave sending the room temperature here near 30°C), so I won't bother putting too much work into it.

I don't think I'd bother upgrading if it's only cooling the CPU - as long as temps are in control at a reasonable noise level, the only difference is how looking at the numbers feels. Changing the fans for something better/quieter might be a more sensible route (if at all necessary).

 

[AusWolf]

It's kind of liberating taking a step away from the oft-repeated truisms found in a lot of water cooling discussion, like "water temps should be close to ambient" and similar ideas with dubious claims to wisdom.

That made me laugh. If water temp is near ambient, it means there's barely any heat transfer between the components and coldplates. You might as well not attach the block, just plug the pump's power connector in, and enjoy the cold water in the loop. Alternatively, downclocking your CPU to its idle speed should do the trick.

I mean, the hotter the liquid, the more efficiently the radiator will dissipate the heat, so as long as component temps are fine I really see no issue at all.

That! If I had to measure the efficiency of a loop, I would base my findings on the difference between water temperatures at the intake and exhaust pipes of CPU/GPU blocks at full load, and not the base water temperature of the outgoing loop.

I don't think I'd bother upgrading if it's only cooling the CPU - as long as temps are in control at a reasonable noise level, the only difference is how looking at the numbers feels. Changing the fans for something better/quieter might be a more sensible route (if at all necessary).

You're making sense. Max CPU temp is around 80-82 °C which isn't bad from a modern CPU, especially at 175 W power consumption. Noise could be a little lower, but 1. a manual fan curve would help, only if I weren't too lazy to create one, 2. it produces about the same noise as the GPU fans, which is loud, but not loud enough to be annoying while gaming, 3. a 280 rad would probably have to go to the front as intake, cooking the GPU which is now enjoying a comfortable 73 °C max load temp. The drawbacks outweigh the benefits, I guess.

 

[Valantar]

That made me laugh. If water temp is near ambient, it means there's barely any heat transfer between the components and coldplates. You might as well not attach the block, just plug the pump's power connector in, and enjoy the cold water in the loop. Alternatively, downclocking your CPU to its idle speed should do the trick.

Yeah, old-school water cooling fans have some weird cultural norms, like radiators having a fixed amount of heat they can dissipate (typically at 1GPM, though fans and airflow seems to be entirely excluded from this number, which ... ). Which in turn leads to people buying massive cases and stuffing them full of a stupid amount of radiators. "Hey, I heard you can only dissipate 65W per 120mm radiator, so I now have 16x120mm rads in my case to account for my 3090 and OC'd 11900K". Of course a lot of these people also insist on fans never running above their minimum fan speed, even in cases where a minor increase would drastically improve performance without a perceptible increase in noise (not to mention that more fans = more noise, no matter how slowly they're running, and if you're running your fans so slow they barely move air through the rads, fewer, faster fans would likely perform better at the same noise level).

I think the "water temps shouldn't exceed ambient" thing stems from early water cooling builds not using any kind of growth inhibitors, and warmer water accelerating the growth of algae and other gunk. Though ... algae still grows at 20°C. Or 15. Or 10. You can't cool your way out of growth. It will just be a bit slower. Heck, you have more of a chance of reducing growth in the loop by heating it up! Still, just using a proper growth inhibitor and/or kill coil will do the trick nicely.

Still, it's fascinating how this collection of truisms and cultural norms serves to maintain those rather pointless "I have six 360mm radiators!" builds. Like, how much benefit is that sixth radiator giving you? Or the fifth? Or fourth? Or third? Of course, if you accept those dubious truths and adopt them as goals, you do need to go that over the top, as there's no other way you'll reach those weird and unnecessary goals - ironically, given that the lower temperature delta lowers thermal transfer, you end up needing that many radiators and those huge volumes of water to maintain any type of efficiency in actually venting the heat to the room at that point. Yes, 30°C water will cool a CPU slightly better than 40°C water. But ... does it matter? I'd argue not.

 

[cvaldes]

Let those people do what they want. I'm okay letting a few nutjobs spend tons of money on too many radiators. That keeps the custom cooling components industry in business. It's their money not yours.

And if they use faulty logic to justify their purchases, that's fine as well. It's not like they are hurting anything except their own bank accounts.

 

[Valantar]

Let those people do what they want. I'm okay letting a few nutjobs spend tons of money on too many radiators. That keeps the custom cooling components industry in business. It's their money not yours.

And if they use faulty logic to justify their purchases, that's fine as well. It's not like they are hurting anything except their own bank accounts.

Well, obviously not - they can do whatever they want. I was just pointing out how following the accepted wisdom of the enthusiast water cooling scene tends to be rather problematic. I generally don't agree that letting people waste their money is okay though - you never know the context of people's lives or how healthy their habits are, and compulsive shopping and spending hurts a lot of people. Advocating for moderation and sensibility in spending is always a good idea. For those rich enough to not care, they'll sadly have to live with that. I'm sure they can take it.

 

[MentalAcetylide]

Still, it's fascinating how this collection of truisms and cultural norms serves to maintain those rather pointless "I have six 360mm radiators!" builds. Like, how much benefit is that sixth radiator giving you? Or the fifth? Or fourth? Or third?

Who the hell puts that many radiators in their case? Are they running a render farm with the cpu, RAM, and 4 graphics cards all overclocked? lol. FFS, I had a hard enough time acquiring a case that supported two 360mm radiators with room for 4 more fans.