Friday, July 16, 2021

 2021 07 16                                 Forte Performance 001

Forte System:

  • AMD Ryzen 9 5950X CPU with 16 cores and 32 threads, 7nm technology
  • G Skill Trident Z Neo F4-4000 32GB Memory
  • ASUS ROG Crosshair VIII Dark Hero Motherboard
  • be quiet brand BK022 Dark Rock Pro 4 CPU cooler
  • WD Black 1TB M.2 NVMe PCIe 4 Drive
  • WD Gold 10TB Enterprise Class rotating SATA disk drive
  • VisionTek Radeon 5450 Graphics card

So far I haven't been able to make the AMD Ryzen 9 5950X CPU work very hard. The highest temperature that it has ever reached, as far as I know, is 65 degrees C. It could easily go up to 85 and higher without harm, but when I tried to manually overclock it a little, the performance got worse, so I set that part of the BIOS back to AUTO.

                           
                  Precision Boost Overdrive

The motherboard has mysterious functions called "Precision Boost Overdrive," which can actually be found in at least three places in the BIOS setup. In all three places that function defaults to "AUTO," with choices of Auto, Disable, Enabled, and Advanced (or Manual):

  • One place to find it is Advanced > AMD Overclocking > AMD Overclocking > Precision Boost Overdrive. 
  • Another is Extreme Tweaker > Precision Boost Overdrive, where there are five available settings, all set to AUTO, including one called (you guessed it) >
  • Precision Boost Overdrive (again!).

When I stress the system with either the CineBench or the HeavyLoad application, every core always shows 100% utilization and the clock frequency of the individual cores can briefly jump as high as 5040 MHz, even though the motherboard is set to the default of 3400 MHz The individual core voltages change as well, as does, of course, the power consumed (and dissipated) by each core. I don't know why the mobo is doing this when, to my knowledge, overclocking is not enabled.

Forte Under Load

I've made changes to the settings of Precision Boost Overdrive (PBO) to disable it, in one or two places, but those changes made no difference in the system's performance with CineBench or HeavyLoad. I'm a little perplexed. PBO has officially been left untouched in this system.

Note: Regardless of PBO, this 16-core system's performance in the 10-minute CineBench R23 comparison test is very good indeed, behind only an AMD 32-core Threadripper system, so whatever is going on is OK with me. But I'd like to understand it.

In a previous experience with a slightly different mobo, the ROG Crosshair VIII Hero (not "Dark" Hero), I found a BIOS entry which could limit the CPU temperature. I haven't found that on this Dark Hero mobo yet, but I suspect it's there somewhere. It seems like that's what's happening - the mobo is automatically overclocking the CPU but also protecting it by limiting the CPU's temperature.

                                                                  Precision Boost 2

Enter Precision Boost 2. See https://www.amd.com/en/support/kb/faq/cpu-pb2  It looks like ASUS does automatically boost the system's clocks and other parameters when under load, using the mathematics of Precision Boost 2 (PB2), which, curiously, is quite different from Precision Boost Overdrive (PBO) despite the similarity in name. PB2 is enabled automagically as the default, and I haven't even found where it's enabled or could be disabled. I think that's where the CPU temperature is being limited, maybe not directly but as a result of some other limitations in PB2, such as current draw, core voltage, or power consumption (current * voltage).

                                                                       EZ System Tuning

In the EzMode screen, in the upper right corner, is EZ System Tuning. Theoretically this allows the selection of "improved system performance," "energy savings," or "normal." It doesn't really work though, as there is something quite wrong with the ASUS BIOS software: Once you have selected "improved system performance," resulting in a CPU clock of 4000 MHz, you can't unselect it again. Even if you get EZ System Tuning to show "normal" again, the CPU clock remains at 4000 MHz, even after a reboot. It's a software bug in the ASUS BIOS, version 3601, Dark Hero.

However, it does improve the system performance. Before selecting "improved system performance" the best CineBench score was 25,076, but with "improved system performance" selected it reached 26,402, an improvement of more than 5%. In my world that's not very important, but it might be important to some gamers. I had to clear the CMOS to get it back to normal, where the CPU clock again starts at 3400 MHz

                                                                     EZ Tuning Wizard

This is not so much a wizard as an assassin. From EZ Mode, I select EZ Tuning Wizard (at the top) > PC Scenario = Daily Computing > Tower Cooler > It says CPU Performance 163% up, DRAM 139% up. That sounds like a LOT, and indeed the computer failed to boot. It tried several times, all by itself, but failed. After I pressed the back panel Clear CMOS button, it booted up OK, no evident harm done, except for clearing all previous selections in the CMOS. The same thing happened when I selected Gaming/Media Editing as the PC Scenario. I'm quite glad that the Clear CMOS button was there. This seems to be another bug in the ASUS BIOS, version 3601, Dark Hero.

                                            "be quiet" brand BK022 Dark Rock Pro 4 CPU cooler

The humongous "be quiet" BK022 Dark Rock Pro 4 CPU cooler is obviously doing a fine job, keeping the CPU down to 62-65 degrees while it is dissipating 125-130 watts of heat. The cooler is huge, but since it's an air cooler, it will never leak water. 😊 By the way, I was ultimately able to add a third fan on the back of that cooler. Thus it has one in the front, one in the middle, and that extra one in the back, all blowing the same direction of course, out the back, where there is a fourth fan, a case fan, mounted an inch and a half away, taking that warm air directly out. Put your hand there, you can feel the warmth.

The entire computer has 12 very very quiet fans, eight of them RGB, either red or blue. I like it.


                                                 Change of Graphics Card and Monitor

VisionTek Radeon 5450 Graphics Card

I don't really understand how the CPU and the graphics card work together, so I though it might be possible that the CPU was somehow limited by the speed or capability of the graphics, and for that reason could not consume more than about 100 watts of power in the cores, preventing it from exceeding a temperature of about 65 C. Forte has a VisionTek Radeon 5450 Graphics card talking to a $70 16-inch TV set in HDMI mode, and my working computer here, Stirling 2021, has an ASUS NVIDIA GeForce GT 710 card talking to two inexpensive 24-inch HP monitors. Neither system is a gamer's dream, to be sure, but at least they're different from each other. Therefore, I wondered if I would see a difference if I swapped Stirling's graphics elements into Forte.

Bottom line - it made no discernable difference. The Forte CPU was still limited to about 65 C. 

                                                        I Give UP

For now. I'm unable to make the AMD Ryzen 9 5950X CPU reach a temperature above 65 C, though it could certainly handle more work and thereby a higher temperature. More research to be done here. In the meantime, Forte is nonetheless a formidable consumer-grade computer.


Friday, July 9, 2021

2021 07 07                                  FORTE WORKS!

The new computer, Forte, is working, and it's pretty powerful. In the CineBench R23 test it beats everything except the Threadripper 2990WX, the AMD 32 core 64 thread predecessor. The 16 core 32 thread Ryzen 9 5950X in Forte scores 25076, where the Threadripper's CineBench score is listed as 30054. Forte also scores just ahead of Intel's Xeon W-3265M, with its 24 cores and 48 threads.
CineBench content creation application


Just as important, Forte does this in a walk. The maximum CPU temperature reported by HWMonitor is 65 C, and it mostly runs about 60 to 62. The CPU can easily handle 80, even 90. It's loafing! The huge Dark Rock Pro 4 air cooler is cool to the touch, barely warmer than the case. According to HWMonitor the CPU "package" is consuming (and therefore dissipating) about 125 watts, while the cores themselves are burning about 105 watts of that, each core about 6.5 watts.

The Forte CPU is unlocked, meaning that it can be overclocked. However, as far as I know it isn't overclocked. I tried manually overclocking it slightly, but the CineBench performance got worse, so it's back to "reset" values. Nominally that's 3400 MHz, though HWMonitor does seem to show that the individual core clocks run from 3800 MHz all the way up to 5040 MHz on occasion during the CineBench test, and I don't know why it's doing that. Another CPU test program is HeavyLoad, and during a HeavyLoad test the clocks run a little over 4000. Clearly, I have something to learn yet about overclocking the Ryzen 9 5950X on the ASUS ROG Crosshair VIII Dark Hero motherboard.
Missing a fan in he upper left of the
photo. Note the HUGEness of the "be
quiet" air cooler. The memory modules
almost disappear beneath it.


In the meantime, the computer isn't quite complete. There are two red LEDs on the top front of the case that blink continually, as if the case thinks that the computer inside is asleep, though it's wide awake and working hard. There's also an empty two-pin header on the motherboard called PLED that might be related, and I suspect that there's a two-pin connector from the case buried somewhere that still needs to be connected. If that exists, I haven't found it yet. It's a mystery.

By the way, Forte really is quiet. Even when the package is dissipating 127 watts or so, and the fans are keeping the CPU temp down to 60-62 degrees, the fans are inaudible with all of the skins off the case. I admit to being somewhat hard of hearing, but I have pretty good hearing aids and this computer is QUIET!

                                                     NewEgg Shipping

Today is July 9.
There's a fan missing. On Thursday, July 1, I ordered two Phanteks 140mm LED case fans from NewEgg, one red, one orange (in case they couldn't scrounge up a red one). According to NewEgg those were shipped by NewEgg, one from Fontana CA on July 1, and the other from Indianapolis IN on July 2. One of them stopped briefly in Stillwater, MN (where I live), and then proceeded to the US Postal Service in Brooklyn Park MN, not to be seen since. The other went directly to Brooklyn Park where it too disappeared. Both were shipped by the UPS "Mail Innovations" service, which UPS describes as follows: "Mail Innovations is not a day definite service."

Well I guess not! According to UPS one of those fans has an estimated delivery of Tuesday, July 06 by 7:00 PM, and the other Wednesday, July 07 by 7:00 PM. Today is Friday, July 09, all day. No fans yet.

One could argue that a delay should be expected because of the intervening Fourth of July weekend. OK, that sounds reasonable, except that the Fourth of July was not a surprise this year. In fact it happens every year on or about July 4. It's actually possible to plan for it, and even predict an accurate delivery date despite the interference of the July 4 festivities. 

Note that it's expedited. Certainly glad it's not
just a normal shipment



I'm totally underwhelmed by the service of NewEgg, UPS, and USPS. In fact, in the meantime, I ordered a different fan from Amazon (for a different computer) on Tuesday, July 6, and received it the morning of Thursday, July 8. Excellent.

The USPS seems to be going downhill fast. Letters within the metro area here, once dependably prompt, now take an indeterminate number of days. The same is true of packages. But NewEgg and UPS know this and persist in using a delivery service which is not "day definite" and not even reasonably prompt. The only solution I have for this is to avoid NewEgg whenever possible. In fact it would have been possible in this case, and if I had done it my computer would be complete and I could button it up. Grrrr.

                                                              UPDATE  Saturday, July 10, 2021

UPS Tracking now says that one of the fans is out for delivery today, only 9 days after the order. No time specified. Nothing on the other fan - apparently it hasn't been in detention long enough yet. I'll post here when either of them arrives.

OK the first fan did arrive today at about 3:45 pm, only 9 days after the order. It's installed.

Saturday, July 3, 2021

 2021 07 03                           FANS

Bill of Materials
Most decisions have been made, the Bill of Materials has been decided and everything ordered and, in fact, most of the parts are here. The one exception is the graphics card. Good ones are available only at scalpers' prices. An inexpensive, low-performing card was ordered and is here; we will see how well it works.



Fans:

In my opinion, fans perform two somewhat different functions:

1. First: They bring in cool air from outside the case, pushing out warmer air. Whether a particular case fan is blowing in or out, it circulates air from the outside to the inside. Consumer-grade computer cases are VERY leaky and meant to be; there is no chance of actually creating a significant pressure in the case, either positive or negative, unless you want to use a lot of duct tape. Air movement in and out is the point. The Phanteks Enthoo Pro "PH-ES614PTG_BK" ATX case comes with one black fan in the front, blowing in, and one in the back, blowing out. In addition, the EVGA power supply has its own fan, pulling air in from outside the case at the back, and exhausting it downward out of the case, or vice-versa, I'm not sure. That fan only goes on when the power supply needs it. I have added five more very-quiet but colorful fans, three blowing into the case and two blowing out. In addition I have replaced the provided front and back black fans with colorful ones. They're the same fans, from the same company, but with white blades and LED illumination.

2. Second: The fans move internal air over the surfaces of components that need that air movement to remove their heat. Best example: the Dark Rock Pro 4 CPU Cooler needs a lot of air flowing over its cooling fins, to remove 100-200 watts of CPU heat. The cooler actually came with two of its own fans for that, and I might add a third. OK I just ordered the third, a quiet 120mm fan. The G Skill Trident Z Neo memory modules also need air flow, as does the WD Black M.2 PCIe 4 disk drive. So does the ASUS ROG Crosshair VIII Dark Hero motherboard (where do these names come from?) as well as the graphics card, regardless what that may eventually be. Further, to improve the air flow within the case, I added a very quiet blue 200mm fan (Big Blue Honker) inside the case blowing on the CPU cooler, the memory sticks, the M.2 PCIe Gen4 drive, the graphics card(s), and of course the motherboard itself.

Fan noise: In my opinion, fan noise has more to do with the speed of the air that's moved than with the amount of air moved. It's about turbulence, especially turbulence right where the fan blades and the air collide. For this reason a fan should rotate as slowly as possible while moving air as efficiently as possible. It seems clear that fans are not all equal in this regard. Some are much better than others, and size does matter, the bigger and slower the better. As an example, the Phanteks PH_F120SP, a 120mm fan, moves air at 54.4 CFM at 1300 RPM with 24.2 dBa of noise, while its bigger brother, the PH-F140SP, a 140mm fan, moves 82.1 CFM at 1200 RPM with 19 dBa of noise. Fan blade design makes a difference too, and may be important in this comparison.

By the way, dBa simply means dB (decibels) over no noise at all. A rock sitting perfectly still on the ground generates 0 dBa of noise. A rock band, on the other hand, may generate as much as 120 dBa of sound (or noise, depending on the perceiver). Of course that much sound can immediately damage a person's hearing. Normal conversation may be about 60 dBa. Technically it's a little more complicated than that, but those complications make no difference in this discussion.

Current state of assembly. Note:
cooler is not yet attached to computer.


A fan producing 19 dBa of sound is very, very quiet, barely audible, if at all, even with your ear right there. Adding another identical fan doesn't mean it's now 19+19=38 dBa. It's all logarithmic, where doubling anything increases the value by 3, so you simply add 3 to 19 and get 22 dBa for two identical fans. Well, that's convenient!

So then let's add another identical fan, to get three. What is the sound level now? I don't know, it's calculable using logarithms, or you can look it up in a table, but why bother? There's an easier calculation. If we add TWO new fans to the two we already had, for a total of four, we just doubled the sound again, so you guessed it - we add another 3 dB to get 25 dBa for the four fans. That level of sound is still very quiet, down in the barely audible range. If we double again, adding four more, we're at 28 dBa, close to what we have in this new computer, possibly less than 30 dBa, certainly less than 33. Some of the fans that you can buy develop 35 dBa all by themselves. Just one fan. Not buying that fan!

What does this tell us? First, it might be better to add several very quiet fans than a few noisy ones. Second, choose fans carefully! We're doing both of those things in this computer. Fans: Seven Phanteks PH-F140SP colorful LED 19 dB 140mm fans blowing in or out of the case (3 top, 1 bottom, 2 front, 1 back), one obscured fan in the power supply, two black ones in the Dark Rock cooler, (a third one ordered in an earlier paragraph of this post, destined for the back of the cooler, got too close to the mobo and was not attached), and an Aero Cool Silent Master 18 dB 200mm 800 RPM blue-LED Big Honker fan inside, almost in the middle of the case. I think that's all. We have a fan in every location that it makes sense to put one, plus Big Honker, for a total of 11 fans. Of these, eight are colored by red or blue LEDs (I deliberated about an orange one). Seven of those eight, the 140mm bunch, also have switches that will disable their LEDs but not their rotation if it just seems like too much of a circus in there.

A quick primer on fan wiring: Fans with four wires connecting to headers with four pins can have their speed controlled by pulse-width modulation (PWM), if so selected in the BIOS. In contrast, fans with three wires/pins are controlled by voltage, also called direct current. Four-wire PWM control is thought to be better, as it offers a greater range of control. The bigger the fan, the more significant this distinction becomes. End of primer.

One fan is taken care of already: the power supply fan is managed w

The hub is right in the middle. Power and PWM 
signal are at the ends, and four of the six output
connectors are occupied.

ithin the power supply. All of the remaining 10 fans require 12 volts to reach full speed. The Phanteks Enthoo Pro PH-ES614PTG_BK ATX Case has a fan-control module (hub) with a PWM input and six outputs. It's called the PWM HUB in the documentation, but that's pretty much exactly what IT IS NOT. It can receive a PWM (4-wire) signal from the CPU_FAN or OPT_FAN header on the motherboard, and separately receive 12V power from a SATA cable, but it cannot pass the PWM signal forward to any fan, because all six of its outputs are 3-wire. See primer above. Instead, it distributes voltage-control direct-current fan power to the six output connectors. Since we have 10 fans, I will use  some two-way fan splitters, thus allowing for plenty of fans to receive voltage control. There are other PWM fan headers on the ASUS ROG Crosshair VIII Dark Hero motherboard as well, and I will use one of those for Big Honker, which takes a little more power than any of the rest. In particular, there is a header called H_AMP (high-current). 

So let's connect one of the cooler fans to the mobo header called CPU_FAN, and the other cooler fan to CPU_OPT. Both headers are 4-wire, meaning that they will send PWM fan control, and both fans are 4-wire, meaning that they will receive pulse-width modulation signals and act upon them to control their speed, and tell the headers how they're doing. I assume that the CPU_FAN and CPU_OPT headers will provide the same PWM signal. The manual doesn't say so, but the internet seems to think that CPU_OPT is slaved to CPU_FAN. That takes care of the CPU cooler fans in the best possible way, connected directly, no hub involved. The CPU fans will both respond to the actual needs of the CPU.

We have eight fans left, connecting most of them to the alleged "PWM Hub." We still need a 4-wire PWM signal from somewhere to correctly drive the hub. It turns out that there are four more fan headers on the motherboard, CHA_FAN1, CHA_FAN2, CHA_Fan3, and H_AMP. All of those headers are 4-wire, and I believe that all of them can be configured in the BIOS to send a PWM signal. I hope. The header I like the best is CHA_FAN1, because it is located in the middle of the motherboard next to the M.2 disk drive, the graphics card, and close to the CPU itself. The PWM signal sent by CHA_FAN1 will be different from that sent by the CPU, more specific to the motherboard's needs. That's a good thing. We will connect the seven identical case fans to the hub, and connect Big Honker to H_AMP. Done connecting fans.