Mussels
Freshwater Moderator
- Joined
- Oct 6, 2004
- Messages
- 58,412 (7.75/day)
- Location
- Oystralia
| System Name | Rainbow Sparkles (Power efficient, <350W gaming load) |
|---|---|
| Processor | Ryzen R7 5800x3D (Undervolted, 4.45GHz all core) |
| Motherboard | Asus x570-F (BIOS Modded) |
| Cooling | Alphacool Apex UV - Alphacool Eisblock XPX Aurora + EK Quantum ARGB 3090 w/ active backplate |
| Memory | 2x32GB DDR4 3600 Corsair Vengeance RGB @3866 C18-22-22-22-42 TRFC704 (1.4V Hynix MJR - SoC 1.15V) |
| Video Card(s) | Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.750v (375W down to 250W)) |
| Storage | 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2 |
| Display(s) | Phillips 32 32M1N5800A (4k144), LG 32" (4K60) | Gigabyte G32QC (2k165) | Phillips 328m6fjrmb (2K144) |
| Case | Fractal Design R6 |
| Audio Device(s) | Logitech G560 | Corsair Void pro RGB |Blue Yeti mic |
| Power Supply | Fractal Ion+ 2 860W (Platinum) (This thing is God-tier. Silent and TINY) |
| Mouse | Logitech G Pro wireless + Steelseries Prisma XL |
| Keyboard | Razer Huntsman TE ( Sexy white keycaps) |
| VR HMD | Oculus Rift S + Quest 2 |
| Software | Windows 11 pro x64 (Yes, it's genuinely a good OS) OpenRGB - ditch the branded bloatware! |
| Benchmark Scores | Nyooom. |
A friend of mine came around with some old water cooling hardware and wanted to build up a retro gaming PC with modern cooling and an SSD, originally intending to SLI two of his GTX 280 cards.
Since I still had my 2500k and motherboard from many years before (It went to family who returned it after upgrading a decade later) I hunted down some matching sticks of RAM from another friend, and let my friend know I was good to go to get started the next day. The board I had didn't support SLI, so he went home to see if he had anything else to try and use.
He never messaged me back as he'd passed away in his sleep that night from heart failure.
He wasn't even 30.
Time passed by and I decided it was time to finish off this build and overclock the crap out of this CPU, even into unsafe voltages and temperatures... and review an old motherboard from back in the era when I first met my friend and the dumb thing's we'd do at LAN parties back then - we blew up plenty of fuse boxes and managed enough power draw once to set the actual powerlines to the building sparking in the rain.
The actual parts he gave me were really mixed with different tubing sizes, brands, and not enough of anything - but I made it all work and added what was missing.
Since I still had my 2500k and motherboard from many years before (It went to family who returned it after upgrading a decade later) I hunted down some matching sticks of RAM from another friend, and let my friend know I was good to go to get started the next day. The board I had didn't support SLI, so he went home to see if he had anything else to try and use.
He never messaged me back as he'd passed away in his sleep that night from heart failure.
He wasn't even 30.
Time passed by and I decided it was time to finish off this build and overclock the crap out of this CPU, even into unsafe voltages and temperatures... and review an old motherboard from back in the era when I first met my friend and the dumb thing's we'd do at LAN parties back then - we blew up plenty of fuse boxes and managed enough power draw once to set the actual powerlines to the building sparking in the rain.
The actual parts he gave me were really mixed with different tubing sizes, brands, and not enough of anything - but I made it all work and added what was missing.
Intel 2500K that i've been keeping in storage for the last 10 years, when I moved to a shiny 4770k.
Gigabyte GA-Z77-D3H - New CMOS battery and thermal pads
2x4GB DDR3 1333
2x4GB DDR3 1600
Team GX1 1TB SATA SSD
Corsair HX750i PSU
MSI GTX 1080 8GB Gaming X / Gigabyte RTX 3070 8GB
Thermaltake P3 Pro TG case
Corsair Commander Core XT (Fan control + Temp probes) - This turned out to not be needed on this board as it has very accurate VRM thermal sensors.
Elmor Labs USB-C PMD Power meter (measures EPS+PCI-E power consumption)
And a custom cooling loop that exists for the special reasons stated in the introduction.
Gigabyte GA-Z77-D3H - New CMOS battery and thermal pads
2x4GB DDR3 1333
2x4GB DDR3 1600
Team GX1 1TB SATA SSD
Corsair HX750i PSU
MSI GTX 1080 8GB Gaming X / Gigabyte RTX 3070 8GB
Thermaltake P3 Pro TG case
Corsair Commander Core XT (Fan control + Temp probes) - This turned out to not be needed on this board as it has very accurate VRM thermal sensors.
Elmor Labs USB-C PMD Power meter (measures EPS+PCI-E power consumption)
And a custom cooling loop that exists for the special reasons stated in the introduction.
My camera really doesn't like my LED lighting here and it's doom and gloom weather so I've used some stock photos here - the stock ones are low res with some component identifying labels removed.
The Feature set is not exciting at all by 2023 standards, but it's also interesting to see how little has changed. Even a decade ago we had SATA 3, USB3, PCI-E 3.0 and very similar motherboard layouts to today.
The CPU: My old 2500K that achieved a 5.2GHz overclock on air, back when it was new.
Using this CPU has a negative of running the system at PCI-E 2.0, which is probably the only reason they aged out when they did - these are the best overclocking CPU's to ever exist and the performance gains were simply insane back then, reaching levels no stock CPU could reach for a very very long time.
I have a new test bench setup, at this stage I didn't have a fancy backdrop for photos. What I did have was an utterly chaotic mess of watercooling components my friend left me, that were quite 'fun' to get installed.
The pump block leaked as did the CPU block and I had only a single spare 120mm fan - changes had to be made, obviously.
(Pics of the completed setup coming soon)
The Feature set is not exciting at all by 2023 standards, but it's also interesting to see how little has changed. Even a decade ago we had SATA 3, USB3, PCI-E 3.0 and very similar motherboard layouts to today.
The CPU: My old 2500K that achieved a 5.2GHz overclock on air, back when it was new.
Using this CPU has a negative of running the system at PCI-E 2.0, which is probably the only reason they aged out when they did - these are the best overclocking CPU's to ever exist and the performance gains were simply insane back then, reaching levels no stock CPU could reach for a very very long time.
I have a new test bench setup, at this stage I didn't have a fancy backdrop for photos. What I did have was an utterly chaotic mess of watercooling components my friend left me, that were quite 'fun' to get installed.
The pump block leaked as did the CPU block and I had only a single spare 120mm fan - changes had to be made, obviously.
(Pics of the completed setup coming soon)
Understanding VRM's and how they work takes an engineering degree I do not have.
I can get the spreadsheets, but if any readers have information to share that they can back up with external sources - please do.
This board is old enough and has an unusual VRM layout - Gigabyte have slapped lowside FETs absolutely all over the place with some VRM doublers in an unusual asymmetrical setup, I'll need to probe the board with a multimeter to identify the phases and honestly, it's not worth the effort for a one-off review and overclock attempt.
12 years on, the VRM thermal pads were still soft and in working order. I did replace them with Arctic cooling 1mm thermal pads but these are still worth recognizing as long-term quality.
This board uses an IR3564 driver that supports 4+1 or 3+2 VRM setups, with at least one IR3598 doubler to K03B7 high side FET's and K0393 low side FETs.
It has three pairs of one K03B7 with two K0393, and then one pair of one each. Some markings have worn off over the last 10 years, making it hard to be certain of some of them.
It looks like a "3+1" setup with doublers on the CPU phases.
The K03B7 have a 30A power rating and a 'power temperature derating' showing that their wattage capabilities decrease massively as temperatures increase, with a maximum temperature of 150c but losing half their power capabilities at around the 75c mark. These should be used to push 12V down to 5V for the second set of FETs to lower even further to the CPU.
The K0393 have a 40A rating and overall very similar stats to the K03B7 with a similar temperature curve. Safe to 150c, but best kept under 75c - or lower.
If this is a doubled 3+1 setup, that would provide 6x 40A (240A) to the CPU at 25c, or half that (120A) at 75c.
They have "channel dissipation" at 30W and 40W matching their amps, so it seems they're fairly lossy and put out a lot of heat.
During the later overclock testing they reached 82C with just 65A of power draw, so it's likely to assume they'd thermal throttle long before reaching that theoretical 120A limit.
I can get the spreadsheets, but if any readers have information to share that they can back up with external sources - please do.
This board is old enough and has an unusual VRM layout - Gigabyte have slapped lowside FETs absolutely all over the place with some VRM doublers in an unusual asymmetrical setup, I'll need to probe the board with a multimeter to identify the phases and honestly, it's not worth the effort for a one-off review and overclock attempt.
12 years on, the VRM thermal pads were still soft and in working order. I did replace them with Arctic cooling 1mm thermal pads but these are still worth recognizing as long-term quality.
This board uses an IR3564 driver that supports 4+1 or 3+2 VRM setups, with at least one IR3598 doubler to K03B7 high side FET's and K0393 low side FETs.
It has three pairs of one K03B7 with two K0393, and then one pair of one each. Some markings have worn off over the last 10 years, making it hard to be certain of some of them.
It looks like a "3+1" setup with doublers on the CPU phases.
The K03B7 have a 30A power rating and a 'power temperature derating' showing that their wattage capabilities decrease massively as temperatures increase, with a maximum temperature of 150c but losing half their power capabilities at around the 75c mark. These should be used to push 12V down to 5V for the second set of FETs to lower even further to the CPU.
The K0393 have a 40A rating and overall very similar stats to the K03B7 with a similar temperature curve. Safe to 150c, but best kept under 75c - or lower.
If this is a doubled 3+1 setup, that would provide 6x 40A (240A) to the CPU at 25c, or half that (120A) at 75c.
They have "channel dissipation" at 30W and 40W matching their amps, so it seems they're fairly lossy and put out a lot of heat.
During the later overclock testing they reached 82C with just 65A of power draw, so it's likely to assume they'd thermal throttle long before reaching that theoretical 120A limit.
One bug I've encountered now on 5 separate occasions (All intel 2nd/3rd/4th gen) is that security updates from intel via windows update somehow corrupt BIOS on these older motherboards.
After a Windows 10 Update Today Overclocking is lost. WTF Microsoft and Intel??? | TechPowerUp Forums
This will never get officially fixed as these systems don't meet the requirements for 10/11.
After windows is installed and updated, you'll reboot and get stuck in the BIOS splash screen - nothing will get the system to POST other than triggering Gigabytes dual BIOS feature to load from the backup BIOS.
You can trigger the backup BIOS by pressing and holding the power button until the system powers on, then off, then release and press normally.
The other method is power and reset at the same time, or by shorting pins 1 and 6 on the primary BIOS chip while powering on the PC (you need three arms for this)
Once reset to the backup BIOS, I could use EZ-flash to update the BIOS and everything behaved normally again - but it may return every time a fresh windows install is ran on these systems.
After a Windows 10 Update Today Overclocking is lost. WTF Microsoft and Intel??? | TechPowerUp Forums
This will never get officially fixed as these systems don't meet the requirements for 10/11.
After windows is installed and updated, you'll reboot and get stuck in the BIOS splash screen - nothing will get the system to POST other than triggering Gigabytes dual BIOS feature to load from the backup BIOS.
You can trigger the backup BIOS by pressing and holding the power button until the system powers on, then off, then release and press normally.
The other method is power and reset at the same time, or by shorting pins 1 and 6 on the primary BIOS chip while powering on the PC (you need three arms for this)
Once reset to the backup BIOS, I could use EZ-flash to update the BIOS and everything behaved normally again - but it may return every time a fresh windows install is ran on these systems.
The above issue aside, BIOS from this era is very similar to the modern designs except it's rather unpolished. Some settings work with mouse, some work with the enter key, some only work with being typed directly into, and yet again others only with the + and - keys.
Many settings are renamed into brand specific names or poor translations. "System Agent" voltage for memory overclocking is instead renamed "IMC" and other things like that throughout the BIOS. It's a good thing to keep in mind vs a modern BIOS, where things are a lot better named and unified between brands.
Any errors or failed overclock settings result in a very fast boot to a 'Failed POST' screen that lets you go into the BIOS and adjust settings without losing what you had changed or like some modern boards will do, booting with stock settings and not informing you leading to thinking your custom settings applied when they did not.
This recovery takes about 5 seconds from the power button press and is overall extremely fast to use, it's definitely nicer than most modern boards achieve.
The "Simple" mode is a terrible gimmick, it has a mouse feature to let you select BIOS items... but they don't correspond to anything in reality. Clicking a southbridge heatsink would give you settings related to Dual BIOS which is in an entirely different section of the board, so this misleads users as to what parts of the board actually do specific tasks.
Despite being 12 years old at the time of writing, the board supports UEFI and Secureboot - the only thing missing for windows 11 support is a TPM 2.0 module.
This system seems to slightly over-volt from what the user has set, being an entry level board that's entirely acceptable as a form of LLC (load line calibration)
Screenshots failed to save on the USB drive I was using at the time, I took some photos instead to show how similar it is to a modern BIOS in 'advanced' mode.
While "simple" mode was a misleading, useless mess - that image isn't even the board in use, and BIOS chips are not under heatsinks on any board.
Many settings are renamed into brand specific names or poor translations. "System Agent" voltage for memory overclocking is instead renamed "IMC" and other things like that throughout the BIOS. It's a good thing to keep in mind vs a modern BIOS, where things are a lot better named and unified between brands.
Any errors or failed overclock settings result in a very fast boot to a 'Failed POST' screen that lets you go into the BIOS and adjust settings without losing what you had changed or like some modern boards will do, booting with stock settings and not informing you leading to thinking your custom settings applied when they did not.
This recovery takes about 5 seconds from the power button press and is overall extremely fast to use, it's definitely nicer than most modern boards achieve.
The "Simple" mode is a terrible gimmick, it has a mouse feature to let you select BIOS items... but they don't correspond to anything in reality. Clicking a southbridge heatsink would give you settings related to Dual BIOS which is in an entirely different section of the board, so this misleads users as to what parts of the board actually do specific tasks.
Despite being 12 years old at the time of writing, the board supports UEFI and Secureboot - the only thing missing for windows 11 support is a TPM 2.0 module.
This system seems to slightly over-volt from what the user has set, being an entry level board that's entirely acceptable as a form of LLC (load line calibration)
Screenshots failed to save on the USB drive I was using at the time, I took some photos instead to show how similar it is to a modern BIOS in 'advanced' mode.
While "simple" mode was a misleading, useless mess - that image isn't even the board in use, and BIOS chips are not under heatsinks on any board.
Using Ventoy to install Windows 11 from a USB 3.2x2 NVME drive (10Gb/s) simplifies things greatly, I have ISO files for Windows 10 and 11 as well as Macrium recovery and Linux Mint on the drive and i highly recommend this setup to anyone, even from a cheap USB flash drive as a recovery media.
Ventoy bypasses the secure boot and TPM requirements, meaning windows 11 installed without complaint on this older machine. The entire install process took about 3 minutes, before I was at the desktop.
As long as you use the "optional updates" feature in windows update, all drivers are installed and the board works perfectly.
It's easy to forget how long we've had SATA III and USB 3.0, since systems of this era were massively held back by mechanical hard drives - back in the windows 7 era, you'd need to use the USB 2.0 ports and install SATA drivers at the first stage of installing the OS.
Installing windows 7 on this machine would take several hours of installing, updates, reboots, updates, reboots, updates, reboots, more reboots, yet more reboots, and manually hunting drivers for every piece of hardware.
Windows 10 and 11 make this incredibly easy, and it's easy for people who never dealt with older operating systems to not be aware of that.
Debloated, windows 11 used 1.4GB of RAM. Installing an Nvidia GTX 1080 changed that clean boot to 1.8GB, using a debloated NVCleaninstall driver.
Keep that in mind next time you blame windows 10/11 for being bloated - 400MB of RAM lost installing an Nvidia card, let alone all the extra bloat from Geforce experience.
Windows 11 got confused by the CPU overclock (older versions of windows did as well) resulting in some fantastic and utterly incorrect values such as 8.20GHz clock speeds in task manager, instead of it's real 4.4GHz at the time.
Ventoy bypasses the secure boot and TPM requirements, meaning windows 11 installed without complaint on this older machine. The entire install process took about 3 minutes, before I was at the desktop.
As long as you use the "optional updates" feature in windows update, all drivers are installed and the board works perfectly.
It's easy to forget how long we've had SATA III and USB 3.0, since systems of this era were massively held back by mechanical hard drives - back in the windows 7 era, you'd need to use the USB 2.0 ports and install SATA drivers at the first stage of installing the OS.
Installing windows 7 on this machine would take several hours of installing, updates, reboots, updates, reboots, updates, reboots, more reboots, yet more reboots, and manually hunting drivers for every piece of hardware.
Windows 10 and 11 make this incredibly easy, and it's easy for people who never dealt with older operating systems to not be aware of that.
Debloated, windows 11 used 1.4GB of RAM. Installing an Nvidia GTX 1080 changed that clean boot to 1.8GB, using a debloated NVCleaninstall driver.
Keep that in mind next time you blame windows 10/11 for being bloated - 400MB of RAM lost installing an Nvidia card, let alone all the extra bloat from Geforce experience.
Windows 11 got confused by the CPU overclock (older versions of windows did as well) resulting in some fantastic and utterly incorrect values such as 8.20GHz clock speeds in task manager, instead of it's real 4.4GHz at the time.
If i listed every step involved in testing and re-testing this we'd be here a very long time, so I'll skip to the combinations that ran stable. Getting even this limited set of testing took four days, with a known CPU.
Despite it's age and low price tag, this board has some quite decent and accurate sensors. The VRM temperature readings always came out hotter than my physical sensors, meaning they were genuinely placed in the hottest spots of the components.
Many newer boards don't give you VRM sensor readings or wildly inaccurate ones, let alone two like this board does (One for the CPU 'VR Loop 1' and one for the RAM 'VR Loop 2') and this is something I definitely want to see exposed in every board.
Power consumption climbs massively when overclocking is involved but not just in the CPU - VRM's have a 'goldilocks zone' and once they pass out of that they start to lose a lot more power as heat, leaving a user with a well cooled CPU and a burning hot VRM.
Overheating VRM's can either result in system instability, or with modern CPU designs the CPU will 'sleep' briefly (for as short as 1ms) until the VRM's output the correct voltage again - this results in 'clock stretching' and stuttering performance.
Some VRM's are rated to high temperatures like the ones in this system, but just because it's stable at 120c doesn't mean it cant melt the plastic off your EPS power cable should they touch and risk an electrical short.
R23:
Stock: 2552
Stock + RAM @1866: 2697 (5.6% faster)
CPU OC 4.5GHz 1.375v 3332 (30.5% faster!)
5.0GHz: 3597 1.45v (40.9% faster)
5.3GHz + 1866 (1.60v!) 3970 (55.5% faster than stock)
Technically if a game uses under 4 threads this CPU is now faster than an i9. Not all of them, but the 7980XE at least.
This is a bigger difference than you get between most CPU upgrades.
A Ryzen 1800x to 5800x is a good example, where it has the same core count (8 with SMT) and performance went up by 69% - with 3 architectures of improvements behind that gain (Zen, Zen+, Zen2, Zen3) and here we are achieving a similar gain overclocking on an entry level motherboard.
Despite it's age and low price tag, this board has some quite decent and accurate sensors. The VRM temperature readings always came out hotter than my physical sensors, meaning they were genuinely placed in the hottest spots of the components.
Many newer boards don't give you VRM sensor readings or wildly inaccurate ones, let alone two like this board does (One for the CPU 'VR Loop 1' and one for the RAM 'VR Loop 2') and this is something I definitely want to see exposed in every board.
Power consumption climbs massively when overclocking is involved but not just in the CPU - VRM's have a 'goldilocks zone' and once they pass out of that they start to lose a lot more power as heat, leaving a user with a well cooled CPU and a burning hot VRM.
Overheating VRM's can either result in system instability, or with modern CPU designs the CPU will 'sleep' briefly (for as short as 1ms) until the VRM's output the correct voltage again - this results in 'clock stretching' and stuttering performance.
Some VRM's are rated to high temperatures like the ones in this system, but just because it's stable at 120c doesn't mean it cant melt the plastic off your EPS power cable should they touch and risk an electrical short.
R23:
Stock: 2552
Stock + RAM @1866: 2697 (5.6% faster)
CPU OC 4.5GHz 1.375v 3332 (30.5% faster!)
5.0GHz: 3597 1.45v (40.9% faster)
5.3GHz + 1866 (1.60v!) 3970 (55.5% faster than stock)
Technically if a game uses under 4 threads this CPU is now faster than an i9. Not all of them, but the 7980XE at least.
This is a bigger difference than you get between most CPU upgrades.
A Ryzen 1800x to 5800x is a good example, where it has the same core count (8 with SMT) and performance went up by 69% - with 3 architectures of improvements behind that gain (Zen, Zen+, Zen2, Zen3) and here we are achieving a similar gain overclocking on an entry level motherboard.
So many sensors! HWinfo has many sensors that all measure the CPU's power, and they all read different values.
This is the different readings when running the stock CPU + 1866MT/s RAM test, and why it matters to always check where readings are from and compare only with the same reading, for accurate results.
These are all from HWinfo, but different sensors
i5-2500K sensors:
CPU Package Power: 69.52W
IA Cores power: 64.3W
GA-Z77-D3H Sensors:
Power (POUT): 63.5W
Power (Input): 75.5W
Elmor Labs PMD-USB
EPS1: 81.4W
At any given time various sensors can read from 64W to 81W as some gets lost to heat each step of the way.
This is why motherboard reviews should use readings from the EPS input to compare their efficiency, while CPU reviews should use their internal power metrics so a motherboard doesn't skew their results. - You cant blame a CPU for high power consumption if the board is at fault.
RAM overclocking drastically increased the power consumption at the EPS socket here, probably due to DDR3's higher voltages. (1.70v for 1866)
CPU Speed | CPU Package power | VR Loop 1 Input | EPS
Stock 3.4GHz: 53W | 54.25W | 60.8W
Stock 3.4 + 1866: 69.52W | 75.5W | 81.4W
4.5GHz: 85.78W / 103W / 109W
5.0GHz: 95W/114W/141W (The efficiency drop is severe here)
5.3: 103/198W/203W
The wattages at the EPS connector blew right out here the cable was definitely warm to the touch as well, since this board only had a four pin EPS connector.
As soon as we passed what the VRM's could easily handle the amount wasted as heat skyrocketed. A better motherboard would not have done this.
"Current out" on the VRM reading was at 65.5A with the 4.5GHz overclock. At 5GHz, this was 80A and absolutely not stable with direct airflow over the VRMs.
Losing 50% of the power as heat is a sign we've well and truly passed the VRM's safe spot, and It simply wasn't stable enough to clock any higher.
VRM temps:
Ambient temperature: 19c
Stock: 55c
CPU OC 4.5GHz: 82c (Light airflow kept VRM's down to 54c)
5.3GHZ: Even with directed air from a 120mm fan, the VRM's went over 80c in seconds.
[\spoiler]
This is the different readings when running the stock CPU + 1866MT/s RAM test, and why it matters to always check where readings are from and compare only with the same reading, for accurate results.
These are all from HWinfo, but different sensors
i5-2500K sensors:
CPU Package Power: 69.52W
IA Cores power: 64.3W
GA-Z77-D3H Sensors:
Power (POUT): 63.5W
Power (Input): 75.5W
Elmor Labs PMD-USB
EPS1: 81.4W
At any given time various sensors can read from 64W to 81W as some gets lost to heat each step of the way.
This is why motherboard reviews should use readings from the EPS input to compare their efficiency, while CPU reviews should use their internal power metrics so a motherboard doesn't skew their results. - You cant blame a CPU for high power consumption if the board is at fault.
RAM overclocking drastically increased the power consumption at the EPS socket here, probably due to DDR3's higher voltages. (1.70v for 1866)
CPU Speed | CPU Package power | VR Loop 1 Input | EPS
Stock 3.4GHz: 53W | 54.25W | 60.8W
Stock 3.4 + 1866: 69.52W | 75.5W | 81.4W
4.5GHz: 85.78W / 103W / 109W
5.0GHz: 95W/114W/141W (The efficiency drop is severe here)
5.3: 103/198W/203W
The wattages at the EPS connector blew right out here the cable was definitely warm to the touch as well, since this board only had a four pin EPS connector.
As soon as we passed what the VRM's could easily handle the amount wasted as heat skyrocketed. A better motherboard would not have done this.
"Current out" on the VRM reading was at 65.5A with the 4.5GHz overclock. At 5GHz, this was 80A and absolutely not stable with direct airflow over the VRMs.
Losing 50% of the power as heat is a sign we've well and truly passed the VRM's safe spot, and It simply wasn't stable enough to clock any higher.
VRM temps:
Ambient temperature: 19c
Stock: 55c
CPU OC 4.5GHz: 82c (Light airflow kept VRM's down to 54c)
5.3GHZ: Even with directed air from a 120mm fan, the VRM's went over 80c in seconds.
[\spoiler]
Tested at 1080p with lowest possible settings, purely to show how the minimum FPS changed the game from being stuttery mess to actually quite playable.
Results are from CapFrameX.
The RAM speed increase alone made a huge difference here, Borderlands 3 is quite sensitive to RAM speed and CPU cache size.
Expressed as Percentages from the FPS low of 25.9, it helps you understand a little easier - the RAM speed increased the lows 40%, while ram and CPU increased them 185.7%
Results are from CapFrameX.
The RAM speed increase alone made a huge difference here, Borderlands 3 is quite sensitive to RAM speed and CPU cache size.
Expressed as Percentages from the FPS low of 25.9, it helps you understand a little easier - the RAM speed increased the lows 40%, while ram and CPU increased them 185.7%
And since someone asked me to test this with a higher end GPU in 3DMark Firestrike, that's been done as well.
Stock+1333 | 5.3+1866
39% faster from the overclocking - a huge boost considering the GPU is entirely at stock here.
Stock+1333 | 5.3+1866
39% faster from the overclocking - a huge boost considering the GPU is entirely at stock here.
This was done to honour a friends accidental final request to beat 5.2GHz on that 2500K and benchmark it against modern hardware, but also a trial run of some testing hardware i've got here and to compare that motherboard and CPU combination to newer ones, TPU style.
How can you know if a modern motherboard is good, without some history from the golden era of overclocking?
Pros:
VRM's rated to 150c
Fantastic thermal pad used on VRMs
RAM ran at 2400MT/s with a 3570K (not shown in review)
Fully unlocked overclocking support with a K series CPU
Extreme overclocking is stable, despite being a budget board.
VRM temperatures perfectly fine with any 1155 CPU at stock speeds
BIOS recovers from errors and failed boots extremely fast
Cons:
Overclocking is locked without a K series CPU
RAM locked to 1600 with a non-K series CPU
VRMs have poor stock cooling - half don't even have a heatsink on them
BIOS has misleading and confusing 'Easy' mode.
PCI-E x4 slot disables all x1 slots
It's 12 years old.
I'm giving this the MSPaint "Hot Shit" brown star award because it's somehow equally shit and awesome.
It does the job fine at stock and could be a boring B or H series chipset board, but somehow it overclocks FAR above its weight limit in a way that puts modern motherboards to shame.
With USB 3.0, SATA 3 and PCI-E 3.0 it can genuinely be used to play any modern game that can handle having just four cores.
[\spoiler]
How can you know if a modern motherboard is good, without some history from the golden era of overclocking?
Pros:
VRM's rated to 150c
Fantastic thermal pad used on VRMs
RAM ran at 2400MT/s with a 3570K (not shown in review)
Fully unlocked overclocking support with a K series CPU
Extreme overclocking is stable, despite being a budget board.
VRM temperatures perfectly fine with any 1155 CPU at stock speeds
BIOS recovers from errors and failed boots extremely fast
Cons:
Overclocking is locked without a K series CPU
RAM locked to 1600 with a non-K series CPU
VRMs have poor stock cooling - half don't even have a heatsink on them
BIOS has misleading and confusing 'Easy' mode.
PCI-E x4 slot disables all x1 slots
It's 12 years old.
I'm giving this the MSPaint "Hot Shit" brown star award because it's somehow equally shit and awesome.
It does the job fine at stock and could be a boring B or H series chipset board, but somehow it overclocks FAR above its weight limit in a way that puts modern motherboards to shame.
With USB 3.0, SATA 3 and PCI-E 3.0 it can genuinely be used to play any modern game that can handle having just four cores.
[\spoiler]
