As we've been covering, AMD's Ryzen 3000 series CPUs often have issues hitting and sustaining the advertised boost clock speed. A recent motherboard firmware update improved things a little bit, but often boosts the wrong core. Results vary based on chip quality, motherboard, and firmware, so I set out to see if adding more cooling would help get my store-bought Ryzen 7 3700X to hit or exceed its advertised top-speed of 4.4 GHz. What I found out is that, even when I used liquid nitrogen to freeze my processor down to -180 degrees Celsius, it was still stuck at 4.35 GHz, 50 MHz below its boost. However, as we'll see below, you can hit boost clocks by manual overclocking.
Testing AMD's Latest Update
My favorite overclocking YouTuber and close friend Roman “Der8auer” Hartung has really exposed and led the drive to get a solution to the problem of the missing MHz from AMD’s new processors. In fact, so much so that AMD has already responded with a possible solution. Usually this means a new AGESA (AMD Generic Encapsulated System Architecture, foundational code for the BIOS) version then handing it off to motherboard companies which do some validations and then pass it onto us, the consumers. Enter AGESA 1.0.0.3 ABBA. I promise not to make any ABBA puns, OK Chiquitita?
Let’s take it for a spin! We’ll use the ASRock X570 Taichi because it's overbuilt with strong PWM and it doesn’t cost much money, money, money, but still has plenty of RGB to satisfy the wildest of Dancing Queens. The Enermax Maxtytant 1250W PSU, which has 104A on the single 12V rail, should be proper overkill. I would have to say we have enough firepower.
On the cooling front, the Enermax Liqtech II 360, which is designed for an astonishingly-high 500W TDP and Intel HEDT processors but fits nicely on AM4, will keep our processor's temperature very nice and low.
CPU | AMD 3700X Processor |
Motherboard | ASRock X570 Taichi Bios 2.1 AGESA 1.0.0.3. ABBA |
Graphics | ASRock 560x GPU |
RAM | G.Skill SPD 4133 B-Die Memory |
Power Supply | Enermax MaxTytan 1250W PSU |
CPU Cooler | Enermax Liqtech II 360 White |
Thermal Paste | Thermal Grizzly Kryonaut LHe edition |
Storage | Team Group L5 3D SSD 256GB |
The latest BIOS available to us on the ASRock X570 Taichi is 2.1 with a release date of 09/09/2019. We’ll keep the memory at bare DDR4-2133 stock settings and NOT messing with the OS at all (limiting memory, removing services, etc.) to best emulate a normal use case as a baseline.
Enter the test layout. Testing with the Cinebench R15 single-core benchmark while logging with the HWiNFO64 utility is the standard requested by AMD. As we can see below; mamma mia, has nothing has been fixed at all? 4.35 GHz is the highest that any core ran during the single thread R15 run. Temps are no issues for the Liqtech II, as you can see it peaks at 58.4C even when the CPU is reaching well over a 1.4VCore.
What now? With easy access to liquid nitrogen I wondered if we could create a scenario where the boost does work correctly. Is this just a temperature issue? Is this is a power issue? Some people (even if only 5% in some cases) can hit the advertised boost. Let’s find a way to make it stick.
I strapped the 8ECC Beast LN2 Pot on and some Thermal Grizzly LHe thermal paste because we are going sub-zero. Starting off, we kept the pot at -40c and (huh?) there is no change at all in the max single-core frequency during R15. -80C, surely this will cause some chaos! Nothing. On and On and On till we hit -180c. Again skunked, the max clocks are the same damn 4.35 GHz!
When you enter your BIOS, all you need to do is set fixed overclocking mode. Set your Vcore and your CPU frequency, and that's it! Your results may vary slightly, but this is just an average retail Ryzen 7 3700X purchased at a big box store.
Let’s check Cinebench R15 performance again as it seems to be a popular metric for this testing and AMD likes to use it to show its threaded dominance over Intel CPUs.
8 Core 16 Threads
Frequency | Voltage | R15 Score | Max Temperature Tdie |
4200 locked | 1.25v | 2164 | 63.3c |
4300 locked | 1.25v | 2214 | 64.6c |
4400 locked | 1.25v | 2274 | 67.8c |
Stock Boost | 1.49v | 2066 | 73.3c (4175-4225mhz moving) |
Single Core
Frequency | Voltage | R15 Score | Max Temperature Tdie |
4200 Locked | 1.25v | 202 | 42.2c |
4300 Locked | 1.25v | 204 | 45.3c |
4400 Locked | 1.25v | 207 | 47.8c |
Stock Boost | 1.49v | 203 | 58.3c (4300-4350 MHz moving) |
There are some great things to notice here. First, the temperatures. Stock boost is really slamming those volts, so heat is climbing quite a bit. 7nm at 1.49V seems a lot to me, and it’s totally not necessary. There also seems to be some CPU cycles used to switch cores because with the same clocks using stock boost the chip is vastly less efficient than if you set it to a fixed clock. We gained a 10% increase in the score from 4.4 GHz fixed vs the stock boost while running 5C less and using 0.24V less to the core.The Wattage read on the MaxTytan shows a drop from 162W peak at stock to 144W using a fixed frequency during the R15 threaded load.
In short, fixed frequency IS the fix for now, and maybe forever. You aren't saving power, and you aren't saving heat or longevity by using the stock boost settings with its high voltage and overall lower clocks.
Are these growing pains? Of course! To be fair the ASRock is quite aggressive with its stock settings, but for AMD to market this as a 3.6 GHz CPU and then you pop it in your motherboard and get 4.2 GHz without making any changes you can hardly be disappointed. Tack on an extra 200 MHz with less time in the BIOS than it takes to play an ABBA song, and YOU are the fix!
Image Credits: Tom's Hardware