Here’s a hypothetical, you have a buddy who is choosing between getting a 7950X or a 9950X for production workloads, and after listening to the reviews you finally want some objective info to help him (or her) make an informed choice.
Don’t worry, I will get to telling you why I think the reviewers are in dire need of a proctologist, because their heads are so far up their own backsides, that they can see daylight, but first the facts.
I did something that nobody else thought of doing (probably because they don’t know how) and that is a direct comparison when I lower the performance of my 9950X to match the maximum performance of my 7950X in CineBench R23.
The way I do the test is to first run CineBench in a standard configuration with no monitor software running, so that I get a clean result. After that, I do another 10 min run and then screenshot the monitoring software (in my case Ryzen Master) when the CPU has reached its maximum temp and power draw.
Of course, I am running on Windows 10 and not WinTel 11.
As far as cooling is concerned, I use an Arctic Liquid Freezer III 360 with three Phanteks T30 fans, and the paste I use is Thermal Grizzly Kryonaut Extreme. My ambient room temp is around 30 degrees Celsius because I have had two spine operations and have spinal arthritis.
My motherboard is just a bog-standard X670 Gigabyte AORUS Elite AX. I don’t need an expensive X670E to get the best out of my CPU.
First off, here is the maximum result I achieved with my 7950X at the maximum safe voltage for TSMC 5 nm which is 1.2 Volts running CineBench R23 for 10 minutes - FYI this result is after having run my 7950X for about 20 months 24/7 only rebooting for driver updates, hardware upgrades or when the OS plaque, as I like to call it, started making itself noticeable (thanks Micro$haft):
The result is not the highest I can get from my 7950X, the thing is though, I benchmark to configure, I don’t configure to benchmark, so this is running my 24/7 safe clocks and voltages.
Here are the stats for that run after my system had reached homeostasis. My system was running at the maximum safe set voltage of 1.2 Volts - what you see under “CPU Telemetry Voltage” is the get voltage (after the voltage droop for all core workloads):
“Not bad”, you might be thinking. My results are a damned sight better than anything that anyone in the Tech Media/YouTube can achieve on a 10-minute run without resorting to exotic cooling (like chilling or LN2).
Now we get to the closest I could achieve to the same CineBench R23 score I got as a maximum for my 7950X result above after a 10-minute run with my 9950X.
Here are the stats for that run after the system had reached homeostasis. My system was running at a voltage of 0.97 Volts Set (and I have been running it 24/7 like this for a couple of days now, chucking various loads at it without any problems).
To get the same result as my 7950X with a 360 rad AIO cooler, I could use a cheap, but good, $20 air cooler with my 9950X and still come out ahead on temp.
I am getting the same result with my 9950X as I did with my 7950X with 65% less CPU power draw.
To put this into perspective, if Intel came out with a flagship CPU that could achieve the same performance as the previous generation flagship with 65% less power, then reviewers would be masturbating themselves into a coma.
“But what about maximum performance”, I hear you ask.
Let’s first turn to the God-King of overclocking, Der8auer.
He got his 9950X, delidded it, applied liquid metal, and then did direct-die cooling and this was his result; his ambient room temp is also about 5 degrees Celsius lower than mine:
His CineBench R23 score was 44599, his power draw was 289.6 Watts and his temp was 94 degrees Celsius.
So what can a no-name rando lowlife such as myself possibly do to compete with the overclocking maestro?
Thank you for asking gentle reader, and I would answer with this:
And here are the stats:
Of course, I did none of those fancy, delidding, liquid metal, direct die, things like Der8auer. All I did was slap my CPU into the socket and whacked on my cooler, like a pleb, (well that, and of course knowing how to configure Ryzen didn’t hurt).
That being said, compared to Der8auer, my score is over 3.3% higher, my power draw is 28.3% lower, and also my temperature is 13% lower than his.
I did find that the latest beta BIOS I was using (AGESA 1.2.0.0a Patch A) still has some ragged edges. Nothing that affected stability per se, but certainly something that will show increased performance in the future.
Sadly, the reviews of the Tech Media/YouTubers of the 9000 Series of AMD CPUs has continued the devolution to the level of glorified infomercials shown in the past generations of Ryzen CPUs.
Let’s compare and contrast shall we?
These self same, “We only test out of the box performance” liars couldn’t break out XTU fast enough when Intel brought out their 12th Gen Alder Lake CPUs, which needed a new OS, just so that they would function.
Of course, these stalwarts of integrity - for the sake of “Fairness” - shifted their reviews of both Intel and Ryzen CPUs over to WinTel 11 which they knew nerfed the AMD CPUs compared to Intel.
So are these reviewers - and I include all the big names in this - too lazy, too stupid, or just plain too corrupt for Intel marketing dollars, to have learned to properly configure Ryzen almost five years after I wrote my first guide on how to get the most out of AMD CPUs?
They have the Intel blinkers on when it comes to Ryzen and their only bloody efforts to do any kind of configuration consists exclusively of “Moar Powa, Moar Gud”.
To end this, I would turn back to the original question I posed, which was, given what I have shown you, would you advise your buddy to buy a 7950X or a 9950X if he (or she) came to you for advice?
