Great review!
We do have Core Isolation/Memory Integrity DISABLED on each platform and CPU
Could you delve a little further on this? I find myself not really familiar with what this is, and why one would disable it.
It sounds like it might be an important security feature?
Also....
*Note* – The FCLK (Infinity Fabric) was set to motherboard defaults with the “AUTO” setting in the BIOS after loading default settings. We confirmed that the FCLK frequency was running at 2100MHz for all of our testing. It was the same for each CPU. We opted to keep it on the default motherboard setting of “AUTO” to represent the ‘out-of-box’ experience with this motherboard/CPU/RAM combination at defaults. We always use default BIOS settings for testing, on all CPUs, with all motherboards, to keep it fair. All AMD CPUs were tested on this same motherboard and RAM kit.
Therefore, we are running AUTO:1:1 – which in effect is 2100:3000:3000 (FCLK/UCLK/MCLK) using EXPO I profile on this RAM, for 6000MT/s DDR5.
Memory appears to behave rather interestingly with this release.
If you guys get around to it at some point I'd love to see your breakdown on memory settings and performance.
The settings you used seem like they would be the best easily achievable RAM settings for this platform, and a good choice for the initial review, but I am curious about other settings as well.
How do they - for instance - compare to DDR5-6400 with fabric at 2133 in a 1:1:1 configuration? Is this stable?
And how does the performance of this compare to - say - DDR5-8000 in a non 1:1:1 configuration? Is it better to go for that perfect 1:1:1 or to shoot for higher clock speed?
Also, I understand CL is less important than it once was, but how important are timings?
I think this could make for an interesting follow-up test.
Wendell over at level1techs also has some really interesting remarks in his Zen5 reviews (both on the main channel and the level1techs Linux channel) about how he thinks the current Windows 11 scheduler might be wonky with Zen5 and is holing it back. Unlike with other CPU's he is seeing much better performance under Linux with this architecture. I'll be looking forward to further investigation into this.
Also,
One more thought,
Your review does not have this data, but in other reviews I've seen, the 9950x - while it does well in average frame rate, it has been
very impressive in 1% and 0.1% lows compared to other CPU's even beating the 14900k in many benchmarks.
I find this to be very interesting and I can't help but wonder why. What about this CPU specifically makes it so good at minimum frame rates?
I suspect some of it may have to do with Intel's E-cores getting in the way, and Intel's much higher memory bandwidth allowing the CPU to run away with it in some cases, pulling up the average, but keeping the minimums where they are. Not sure though
I've been a long time believer that 1% and 0.1% lows are of much greater importance than average frame-rates. That's what I determine my "playability" on. I want minimums/lows to be 60fps or greater. IN reality this tends to result in averages at or about 90fps, but some CPU/GPU's have a smaller difference between averages and lows than others.
