The Ryzen 9 9950X, a new flagship 16C/32T desktop processor from AMD that promises to deliver another leap in performance with its brand-new Zen 5 cores, is up to 45% faster than the Ryzen 9 7950X, the previous generation’s non-X3D leader, according to new benchmarks that have surfaced at an online forum. AnandTech forum member igor_kavinski, who shared the numbers, clarified that this is an early engineering sample, one that was running on an AM5 motherboard with DDR5-8000 (34-45-40-42) memory.
Percentage improvements, per igor_kavinski:
| ​ | 7950X​ | 13900K​ | 7975WX​ |
| AES​ | 45+​ | 55+​ | 11+​ |
| FP32​ | 39+​ | 60+​ | -13​ |
| FP64​ | 39+​ | 60+​ | -16​ |
Some screenshots alluding to the chip’s performance (AES, FP32, FP64):
A comparison between the new, and older, flagships derived from the numbers above:
From a report:
While the benchmark itself showcased almost 2x performance gains in these aspects, the actual performance of the chip may not showcase similar gains since applications and workloads aren’t based entirely on such instructions and use multiple different parts of the chip. These are also the floating point figures so we can see a different story in the integer performance tests but one thing is for sure, Zen 5 does offer a significant gain through its brand new architecture…
Join the discussion in The FPS Review Forums...
Discussion (9 replies)
Join Discussion →But will it make Civ V any faster? /s
"Burticus, post: 86622, member: 297" wrote:But will it make Civ V any faster? /s
Can't remember if Civ V took advantage of the X3D cache or not... but if not, then probably?
Waiting for more balanced testing.
Only place in saw those numbers reported was wccf tech. Gives me hope though for reviews to come. I know my 5900x is holding back my video card.
"Elf_Boy, post: 86630, member: 438" wrote:Waiting for more balanced testing.
Yep. That's the way to go. We don';t need another Bulldozer hype machine.
I will reserve any and all judgment until reputable reviewers have actual silicon in hand.
Speculating, we should expect higher IPC - that's really the only way for AMD to go, and they're behind Intel's P-cores at the moment. Cache and interconnect optimizations should also be present, as well as memory controller advancements (already seen from AMD with their APUs being able to push 8000MT/s+).
Granted, I've seen it reported that Zen 5 without 3D V-cache will not be faster than Zen 4 with 3D V-cache for gaming, per AMD, so gains will probably be limited to compute-heavy scenarios.
"LazyGamer, post: 86700, member: 1367" wrote:Granted, I've seen it reported that Zen 5 without 3D V-cache will not be faster than Zen 4 with 3D V-cache for gaming, per AMD, so gains will probably be limited to compute-heavy scenarios.
"Zarathustra, post: 86704, member: 203" wrote:
Sounds like a terrible idea...
"LazyGamer, post: 86714, member: 1367" wrote:Sounds like a terrible idea...
I mean, they already have SKU's with higher TDP than that, so why not. It's not like it will be constantly at the high power state, so it's fine. It will hum along on the desktop at just as low power as a 65W part.
Discussion (9 replies)
Join Discussion →But will it make Civ V any faster? /s
Can't remember if Civ V took advantage of the X3D cache or not... but if not, then probably?
Waiting for more balanced testing.
Only place in saw those numbers reported was wccf tech. Gives me hope though for reviews to come. I know my 5900x is holding back my video card.
Yep. That's the way to go. We don';t need another Bulldozer hype machine.
I will reserve any and all judgment until reputable reviewers have actual silicon in hand.
Speculating, we should expect higher IPC - that's really the only way for AMD to go, and they're behind Intel's P-cores at the moment. Cache and interconnect optimizations should also be present, as well as memory controller advancements (already seen from AMD with their APUs being able to push 8000MT/s+).
Granted, I've seen it reported that Zen 5 without 3D V-cache will not be faster than Zen 4 with 3D V-cache for gaming, per AMD, so gains will probably be limited to compute-heavy scenarios.
That might be changing.
Sounds like a terrible idea...
I mean, they already have SKU's with higher TDP than that, so why not. It's not like it will be constantly at the high power state, so it's fine. It will hum along on the desktop at just as low power as a 65W part.