AMD Ryzen 5 7600 / Ryzen 7 7700 / Ryzen 9 7900 Linux Performance

[phoronix]

Phoronix: AMD Ryzen 5 7600 / Ryzen 7 7700 / Ryzen 9 7900 Linux Performance

Last week at the AMD CES 2023 keynote hosted by Lisa Su, new 65 Watt Ryzen 7000 series processors were announced. These more affordable Zen 4 processors are going retail this week and today marks the embargo lift. Up on the Linux testing block are the Ryzen 5 7600, Ryzen 7 7700, and Ryzen 9 7900 processors.

https://www.phoronix.com/review/ryzen-7600-7700-7900-linux

[Anux]

Yeah, my 5700G still leads in efficency and price/perf.
Are there any announcements regarding 7000X3D release dates?

[avis]

Goes to show how egregiously needlessly OC'ed all the X parts are.

Pricing is fine, would be nice to upgrade once motherboards and DDR5 RAM become cheaper. I got my 2x16GB of DDR4 3600MHz CL17 RAM for $125. There's nothing even close to this pricing for comparable DDR5 6000 CL32 (I want 2x32GB since DDR5 works horribly slowly if you populate all the four slots).

RPL is left in the dust:

Yeah, my 5700G still leads in efficency and price/perf.
Are there any announcements regarding 7000X3D release dates?

They should become available before February 28.

[schmidtbag]

Hmm, this makes the 7700 a bit more enticing than I thought. Hardly any slower than the 7700X but significantly lower power draw. I've been considering the 13500 too but seeing as the 13600K is overall slower, costs roughly the same, and is a lot more power hungry, I'm beginning to think I might just commit to the 7700. I'd still prefer to upgrade my GPU first, though.

[DoofusOfDeath]

Michael Thanks for doing these benchmarks, I find them super useful.

For fixed-size workloads like code-compilation, have you considered generating 2D scatter plots like the following?

• X axis is total watts consumed (either by the system, or just the CPU)
• Y axis is time needed to complete the workload
• each data point is a different CPU

No idea how many of your readers would be interested, but I like the idea of clearly seeing the tradeoff between electricity-bill vs. productivity.

[Anux]

Hmm, this makes the 7700 a bit more enticing than I thought. Hardly any slower than the 7700X but significantly lower power draw. I've been considering the 13500 too but seeing as the 13600K is overall slower, costs roughly the same, and is a lot more power hungry, I'm beginning to think I might just commit to the 7700. I'd still prefer to upgrade my GPU first, though.

The only dick move here is the "no overclocking" (probably because that would make their X CPUs worthless). I would wait for the 8 core 3D cache variants, unless your workloads do not profit from large caches.

[Michael]

Michael Thanks for doing these benchmarks, I find them super useful.

For fixed-size workloads like code-compilation, have you considered generating 2D scatter plots like the following?

• X axis is total watts consumed (either by the system, or just the CPU)
• Y axis is time needed to complete the workload
• each data point is a different CPU

No idea how many of your readers would be interested, but I like the idea of clearly seeing the tradeoff between electricity-bill vs. productivity.

All my graphs are automated so would need to think of a way for determining the fixed-sized workloads effectively, etc. Would likely be a lot of work without not knowing how useful it would be to readers.

[r1348]

The only dick move here is the "no overclocking" (probably because that would make their X CPUs worthless). I would wait for the 8 core 3D cache variants, unless your workloads do not profit from large caches.

The non-X CPUs can be overclocked too, there's nothing artifically blocking that. Of course, you shouldn't expect the same results of the X CPUs.

[Anux]

All my graphs are automated so would need to think of a way for determining the fixed-sized workloads effectively, etc. Would likely be a lot of work without not knowing how useful it would be to readers.

Me would like that too. Maybe it's easier if you do it only for the last 2 tables (average power over all tests and geomean)? That would be less accurate but probably less work?

[CHECK_12345]

Wow the Ryzen 9 7900 non-x seems to be pretty good performance per-watt at pretty good price compared to x-series. While seeing that It could be $30USD more than Ryzen 7 7700x when I search it up in internet for the prices. I am guessing that the Ryzen 9 7950x could be the most power efficient because is has more cores/threads. It was G-Series there could be even more power efficiency because of G-Series being Monolithic instead of Chipset.

More cores + Less Core Clock = Good Performance Per-watt

I am using Ryzen 7 1700 with 2.5gHz set at 0.768V and also set 1.55gHz set at 0.612V.