In my weekly search for AMD Zen news, I came across this article by WCCFTECH about the AMD Zen "Naples" 32 core "beast" of a CPU. They give a Geekbench 3 score and a Geekbench 3 Multicore score, but it doesn't actually seem that impressive. Now, the Geekbench scores were run on a 64bit Linux kernel, but they seem to be, well, sort of miniscule. Looking at 64bit Linux scores on Geekbench 3, there is abolutely nothing impressive about this CPU. I don't really know how to interpret this, maybe I'm missing something. I still have high hopes for Zen, but new I'm a tad more worried that it won't be the breakthrough architecture that AMD needs. Sigh.
1 Like
Yeah, not very good, but the clock speed is a miniscule 1.4 GHz also, if it could pull 2.4 GHz, it would definitely take on most Xeon processors (Except maybe the 22 Core). But I am not counting on that.
1 Like
The numbers do not add up, at all! That is more bulldozer level, not what they showed of against Intel at their event.
Maybe if AMD could send me a Zen chip on fitting mobo to play with....
2 Likes
The clock speeds for all the pre release Zen stuff are very low and are expected to be much higher on release. The other thing about the clock speeds is that this is 32 cores on one chip so heat management is a serious consideration, like the Intel parts as the core count goes up the clocks come down.
Is this the same bench scores from when they threw it against the 5960X and ever slightly beat it. A bit one sided with the Zen being 32 cores against 10+10 more in hyprethread if those count in benches.
A little dissapointing but for consumer side I expect 8 core to be around the sweet spot of performance clock to core and that is what a lot of people want. The massive 16 and 32 core ones are server aimed, I am sure some of the user wan them but they are not the main market for them.
1 Like
The leaked geekbench scores are probably useless. Some bored tech clicked the wrong button, then deleted. Some are probably full on fakes. Geekbench can't really detect stuff properly, clockspeed is notoriously wrong for example. So you can't really draw any conclusions from that data.
Add to this that wccf is a kinda terrible rumor site. They vacuum all the forums like anandtech, beyond3d, semiacc etc etc and then post it as "news". It has been a joke for some time, "don't post that, it will be news on wccf tomorrow". I would ignore that site or at least take what they post with a ton of salt. They want your clicks, nothing more.
1 Like
Pretty sure the geekbench benchmarks are for 16 core, 32 thread. Meanwhile, the demo against Intel was core per core, thread per thread, clock for clock, as a way of demonstrating the IPC gain that they were boasting about.
the secret behind most of amds ipc gain is that zen will be using 16nm chips unlike their latest APUs which are 28+nm. if you look at the scores on passmark. the a12-9800(AM4 socket) is more or less on par with a i5-2500S. it's 7 points faster on multithread but the 2500s has a lead of about 200 points on single thread. the 9800 is still about 100points faster then a 2400S. the 9800 turbo boosts to 4.2ghz while the 2500S is at 3.7ghz and the 2400S is only 3.3ghz.
1 Like
"We Can't Confirm Fuckall" Tech.
8 Likes
Accurate!