Hacker News Reader

Comment by mrb | Hacker News Reader

This was expected because the 3900X is optimized for fewer faster cores (12 at 3.8GHz) while the 9980XE is optimized for more slower cores (18 at 3.0GHz.)

A better and far more impressive comparison would be the 3950X vs 9980XE which have comparable number of cores (16 vs 18). And in that instance AMD still beats Intel in both single-threaded and multi-threaded workloads! And by wide margins! And with AMD being 62% cheaper ($750 vs $1950)! And with AMD's TDP 36% lower (105 vs 165W)! And despite having 2 fewer cores! And despite the 3950X engineering sample below being clocked under final specs (3.29GHz vs 3.5GHz)! See this leaked 3950X engineering sample codenamed "Myrtle": https://browser.geekbench.com/v4/cpu/compare/13612629?baseli...

Intel has never been so thoroughly defeated since the 1999 K7. Interesting times...

View on HN

Also doesn't Intel have a lot more known security flaws?

One more, assuming you mean meltdown. But that's the easy one to fix.

Not true. Copy-pasting from my Ryzen terminal:

   $ spectre-meltdown-checker | grep "NOT VULNERABLE" | wc -l  
   12

So? The same script reports 11 lines on my 4 year old Haswell. It should read as "mitigated or not vulnerable" instead of simply "not vulnerable".

Edit: oh wait, it's actually 12 lines, the same as on your system. The reason it reported 11 on the first run is because KPTI has been disabled manually here.

:) I grepped for the wrong thing. Here's a better version:

   $ sudo spectre-meltdown-checker | grep "Not affected" | wc -l


   On Skylake: 0
   On Ryzen  : 10

I don't have an Ivy Bridge CPU to test this on, but AMD has suffered from _significantly_ fewer number of vulnerabilities than Intel, not just "one more" as the top commenter suggested.

What about the recent one ZombieLoad?

You're forgetting Intel's trouncing in the Pentium4/Athlon XP days also. It's been a while, but I for one am looking forward to a very competitive marketplace for desktop CPUs.

He was referring to the K7 architecture which includes the Athlon XP brand.

This is what happens when the 3900/3950X are on TSMC's 7nm FINFET while Intel's still on 14nm base Lithography

Does the old “there are more software written with Intel in mind than AMD” still hold? Just like Nvida.

A critical difference is that GPU architectures differ far more, AND different GPUs (well, their driver's) effectively use different compilers with different optimizations, etc. (Yes, the input is an IL with some optimizations already done, but ultimately codegen is done by the driver.)

You can micro-optimize for a particular CPU- say, not using AVX-512 on Zen, or hand-placing instructions so the latencies all line up juuust right- but it's not nearly as important as for a GPU.

Plus, GPU vendors will work with software devs to optimize the driver just for that software (mostly via a collection of weird workarounds, hacks, feature flags, etc. in the driver, automatically applied when it detects that software running); often the software dev will take, essentially, a bribe from a GPU manufacturer to only work with them.

Then you have stuff like PhysX, CuDNN, etc...

Reply on news.ycombinator.com