AMD EPYC 7502 + EPYC 7742 Linux Performance Benchmarks

[oleyska]

Exactly. The difference was even bigger back then, with the Athlon 64 performing about 60% faster than a similarly clocked Pentium 4.

I kind of beg the differ.
Yes amd's IPC might have been there, and for consumers yes for the normal stuff it was a lot in amd's favor.
however Intel always had the process advantage and just brute forced the performance crown with hilarious cache's and in server their performance didn't really differ that greatly but amd was still better but not by these margins by a long shot.

[hoohoo]

It is the end of the world as we know it.

[wizard69]

With all of the AMD product releases in the last couple of months I do feel that you do deserve a nice holiday!

I know I support this idea. Even if it is only a week! Running Phoronix 24/7 has to wear on you after awhile.

[Michael]

I know I support this idea. Even if it is only a week! Running Phoronix 24/7 has to wear on you after awhile.

It wears me down a lot, but good Bavarian beer provides great recovery :P

[wizard69]

Your post has me wondering which 7nm process at TSMC they have built this series on. TSMC has at least 2 “7nm” processes that I know of. If AMD wanted to they may be able to jump to the newest version and get a minor boost in 10 months. They could easily call it Zen 2+ and give themselves some padding to put more effort into Zen3.

So except for certain lightly-threaded and database workloads, this looks like a blowout. AMD looks to have the space to itself, at least until Ice Lake-SP in 10 months. If I were AMD, I would wait like 7 months, and then start trickling out details about Zen 3 in order to get people to wait for that release instead of buying ICL-SP.

If ICL-SP is actually going to be competitive with Rome, we should expect to start seeing benchmark "leaks" soon. Silence is a red flag here.


According to AnandTech's review, this is up to the motherboard manufacturer. By default it still uses half but you could make a motherboard that allocates fewer lanes than that to IF.


Maybe the boost algorithm doesn't idle the CPU until after a certain amount of time. If it really can't idle, then that would be a serious problem. It would definitely be worth measuring idle power consumption against Xeon, as that is something a server customer would care a lot about.

[wizard69]

It wears me down a lot, but good Bavarian beer provides great recovery :P

See I knew you needed a vacation! October fest season is coming up, we promise not to mis you for one whole week!

By the way keep up the good work, this AMD release appears to be the most interesting in at least a decade. Now all we need to see is a recovery on the GPU side.

[andreano]

The future of Intel-sponsored software projects like SVT-AV1 looks a little bleak now that AMD beats them in their own game.

[edwaleni]

Your post has me wondering which 7nm process at TSMC they have built this series on. TSMC has at least 2 “7nm” processes that I know of. If AMD wanted to they may be able to jump to the newest version and get a minor boost in 10 months. They could easily call it Zen 2+ and give themselves some padding to put more effort into Zen3.

The chip is manufactured on TSMC's 7nm HK-MG FinFET process using SAQP. The over die is 0.34x the size of their 16 nm process version. TSMC's 7nm process density is 1.6X compared to their 10nm process. Minimum metal pitch is 40 nm, as reported at IEDM 2016. TSMC claims their 7nm process will deliver a 20% performance improvement and a 40% reduction in power consumption.

The 7nm node will come in two variants, one optimized for mobile applications and a second one optimized for High performance applications. TSMC plans to introduce a second improved process called 7nm+ a year later, which will introduce some layers processed with EUVL. This will improve yields and reduce fab cycle times. The 7nm+ process will deliver improved power consumption and between 15-20% area scaling over their first generation 7nm process.

https://en.wikichip.org/wiki/7_nm_lithography_process

[Adarion]

It doesn't fit on the screen, even on 32c systems I don't believe it ends up looking anything special as it gets chopped off.

Awww! But thanks!

[bigless]

Maybe dumb question, but did you benchmarked MariaDB configured for NUMA? I quess its off by default.