AMD Opteron Vs EPYC: How AMD Server Performance Evolved Over 10 Years (phoronix.com)

← Back to Stories (view on slashdot.org)

AMD Opteron Vs EPYC: How AMD Server Performance Evolved Over 10 Years (phoronix.com)

Posted by BeauHD on Wednesday September 20, 2017 @10:00AM from the benchmarks-and-performance-per-watt dept.

New submitter fstack writes: Phoronix has carried out tests comparing AMD's high-end EPYC 7601 CPU to AMD Opteron CPUs from about ten years ago, looking at the EPYC/Opteron Linux performance and power efficiency. Both on the raw performance and performance-per-Watt, the numbers are quite staggering though the single-threaded performance hasn't evolved quite as much. The EPYC 7601 is a $4,200 USD processor with 32 cores / 64 threads. The first of many tests was with NAS Parallel Benchmarks: "For a heavily threaded test like this, going from a single Opteron 2300 series to the EPYC 7601 yielded around a 40x increase in performance," reports Phoronix. "Not bad when also considering it was only a 16x increase in the thread count (4 physical cores to 32 cores / 64 threads). The EPYC 7601 has a lower base clock frequency than the Opteron 2300 CPUs tested but has a turbo/boost frequency higher, among many architectural advantages over these K10 Opterons. With the NASA test's Lower-Upper Gauss-Seidel solver, going from the dual Opteron 2384 processors to a single EPYC 7601 yields around a 25x improvement in performance over the past decade of AMD server CPUs. Or in looking at the performance-per-Watt with the LU.C test, it's also around a 25x improvement over these older Opterons."

34 comments

Min score:

Reason:

Sort:

Not that impressive by speedplane · 2017-09-20 10:15 · Score: 1

The article notes speedups ranging from about 3X to 40X depending on the test... while that initially sounds like a lot, but that's only after 10 years of development. If performance doubled every 18 months, the speedup should be 680X.

--
Fast Federal Court and I.T.C. updates
1. Re:Not that impressive by ERJ · 2017-09-20 10:21 · Score: 4, Informative
  
  I'll be that guy...
  
  Technically Moore's Law (I assume that is what you are referencing) says the number of transistors per square inch of wafer will double every 18 month, not that performance will double.
2. Re:Not that impressive by night · 2017-09-20 10:24 · Score: 3, Insightful
  
  Moores "law" was doubling of transistors, not performance. But it's not a law and has already failed in both senses of the word. Intel just delayed their 10nm CPU again. AMD announced their next generation shrink will be 12nm, given that the current CPUs are 14nm that's going to be a rather slow increase. Generation to generation improvements are on the order of 10-33% on most metrics and take quite a bit longer than 18 months to come out.
3. Re:Not that impressive by Anonymous Coward · 2017-09-20 10:33 · Score: 1
  
  Unless I'm mistaken 10 years = 120 months which is 6.66 doubling iterations so somewhere between 32 and 64x, even if we ignore the fact that Moore's "Law" refers to transistor counts rather than performance.
4. Re:Not that impressive by Anonymous Coward · 2017-09-20 10:48 · Score: 0
  
  Hmm, 2^6=64, so you are off by a factor of 2: it should be between 64 and 128.
  But Moore's law characteristic period of 18 months was only really valid when the law was stated, it has now increased to well above 2 years and closer to 3.
  Yes, we are hitting the wall in transistor density for logic chips (for flash, 3d works and 32 to 64 layers seems to be quite normal these days).
5. Re:Not that impressive by barc0001 · 2017-09-20 10:55 · Score: 4, Informative
  
  It's been dead for some time. My home rig is an i5 2500K that I bought in the spring of 2011. It's only this year that I've found a decent midrange "doubling" candidate for building a new machine around, interestingly enough, the Ryzen 5 1600. Benchmarks suggest that I'll get about double the performance out of that, and it's in the same price bracket as when I bought the 2500K. 6.5 years and only 2x the oomph on the desktop in the midrange price bracket. I used to see that sort of improvement (and upgrade accordingly) every 2 years, but no longer.
6. Re:Not that impressive by night · 2017-09-20 11:17 · Score: 4, Insightful
  
  Agreed, which if you think about it justifies more expensive desktops on a longer replacement cycle. Go ahead and spend the extra few hundred on ram, cpu, motherboard and case. The plan on keeping it for 7-10 years. The next doubling is going to take even longer than the last one.
7. Re: Not that impressive by alvinrod · 2017-09-20 11:33 · Score: 2
  
  It's still pretty impressive especially if you account for most of the growth being towards more cores. That's going to have higher diminishing returns than single core improvements.
  
  Anything else improving 3x in ten years would be pretty much miraculous. Imagine if cars now got that much improvement in fuel efficiency. We're so spoiled it's not even funny.
8. Re:Not that impressive by K.+S.+Kyosuke · 2017-09-20 12:03 · Score: 2
  
  I'll be that guy...
  Technically Moore's Law (I assume that is what you are referencing) says the IC complexity for minimum component costs will double every two years, not that the number of transistors per square inch of wafer will double.
  
  --
  Ezekiel 23:20
9. Re: Not that impressive by K.+S.+Kyosuke · 2017-09-20 12:20 · Score: 1
  
  Shouldn't you ask some sucking nerds about that instead?
  
  --
  Ezekiel 23:20
10. Re:Not that impressive by Kjella · 2017-09-20 12:59 · Score: 5, Interesting
  
  It's been dead for some time. (...) 6.5 years and only 2x the oomph on the desktop in the midrange price bracket. I used to see that sort of improvement (and upgrade accordingly) every 2 years, but no longer.
  It's getting very near the end, 10nm is already shipping but not on desktop chips, 7nm gets exotic with EUV but is probably doable but 5nm is a "maybe, if we get all the crazy quantum effects worked out". Even if they pull another rabbit out of the hat the silicon lattice constant is 0.543 nm which is a lot more fundamental problem than all the other issues they've found workarounds for. You're literally down to counting atoms, my guess is that by 2025 they've reached the end of the line. Not just a speed bump but like permanently. At least for anything remotely resembling the processors we have today.
  
  --
  Live today, because you never know what tomorrow brings
11. Re: Not that impressive by Anonymous Coward · 2017-09-20 15:39 · Score: 0
  
  Technically, yes.
12. Re:Not that impressive by Anonymous Coward · 2017-09-20 16:54 · Score: 0
  
  It's getting very near the end, 10nm is already shipping but not on desktop chips, 7nm gets exotic with EUV but is probably doable but 5nm is a "maybe"
  If you add a zero to all those number I seem to recall them saying the almost exact same thing about 15 years ago.
13. Re:Not that impressive by Anonymous Coward · 2017-09-20 19:32 · Score: 0
  
  "IC complexity", in that case, referring to the number of components on a single quarter-inch semiconductor. So yes, transistors per square inch.
14. Re:Not that impressive by Anonymous Coward · 2017-09-20 20:00 · Score: 0
  
  What do you mean, "probably doable"? The big 4 are already all running 7nm test lines.
  Semiconductors have had "crazy quantum effects" to "work out" for decades. A 5nm transistor has been demonstrated 15 years ago. IBM/GF has said they've created 5nm chips recently.
  You might be right about only one more shrink left after 5, but alarmist about 7 and 5.
15. Re:Not that impressive by K.+S.+Kyosuke · 2017-09-20 22:19 · Score: 1
  
  No, number of components in a single package. The size of dies within those packages has increased over time.
  
  --
  Ezekiel 23:20
16. Re:Not that impressive by Anonymous Coward · 2017-09-21 00:41 · Score: 0
  
  18 months is 3/2 years. There would be 10 / (3/2) = 20/3 doublings in 10 years.
  2^(20/3) ~ 101
17. Re:Not that impressive by speedplane · 2017-10-03 20:51 · Score: 1
  
  I'll be that guy... Technically Moore's Law (I assume that is what you are referencing) says the number of transistors per square inch of wafer will double every 18 month, not that performance will double.
  Ha, yes I saw that coming. I know what Moore's law says, which is why I didn't directly reference it. That said, the industry's success has been rooted in it's historical exponential performance gains. As a practical matter, no one cares how many transistors are on a trip, but what that chip can do.
  
  --
  Fast Federal Court and I.T.C. updates
18. Re:Not that impressive by speedplane · 2017-10-03 20:55 · Score: 1
  
  It's been dead for some time.
  We're still too close to it to tell when Moore's law really ended. My guess is that historians will later peg somewhere within 2010-2015.
  
  --
  Fast Federal Court and I.T.C. updates
19. Re: Not that impressive by speedplane · 2017-10-03 20:57 · Score: 1
  
  Anything else improving 3x in ten years would be pretty much miraculous. Imagine if cars now got that much improvement in fuel efficiency. We're so spoiled it's not even funny.
  This is a good point, but the entire computing industry has been predicated on exponential growth. It'll take centuries to get real AI with linear growth of the underlying hardware.
  
  --
  Fast Federal Court and I.T.C. updates
WOW! by Anonymous Coward · 2017-09-20 10:18 · Score: 1

I want it to have my babies.
Wasted comparison by Anonymous Coward · 2017-09-20 10:23 · Score: 0

I mean, compare it to Intels latest or even Power9 series.
Even comparing it to ARM designs is more useful.
1. Re:Wasted comparison by Anonymous Coward · 2017-09-20 10:50 · Score: 0
  
  Are Power9 systems shipping now?
  Don't get me wrong, I'd love to run some scientific code on them, but I've not seen any yet.
2. Re:Wasted comparison by Anonymous Coward · 2017-09-20 21:50 · Score: 0
  
  Still in "pre-order acceptance", so they'll take your money. Will they run, or will it get you a machine to run?
  The problem with AMD is the same as with intel, these days: Can't trust the platform because of the extra "security" and "management" crud they snuck in. You'll never reallly own the thing since they retain the keys. That makes the whole thing nigh useless, whatever the speed gains.
  Wish they'd get on with the POWER9 already. Maybe we can poke fujitsu to do some sparc-y goodness. Or if someone has a billion or so pocket money to spend, resurrecting alpha on newer process might be interesting. Shouldn't be too hard to stick a sackful of'em on a single die, maybe with a nice interconnect like a siliconized crossbar switch like the starfire's, or something. So many possibilities, so much effort and money wasted on mediocre crap.
In case you missed it... by Gravis+Zero · 2017-09-20 11:01 · Score: 3, Interesting

They did a comparison between the highest end Intel chips and the EPYC 7601. Not to spoil it but EPYC blew the panties straight off of Intel's chips while using less power. It's no wonder Intel has been flailing in the media.

--
Anons need not reply. Questions end with a question mark.
1. Re:In case you missed it... by Anonymous Coward · 2017-09-20 11:16 · Score: 0
  
  ya top end offering vs midrange offering, i wonder why one would win over the other
2. Re:In case you missed it... by drinkypoo · 2017-09-20 12:05 · Score: 2
  
  ya top end offering vs midrange offering, i wonder why one would win over the other
  The only comparison that matters is TCO/performance. Intel's current chips are just as power-hungry as AMD's, so the ratio is going to be more favorable on the AMD system.
  
  --
  "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
3. Re:In case you missed it... by K.+S.+Kyosuke · 2017-09-20 12:12 · Score: 5, Insightful
  
  You wonder why the more expensive Xeon was slower than the cheaper Epyc? Probably because Intel has been milking you...
  
  --
  Ezekiel 23:20
4. Re:In case you missed it... by Anonymous Coward · 2017-09-21 05:27 · Score: 0
  
  msrp for price/performance comparisons are pretty silly and seem to be the only thing amd apologists have going for them.
  If you pay more then 50% of msrp for your intel systems, you probably need to shop for a new vendor, as your current vendor is milking you
CHART JUNK ABOUNDS! by Anonymous Coward · 2017-09-20 11:27 · Score: 0

Junk out makes it all just a fool's waste of time!
Upgrade my potato by PopeRatzo · 2017-09-20 11:39 · Score: 1

So, will this shit run the new Wolfenstein II: The New Colossus? My old Gen 1 i5 doesn't have enough cores or threads or gigabytes or some bullshit.

--
You are welcome on my lawn.
Re: AMD astroturfing? by Anonymous Coward · 2017-09-20 19:02 · Score: 0

Why would they? There is plenty of positive news for AMD these days, so the stories will appear anyway.