AMD Epyc 7763 CPUs break Cinebench world record, crushes Intel Xeon in performance
Click here to post a comment for AMD Epyc 7763 CPUs break Cinebench world record, crushes Intel Xeon in performance on our message forum
kapu
Intel counterpart looks like decade old tech looking at scores and TDP.
King Mustard
They're cheaper and faster but cost a little more to run, looking at the higher TDP.
I imagine that matters to datacenters.
Still - overall better from the looks of things
RED.Misfit
The IPC looks "roughly the same" between the two.
52%+ performance for AMD for 60% more cores.
At this point, Intel is just lacking cores.
Ratilt
Just lacking cores? I don't think you see the bigger picture here. If it is "just cores" why Intel not mach it or come close? Efficiency also matters for the server market and AMD product is leaps and bounds ahead of Intel's. I wonder what would've been the TDP of Xeons if Intel add 20 more cores. Besides, I've found Intel's TDP inaccurate of what it actually is.
kapu
Not ture at all . Look at performance per watt . For datacenters it is what is most important and at this point difference is huge , just another level . Like i said before , like a decade of performance leap.
Kaarme
Dude, they get much more work done in the same time, with marginally higher TDP. It's like transporting people with a van, compared to a regular car. Even if the van used 10% more gas, it wouldn't matter since you could move twice as many people.
Richard Nutman
What matters is work per watt. And the Cinebench score per watt on AMD's is miles ahead of Intel.
You need roughly 1.52x more Intel dual sockets to match the AMD score, which would give you a TDP of 410 watts! Far higher with the Intel solution.
rl66
It's more due to:
- less core
- less frequency
- less cache
counterpart considerating above:
- The Intel's price make the Epyc even more desirable for a company...
AMD play well to regain the PRO segment, last week it was aviability of pro GPU, now this...
rl66
Yes and No, it depend on what you do per W and what it cost and gain... If cooling a bit more and it does enough more to compensate this charge it's OK.
It depend on what the DC is intended to be used to.
Kool64
also there's another aspect of having so many cores all in one place: Space savings... can have a much smaller server farm which would potentially save on rent or allow more space for a sea of cubicles or something like that.
RealNC
Who cares :P
schmidtbag
Seems a little weird to me you're not considering performance-per-watt. We're talking a roughly 35% performance increase for a 4% increase in power consumption. This of course is assuming TDP translates to power consumption, which overall it doesn't really, but in either case, the Epyc is very obviously doing more work while using proportionately less power. That's a win for datacenters, where as you said, this sort of thing matters.
The extra performance also means they need fewer servers, which in turn means a lower electricity bill.
H83
This^^ AMD´s CPUs are better overall but their biggest advantage over Intel is the core count, if Intel can match AMD on that aspect, then things become pretty even.
And that´s why i´m amazed that Intel is still to announce something similar to EPyc/Ryzen regarding their multi core approach...
kakiharaFRS
my experience from comparing a 8c 9900k to 24c 3960x both oc to whatever I could with the same liquide cooling setup (minus the cpu blocks)
around 1.1x watts for 3x the cores and 2.8x cinebench scores
intel max temp 89°C amd 67°C (same fans same rads same case !)
the temps and watt/perf were pretty insane on AMD the only things where it suffered from the comparison were the overall system power usage (all pcie 4.0 and several times the lanes isn't free) and idling power because even if you tune them down there's still 24 of them rather than 8
with high cpu usage tough the tables reverse and AMD becomes so much better
can't wait for zen 3 threadripper so i can ditch "I has no lanes" intel
edit : H83 not only the core count, the biggest problem for Intel is that a single tiny die cannot be cooled properly, amd design helps cooling a lot with my current setup the cpu goes from max temp at 100% cpu usage to idle temp in <1s whatever highest temp was even 100°C so it's not that my loop can't cool it's just that a small surface can't "exhaust" or whatever you call that enough heat
BLEH!
Broadwell is 6.5 years old give or take, so not far wrong!
itpro
@nizzen says these CPUs are only good for playing cinebench and they are cheap because they aren't exotic.
nizzen
Laughing when rendering with 2x RTX 3090 on my old x399 with 1950x 16 core 😛
rl66
Totaly agree, i have made somewhere in here a post about it.
In most case company change material around 4 or 5 year.
Epyc started to be very intersting around 2 year ago...
That mean in 1 year for early adopter, and in 2 or 3 year, we will see if they will keep Epyc on renew to have this info.
If yes it might be for AMD the return on the Pro segment for time, but for that they have to stay on the run whatever it's positive or not.
(what might be a good point is that pro motherboard maker seem to extend the AMD line, if they do is that there is a demand).
waltc3
If Intel could make a competitive CPU it would, but it can't, so it doesn't. Intel will try and sell "productivity benchmarks" over performance per watt. I wish them luck because they will need lots of it!...;)
Kaarme
They can, it'll just take a bit more time. In the meantime they are still making as much money as ever, so business, let alone existence, wise they have got nothing to worry about. They kicked out the technology hater CEO, so things ought to be rolling for them again in the near future, once research starts to bear fruit. Of course it also depends on how well they get their 7nm process working, but I wouldn't expect a catastrophe there. They had giant problems with the 10nm, but as we all know, a failure can teach more than 10 successes.
