It doesn't matter what the test results end up being, it's about appearances. I expect that everyone that had an issue with Mooley Eden will be very vocal about this.
Well then, independent testing will soon confirm or deny that score adjustment.
It's not unfair. AMD can build their own compiler if they want to take advantage of their processor's specific optimizations (so long as they translate in actual application deployments).
That said, it would be helpful to see both products compared using GCC or the like.
That sounds like explicit suggestion to use icc with Naples instead of multiplying Intel result by arbitrary number, frankly. Especially since we do know that icc generates perfectly okay code for Ryzen.The Intel compiler is indeed "optimized for SPEC" - it employs optimizations that are either illegal in a language such as C++ or not applicable to pretty much any real program outside the SPEC benchmarks that is larger than a few hundred lines of code. You have to use a magical combination of flags to get anywhere close to the numbers they publish, and if you try that on other programs you either don't see any improvements or might introduce runtime bugs.
I'll just mention that Geekbench 4 uses LLVM for all platforms.Comparing hardware when the platform compiler is different (think Mac, Linux, Windows) is also an issue for other benchmarks, such as Geekbench
I'll just mention that Geekbench 4 uses LLVM for all platforms.
Eh? I mean, come on, it leaves the raw numbers for workloads too.Geekbench is useless for inferring workload-specific real-world performance because it is a weighted score.
Huh, i must be remembering wrong, cause i swear i read somewhere they switched to Clang on every platform. Oh, wait, i kinda get where my confusion originates from now.Where is that mentioned? The only thing I could find was this: http://geekbench.com/doc/geekbench4-cpu-workloads.pdf, which indicates on page 4 that VC++ is still used on Windows.
For vacuum comparisons you're right.The only good way to compare hardware is to use exactly the same OS, same compiler and same compiler flags. This is one reason why you cannot really compare the Ipad Pro GB4 results with Windows laptops, for example.
But ultimately, geometric mean with libquantum is worse than plain weighted mean, if you ask me.Geometric means are better than weighted means when looking for baseline performance.
Geekbench subsection scores are also a geometric mean of the benchmarks under each subsection.But ultimately, geometric mean with libquantum is worse than plain weighted mean, if you ask me.
Intel's response to the server launch is indeed EPYC
"... AMD’s approach of stitching together 4 desktop die in a processor i
If it was, it was not Intel's complaint/excuse.Wait , was it not just a few months ago that it was a complaint and/or an excuse that Ryzen 7 was just a server CPU ?
google translate said:When memory benchmark stream on a 2P system with EPYC 7601, the transfer rate is with measured 247 GByte / s (Triad) in well twice the value of the system with two Xeon E5 2699Av4. At the Linpack, Epyc expected to be less than the competitor, with 1050 to 1520 GFlops, it comes to about two thirds. The test device ran under Ubuntu Server 16.04.
Cinebench R15 has changed to Windows Server 2016 [-] is also running. The Cinebech value first swayed between quite a decent 5400 to 6000, but AMD technicians were still working on the configuration and ultimately values came to 6879, about 1300 more than the Xeon E5-2699Av4.
For business confidence it's good to see the high quality testimonials that AMD Epyc processor powered systems are receiving. You can see a summary of several such statements from industry heavyweights above. For further details AMD has published individual testimonial videos from the likes of Microsoft, Samsung, Baidu and 1&1.
Technically, they did the same for Bulldozer Opterons, didn't they.Further to my previous comment it seems Dell EMC, HPE, Supermicro and Tyan know something Intel does not and all have announced Epyc servers:
Technically, they did the same for Bulldozer Opterons, didn't they.
So, i will wait, even though perf numbers look great.
Said they are offering it for customers, that's important distinction, just in case. Similar with 1n1 from my perspective.
Said they are offering it for customers, that's important distinction, just in case. Similar with 1n1 from my perspective.
Samsung left me confused because most of it sounded like them promoting their memory and (subtly) process (cough-cough), but it'll do cause I/O and memory capability is a real advantage of these, even if it is the one i am not particularly interested in because most of it is known value already.
Baidu was talking about efficiency and TCO, so once again, not something i am interested to know (beyond knowing that they do keep them right in TDP limits).
But that aside, all the OEMs you mention offered Bulldozer systems too, so them adding EPYC to portfolio is nothing interesting, especially since EPYC objectively can't be a bulldozer tier chip. The interesting part is what market will that 8 socket in 2 socket space trickster reach.
Did you miss the part where i have repeated twice that OEMs including Microsoft offer Bulldozer systems too? Fine, i have just repeated it the third time. OEMs offering something is not particularly interesting. Baidu using it is interesting, but they are clearly going for efficiency points here and that's great (for AMD). But does not address the Intel's nitpick at nature of Naples, like you think it does.
Why, i could. I could say that every major OEM offered and/or used Pentium 4 Xeons and that obviously makes Pentium 4 a great server CPU.Trying to name drop Bulldozer all the time is really weird - why don't you name drop the Pentium 4 in the Intel threads then??
Yeah, that's where it all becomes mostly clear. It is obvious they had to be sampled around for months, if AMD does not secure in any big orders in Q2 or Q3... it would be a curious case.The uptake and marketshare is what gives us real insight into the performance. Q2 should give us some hints.
Did you miss the part where i have repeated twice that OEMs including Microsoft offer Bulldozer systems too? Fine, i have just repeated it the third time. OEMs offering something is not particularly interesting. Baidu using it is interesting, but they are clearly going for efficiency points here and that's great (for AMD). But does not address the Intel's nitpick at nature of Naples, like you think it does.
Why, i could. I could say that every major OEM offered and/or used Pentium 4 Xeons and that obviously makes Pentium 4 a great server CPU.
Because in regards to Intel threads, P4 was not relevant since at the worst Nehalem.I have not seen you once name drop the Pentium 4 in any Intel thread in the last year
What is related is the fact that OEMs offered Bulldozer systems, just like they offer EPYC systems now. As such, talking about OEM offering products as stamp of product quality is a fallacy. Come on man, you're not that dense, are you.yet you seem to desperate to keep talking about Bulldozer which was launched six years ago in this thread when they are not related
Well, this is also the first thread i get to see where people mention OEM offering products based on a CPU as sign of quality of said CPU. That fallacy deserves to be punished even if EPYC was 10 times faster than Skylake-SP at half the TCO. And see, you proceed to repeat that same fallacy for the 4th time with the different CPU. Maybe i am wrong about you, after all.This is the only Epyc thread on a few big tech forums I have had a quick glance where people are trying to keep on name dropping Bulldozer.
Thanks, i knew i was missing something on my own ignore list.Do it. It'll save you hassle!