Intel Dual Socket Ice Lake-SP Processors Pitted Against AMD EPYC Rome
Click here to post a comment for Intel Dual Socket Ice Lake-SP Processors Pitted Against AMD EPYC Rome on our message forum
DeskStar
I likes me learning new stuffs.....
RED.Misfit
Not the same OS, not the same memory size (don't know what impact it could have).
But the best is yet to come, Ice Lake-SP should have challenged Epyc Rome. Sadly for Intel, they are a little bit late and it's gonna have to face Milan instead at the end of the year.
caldas
Started very well, Kernel Windows 10 Vs Kernel Windows Server 🙁
NewTRUMP Order
Ok so lets get down to what really matters in the business world. What does each one cost?
Vincenzo
If i remember right.. windows 10 Pro have some problem using processor with to many core (32+), instead the server edition, or the workstation edition work much better.
Fox2232
Depends on power consumption. Cost is like marathon and upfront price is often of no importance in long run.
schmidtbag
That is going to be interesting. The Epyc series has often been nearly half the price for the same amount of cores. Intel has been lowering prices but even in the unlikely event these Ice Lakes are a better value, Epyc also has dual-socket options that Intel just has no answer to at all.
Fediuld
One note. The EPYC multicore prerformance is hampered by the Win10Pro multicore restrictions also the benchmark is of different versions between the two systems. The later version is been patched to work better with 128+ core CPUs.
JamesSneed
Yes because Geekbench is how you test high core server chips. The results of this regardless who wins is pointless. This will not translate to actual server workloads so nobody looking to buy these chips will even care about this test.
JamesSneed
You might be surprised most large enterprises don't care about the purchase price all that much. The licensing of the software that runs on top of the systems is usually many multiples the cost of the hardware. They end up caring about floor space used for x performance and performance they get for the power used as both of those are finite in a data center.
Borys
MAN.... Intel is DEAD!
Zeka
Should not bother releasing such info, different OS, different amount of RAM, different version of Geekbench, single cpu vs dual cpu... no point in comparing really
Gomez Addams
"Arguably you can discuss the results."
Arguably? What the heck does that mean? Of course we can discuss the results! There is absolutely no reason we can't. It is NOT arguable at all.
The question is why would we bother to? It is Geekbench. Who actually cares about the results of that test? It is about as irrelevant as a test can possibly be.
Deasnutz
I've yet to see a single AMD based server in the enterprise field, since 2005, so it will take time. Tons of products and server appliances use Intel based platforms, so it will take more than a few benchmarks to kill that.
TheCapulet
JUST built one last week. Was a fun little "money is no object" project from one of my clients that runs individual VMs for everything. Few things are better than holding an $8k processor in your hands.
moo100times
We have seen this approach from intel before, so I am suspicious at all the above.
My biggest gut feeling is that if Intel did have something, they would be showing outright how much "something" it is, rather than such dodgy comparisons. Is this simply avx512 support?
In broader news - are we watching Rome fall?
NewTRUMP Order
Exactly and what does each one cost in power usage? Back to what matters in business=cost.
If you are not watching your cost on EVERYTHING in your business then you are running a losing proposition.
NewTRUMP Order
There are 100 Epyc based servers in the world. Apparently your vision is limited.
GSDragoon
As others have mentioned, the test is very flawed. More Intel propaganda?
Dazz
Geekbench 4.2.2 is from February 2018, ROME came out in August 2019 so the software is already a year and a half out of date, GeekBench 4.3.3 from March 2019 i would assume Geekbench had problems in utilising 128+ threads since it predates such multi threaded processors. It also calls into question with the use of Windows 10 Pro and if they have the patches to utilise the cores the software they are testing is clearly won't utilise a ROME Based processor.
A google search says the ROME from July this year can get 35k with Windows 10 Pro
https://browser.geekbench.com/v4/cpu/14856570
Then Linux has 76k with the same processor and 64 cores
https://browser.geekbench.com/v4/cpu/14569217
So changing from the Windows 10 Pro handicap version to one that can utilise the cores increases performance over 2x.
