Evaluating CPU cores for optimisation - page 5
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
The 2990WX is different. It consists of four Zeppelin chips, with 32 processing cores. On the X399 platform, AMD has imposed some restrictions on this processor so that it doesn't hurt sales of EPYC server chips.
Chief among these limitations is the presence of only four memory controllers. Although there are two more Zeppelin chips, AMD calls them compute chips. This means they don't have access to local PCIe or DRAM, for that they have to address I/O components via Infinity Fabric. Since there are twice as many crystals, the bandwidth of the Infinity fabric is twice as low, around 25Gb/s if DDR4-3200 memory is used.
Here, if there is any active memory handling, it's only occasional reading - reading 64 times the EA code from memory is very fast, and clearly doesn't take 100 seconds!
The2990WX cannot be slower than FX-8350 under load of equal number of cores!
Besides, you and I loaded this processor in R, and there the performance was significantly better than that of the FX-8350, with each thread eating up around 100 megabytes.
It looks like the compiler is tuned to some peculiarities of Intel processors.
Although, it could be that agents are independent of each other - it's like different programs and then they can load the bus with constant data overload to get new jobs (pieces of code for execution) for each agent, but I'm not an expert in this matter, obviously.
If this is true, then it is time to change the agent ideology by making the program (EA) common to all cores, and execute the code itself synchronously - on more cores this may be faster than the current asynchronous execution.
Build 2085
Debian9 Wine 4.0.1
Asus P8P67PRO
Intel@ Core i7-3770K CPU @ 3.50GHz
RAM 4x4 16Gb
Tree_Brut_TestPL
Tree_Brut_TestPL_F
RAM 4x4 16Gb
Tree_Brut_TestPL
Tree_Brut_TestPL_F
Thanks for the tests - updated the rating.
Here you can see that without hypertrading it's not good at all - slower than Celeron G3900...
I think anything slower than a Celeron should be changed...Thanks for the tests - updated the rating.
Here you can see that without hypertrading it's not good at all - slower than the Celeron G3900...
I think that slower than the Celeron, it's advisable to change already...Maybe the test is not correct because of Wine.
Because under Wine the agents show Intel Pentium 4 2.40 GHz
It's not clear what configuration was actually used.
Perhaps we should add Wine correction into the rating.
I will try to do a test on a virtual machine in Windows10 a bit later.
It is possible that the test is not correct because of Wine.
Under Wine the agents show Intel Pentium 4 2.40 GHz
And it is not clear what configuration is actually used.
Perhaps we should add Wine correction into the rating.
I will try to do a test on a virtual machine in Windows10 a bit later.
Can't you just try it in Windows 7 /10 , without virtual machines?
Can't you just try it in Windows 7 /10 , without virtual machines?
I have linux as my primary system, now because of the test to reinstall the operating system, as it is not kamilfo)).
I may try it after some time, when I reinstall it back to Windows because I am convinced that Wine is not suitable for mt5.
I have linux as my primary system and I'm not really comfortable reinstalling the OS for testing purposes )).
I may have done that after some time, after reinstalling the system I should switch back to wine. I am convinced that Wine is not suitable for mt5.
If the tests are any better, it's a good reason to have a Windows system, at least for testing purposes...
The virtual machine didn't help.
Most likely the problem is virtualization, both on Wine and VM.
Because it's kinda weird that an i7 4 cores, is inferior to a celeron 2 cores.
Build 2093
Windows10 VM VirtualBox
Asus P8P67PRO
Intel@ Core i7-3770K CPU @ 3.50GHz
4x4 16Gb RAM
Tree_Brut_TestPL
Tree_Brut_TestPL_F
I decided to look into what instructions differentiate between processors - well there can't be such a performance boost for an i7-8700 out of the blue, so for comparison I took the 2990WX, FX-8350, E5-2670.
Here's a map of the instructions on which:
Grey - instructions are all there.
Green colour - instructions are not available for all
Pink - similar technologies/instructions
Blue - unique processor instructions
Yellow - shows lack of instructions compared to i7-8700
Source of information.
We see that 2990WX has all instructions which are present in FX-8350 and i7-8700, which means the performance of cores should be comparable for the same task (may be a little bit slower due to frequency, but it's in theory, if we ignore the progress of microprocessors, purely on the availability of logic). At the same time FX-8350 has instructions which were dropped in 2990WX, and maybe not dropped, but just gave another name (marketing) - it's good to check, for those who understand.
Further, let's compare i7-8700 with E5-2670 and pay attention to its instructions presence and relatively to FX-8350 - we see that other processors have no instructions BMI1, F16C, FMA3 - what are they responsible for, and is their absence critical - that's the question!
The virtual machine has not helped.
Most likely the problem is in virtualisation, both on Wine and VM.
Because it seems odd that an i7 4 cores, is inferior to a celeron 2 cores.
Yes, there are some oddities here - we need to get more statistical data to assess the situation.
Thei7-3770K has no BMI1, FMA3 instructions - maybe that's the reason.