and you are talking about code that was never written for that purpose,
even though they have tried to offload some it to utilise some
parallelisation recently. But it needs to be there from the start to get
much meaningful from it - and then have compilers and programs to
provide it (and they definately were not common back in 2012 when this
was released, much less for the years previous it took to write it onto
RoF engine).
Whilst MS2020 may have it, I doubt many combat flight simes are going to
provide it for a good few years. Its almost a restart
TTFN
Stickz
On 23/10/2020 20:41, 56RAF_phoenix wrote:
I should have added: it's possible the devs have put in a lot of fine-grained parallelism at the loop level to exploit hyperthreading. But that's a lot of effort that usually doesn't pay off unless you have large amount of independent work on each loop interation. That seems fairly unlikely in a flight sim. You only get really efficiency using coarse grained message passing (MPI).
As you can tell, I'm pretty skeptical about parallel code efficiency... and I run parallel programs that I founded on /tens of thousands/ of processors.
56RAF_phoenix
On 23/10/2020 20:33, 56RAF_phoenix wrote:
when I turned off hyperthreading and just run on 4 cores those cores looked liked they were being pushed harder % wise than if I had hyperthreading onAs a former expert in parallel programming for similarish coding, I'll say that just because the cores look busy doesn't mean they're doing /useful/ work. Just doing the context switching and reloading the cache will be counted as cpu work, when all you're doing is changing threads. Every time you switch hyperthreads it costs useless effort. It's only worth while if the other thread is in some wait state at a sync barrier.
You'd have to add the % usage of the 4 cores together, then the 8 hyperthreaded "cores" for the same instant of flight, and even then you'd probably find the totals for 4 cores might be lower.
56RAF_phoenix
On 23/10/2020 18:50, Joe Stronach (Redacted sender jstronach for DMARC) wrote:Cheers Col, agree.
I think there's is still quite a bit of legacy 'buy the most expensive Intel cpu you can afford' opinion on the forums, but maybe that will have turned around completely quite soon.
Personally I'm pleased that the B450s have been given an upgrade path by AMD after all. Although my 3600 will hopefully suffice for quite some time yet.
Roke
Get Outlook for Android <https://aka.ms/ghei36>
------------------------------------------------------------------------
*From:* 56raf_firebirds-bounce@xxxxxxxxxxxxx <56raf_firebirds-bounce@xxxxxxxxxxxxx> on behalf of Colin Pack <dmarc-noreply@xxxxxxxxxxxxx>
*Sent:* Friday, October 23, 2020 5:39:28 PM
*To:* 56raf_firebirds@xxxxxxxxxxxxx <56raf_firebirds@xxxxxxxxxxxxx>
*Subject:* [56raf_firebirds] Re: AIO Radiator
Joe, when we did some testing just after you started playing IL2, we found that the 3600 6c/12t 4.2ghz was not struggling on VR at the 11.1ms frame time using fpsvr. The issue if anything was your Titan like our 1080ti’s were getting to the 11.1ms, so could not push the graphic settings any higher. The same for my overclocked i7 4c/8t 4.7ghz staying below the 11.1ms.
If you start to pushing the refresh rate up, which lowers the frame time:
72hz 13.9ms
80hz 12.5ms
90hz 11.1ms
120hz 8.3ms
144hz 6.9ms
That’s when you start getting, graphic problems, etc.
The Index can be played at 80hz which I do use, I don’t hit problems until 12.5ms, so as long as mainly object loading does not push my cpu above that 12.5ms, that side of the game runs fine.
As a side note when I turned off hyperthreading and just run on 4 cores those cores looked liked they were being pushed harder % wise than if I had hyperthreading on as all the cores threads seemed more lower and balanced out. This was using fpsvr to watch the cores not sure how accurate it is.
Now with the gpu if I setup the graphic settings that did not push the gpu above 12.5ms then again great.
So like you say it’s the graphics card that is the main problem, the cpu online will not be the problem most of the time. As long as you don’t meet up with 20+ aircraft and firing at once or flying over a large target covered with loads of flak guns, which is rare online.
Colin
Sent from my iPad
On 23 Oct 2020, at 15:46, Joe Stronach <dmarc-noreply@xxxxxxxxxxxxx> wrote:
I have one on order Pheonix, which I'm going to pair with a Ryzen 3600 initially. I think your analysis is going to be proved right tho...
Roke
Get Outlook for Android <https://aka.ms/ghei36>
------------------------------------------------------------------------
*From:* 56raf_firebirds-bounce@xxxxxxxxxxxxx <56raf_firebirds-bounce@xxxxxxxxxxxxx> on behalf of 56RAF_phoenix <phoenix@xxxxxxxxxxxxxxxxxxxx>
*Sent:* Friday, October 23, 2020 11:46:07 AM
*To:* 56raf_firebirds@xxxxxxxxxxxxx <56raf_firebirds@xxxxxxxxxxxxx>
*Subject:* [56raf_firebirds] Re: AIO Radiator
This <https://docs.google.com/spreadsheets/d/1gJmnz_nVxI6_dG_UYNCCpZVK2-f8NBy-y1gia77Hu_k/edit#gid=1719385058> spreadsheet posted on the forums is very interesting. Though I suspect there are too many individual rig and settings variations to make it definitive, there's no indication that I can see that the current Ryzen CPUs are any slower in VR, the Min Frame Rate looks graphics card dominated. I'd love to see 3080 numbers when someone gets one.
56RAF_phoenix
On 17/10/2020 20:38, Colin Pack (Redacted sender bart_56 for DMARC) wrote:https://www.overclockers.co.uk/amd-ryzen-5-5600x-six-core-4.6ghz-socket-am4-processor-retail-cp-3cc-am.html
Sent from my iPad
On 17 Oct 2020, at 19:15, tim foster <dmarc-noreply@xxxxxxxxxxxxx> <mailto:dmarc-noreply@xxxxxxxxxxxxx> wrote:
ok forget it see there is a new 5600 series (why clash bloody numbers fgs) - but I dont see anyone quoting uk price
