Ugrading the video card

[gptech]

You're mixing up threads and cores, it doesn't matter that TS natively is written to recognise dual cores, it's how many concurrent threads that are running that's important. No application "fully loads cores", to do so locks that core up.

I just can't quite justify spending a couple of hundred quid, just so I can have shadows and a bit less catenary and fence shimmer... it runs well enough for me if I'm honest, but there's definitely (a little) more available from a newer card.

As you can't justify spending to get little gain--your words---why not try overclocking the CPU; you should hit 4.2GHz with little bother.

[peterfhayes]

why not try overclocking the CPU; you should hit 4.2GHz with little bother

.
But surely, Gary, then, if he didn't have a video card capable of receiving and rendering the faster data (ie the gpu matched to the cpu) you would still see poor performance - possibly the dreaded (blue) bottlenecking??.
IMO good performance comes from matched components from the motherboard to the case.

TSxx has been multicore ready since TS2012 and I see from "live" measurements that although cpu 0 is the predominant cpu core (used also by the OS) cpu Cores 1 to 3 are all used to some extent. That is the way that Windows works. If you want to see how TS 2015 runs on a SINGLE core, open task manager then open TS 2015, tab back to task manager and highlight "railworks.exe" right click and choose " set affinity" set it to say cpu 1 (TS2015 will now run only on cpu 1) - then go back to TS 2015 and see how well it now runs (doesn't). If your "set affinity" is not set to use all cores (default) then that may be your issue.

pH

[gptech]

But surely, Gary, then, if he didn't have a video card capable of receiving and rendering the faster data (ie the gpu matched to the cpu) you would still see poor performance - possibly the dreaded (blue) bottlenecking??.

True Peter, but what we have here is a 'gaming class' GPU coupled with an entry level dual core Pentium; OK, it's not your average entry level dual core Pentium as it's overclockable, very overclockable in fact. Overclocks of 50% above stock are well documented and whilst it's true that not every individual chip will hit those heights, 4.2-4.5 GHz should be attainable. I wonder if the Passmark figures are a little skewed by overclocked systems forming a fair sized chunk of the samples tested; every review you'll find mentions something along the lines of :
"system builders can design and build a system based around Haswell and still reap similar performance to that of a Core i5/Core i7 chip if they apply an overclock..." (http://www.vortez.net/articles_pages/in ... ew,15.html---emphasis by myself) so it sitting above i3/i5 CPUs at it's stock 3.2GHz doesn't add up.

[peterfhayes]

Agreed.
pH

[ttjph]

You're mixing up threads and cores, it doesn't matter that TS natively is written to recognise dual cores, it's how many concurrent threads that are running that's important. No application "fully loads cores", to do so locks that core up.

OK, if we're going to that level, TS (when I looked a while ago) ran upwards of about 15 threads. However, one of those threads is the limiting factor (in a CPU-limited setup): for practical purposes (ignoring how Windows spreads the load across cores) one core is fully occupied by that thread, in that as soon as it finishes the calculations for a frame, the other threads have already finished (using less that half the time available on one other core) and the limiting thread starts again immediately for the next frame.

Having one core would definitely be a step backwards, as that would obviously have to do everything (and there would therefore be a delay between finishing and restarting the limiting thread while it processed all the others); but having just two cores is enough that (the equivalent of) one core can be dedicated to the limiting thread, and the other core can do everything else and still be left twiddling its electronic thumbs.

In my case, all the CPU threads have time to complete before the video card finishes rendering a frame, and asks for the next lots of data, so both cores get to twiddle their thumbs for at least some of the time.

As you can't justify spending to get little gain--your words---why not try overclocking the CPU; you should hit 4.2GHz with little bother.

Because all that would do is increase the CPU idle time between frames, and make the room slightly warmer...

Following another thread I'm now rather liking the look of a GTX 480, as they go for relatively little on Ebay, are dual-slot (my preferred cooling solution - get the hot air straight out of the case), and are a big step up from the 260 on all the obvious metrics.