Building a gaming system

[Wild Cobra]

$225 is actually a pretty terrible price for a 32GB DDR3-1600 kit. Here's a good one for $145. RAM prices have fallen a lot in the last 12 months.

http://www.newegg.com/Product/Produc...82E16820231490

Well, did you look a the timing and latency difference?

I like the Viper better. It's access is faster.

I took 2 x 4 GB out of the system with an 11 CAS timing. The Vipers have 9 and I put 4 x 8 GB in, and the ones in your link have 10.

[Wild Cobra]

Disabling VM is not a good idea on modern OS'es. If anything, you want to have a faster drive, let me explain why.

Modern OSes will use up all the RAM you put at their disposal. That's because they use any unused RAM as disk cache. But it will release it to apps as soon as is needed.

Furthermore, OSes in this day and age suspend and swap out processes that are hidden and are not doing compute or server tasks. This is a good thing. It gives active apps have more RAM and compute cycles. Once all active processes do eat up all RAM, then you'll inevitably hit paging, and things will slow down (including the fact there's not enough RAM for a disk cache, which really makes paging slow).

Now imagine if there's no VM available. Processes can be suspended, but they must remain in memory. That means more memory fragmentation and potentially less RAM available to all processes, including the front most process.

I say "imagine", because you really can't turn off Virtual Memory on Windows. What you're allowed to control (whether you have a pagefile or not, and what size is it) is only for "private commited" memory, one "kind" of memory, but will likely cause other type of memory (ie: mapped memory) to swap out more often. That's why, if you're really short on memory and you suspect you'll page out, a faster drive makes more sense.

Here's a good writeup about the misconceptions of Virtual Memory on Windows:
http://azius.com/blog/pagefile-yes1/

Well, if I have problems, I'll change back.

The Tower is nice now that it has 32GB in it. I haven't tried the game in it yet. I copied the ship file over to a USB stick, and will do so later. Once I installed the four Vipers, I made sure they worked, then powered down and transplanted my 500 GB Barracuda from my Lenovo XP computer so I have all those files on hand. This tower now has the original 1 TB and DVD writer, an added Bluray player, and the extra 500 GB Baracda. Took be a while to figure out the permissions, even in admin mode since I never did this before, it was tricky for me to see the locked drive and make it available to admin and my user profile.

Now I have a Lenovo desktop with no HD that I need to find a new home for. It has a win 7 license, but not installed. I bough it with XP. I don't need the extra 500 GB. I might just copy everything over I want to keep, and put the HD back in, format it, and install win 7. Who knows, maybe I'll just dedicate it as a backup drive?

I wonder if the 2.5 HD from my decade old laptop will plug~n~play with it?

[Cry Havoc]

Well, did you look a the timing and latency difference?

I like the Viper better. It's access is faster.

I took 2 x 4 GB out of the system with an 11 CAS timing. The Vipers have 9 and I put 4 x 8 GB in, and the ones in your link have 10.

Latency timings are probably the last bottleneck in the world you'll experience with RAM. You are talking about paying almost $100 more for *fractions* of a second. It's not worth it, you could easily pour that into say, a SSD, and have much better results.

[ElNono]

You definitely overpaid... here's a PNY 32GB DDR3-1600, Cas 9 kit for $139.99:

http://www.amazon.com/PNY-1600MHz-PC.../dp/B00KMXUOOW

Here's the Crucial Ballistix, also Cas 9:
http://www.amazon.com/gp/product/B007HAXMGA

then again, it's your money, tbh...

[Wild Cobra]

Latency timings are probably the last bottleneck in the world you'll experience with RAM. You are talking about paying almost $100 more for *fractions* of a second. It's not worth it, you could easily pour that into say, a SSD, and have much better results.

Well I prefer for one, to support the local community than mail order anyway. We ave enough loss of jobs without mail order cheapskates making it worse.

I also didn't want to wait and I hate UPS.

[Wild Cobra]

You definitely overpaid... here's a PNY 32GB DDR3-1600, Cas 9 kit for $139.99:

http://www.amazon.com/PNY-1600MHz-PC.../dp/B00KMXUOOW

Here's the Crucial Ballistix, also Cas 9:
http://www.amazon.com/gp/product/B007HAXMGA

then again, it's your money, tbh...

Yes, and Amazon has the same memory I bought for $175.

[Wild Cobra]

why do u need more then min 8gb-16gb ram wild cobra?

u running a server or some at home?

Because it can be done.

[Cry Havoc]

Well I prefer for one, to support the local community than mail order anyway. We ave enough loss of jobs without mail order cheapskates making it worse.

I also didn't want to wait and I hate UPS.

I wasn't taking issue with your place of purchase, I was commenting on how little of a performance difference you see with latency differentials. Which is to say, practically none.

[Wild Cobra]

I wasn't taking issue with your place of purchase, I was commenting on how little of a performance difference you see with latency differentials. Which is to say, practically none.

Sometimes, every little bit helps.

A system isn't any better than it's weakest link. I think for 32 Gigs, the price difference is acceptable.

[ElNono]

So I got me the Y cable for the fans, now the side fan is working too and CPU is idling at 35C...

Also snagged a cheap Chinese XBox wireless receiver, working like a champ on Win10

[DJR210]

Because it can be done.

Well then 4-way SLI 980Ti or your PC sucks

[Wild Cobra]

Well then 4-way SLI 980Ti or your PC sucks

LOL...

$640...

I don't need that type of graphics speed, but I sometimes have dozens of large applications open.

[baseline bum]

This Transformers game looks awesome.

[baseline bum]

Well then 4-way SLI an X or your PC sucks

fify

[Wild Cobra]

Your CPU load on the other screens doesn't look that bad. Memory seems not to really go over 4GB, so adding more RAM isn't going to change that.

It's entirely possible that the game engine just isn't coded to push hardware to the limit. That's pretty common when you're not using a commercial engine like Unreal, etc.

You might just need to bug the devs to optimize their code.

The game uses Unity 4, only supporting a single thread. I doubled my game performance of my tower by disabling hyperthreading in the BIOS. I discovered the game was maxing out at 15.04% CPU usage. I assume unity was using 12.5%, fully utilizing one logical core, and other aspects of the game operating in other cores. After changing the computer from eight logical cores to four, matching the four physical cores, I now get double the framerate and the game is capable of utilizing more than 27% of the CPU.

I can't disable hypertrhreading in my laptop. It has the latest E175AIMS.10H Bios version, but no thread switch. In searching, I found claims of editing the NVRAM without the BIOS, but it required downloading executables I have a hard time trusting, and for a different MSI computer.

If anyone knows how to change my laptop from eight logical cores, to four logical cores, please do tell.

It's the MSI GP70 2QF. Also known as the Leopard Pro.

[Wild Cobra]

Too bad they don't use Nvidia PhysX to allow your graphics card to do the physics calculations and let the CPU just deal with game logic, but I guess that would be screwing over people who bought AMD cards.

Up until Unity 5.0 the engine was using a fairly outdated version of Nvidia's PhysX physics middleware. Unity 5.0 included version 3.3, which is standard among Triple-A games.

https://en.wikipedia.org/wiki/Unity_...ine%29#Unity_4

There is a reported rumor in the KSP forum that an updated version of the game that will use unity 5. I read it allows actual mul hreading and 64 bit instead of 32 bit in the game memory usage.

KSP isn't using the CPU for the graphics, but for the calculation of stress, heat, etc. of the individual parts. It doesn't start slowing down until the CPU has to do the calculations for several hundred parts.

My tower has a very fine processor, but an inferior graphics card compared to my laptop, and it doesn't even break a sweat for the game. It's the GeForce GT 720.

[baseline bum]

That's annoying when a game is parallelized so poorly turning off HT makes it run better. Far Cry 4 is like that too, your performance is dominated by one core. Thankfully most games run better with hyperthreading on.

[Wild Cobra]

At least the CPU in my computer is in 5th place for single thread benchmarks:

http://www.cpubenchmark.net/singleThread.html

[baseline bum]

Apparently all RAM isn't the same for gaming. It turns out when you're CPU bound the speed of your RAM can make a noticeable difference, or in the case of Ryse Son of Rome, a night and day 50 fps difference.

[Wild Cobra]

Apparently all RAM isn't the same for gaming. It turns out when you're CPU bound the speed of your RAM can make a noticeable difference, or in the case of Ryse Son of Rome, a night and day 50 fps difference.

Yep.

And here some joker thinks I had no reason to buy CAS 9 RAM.

When memory starts constraining you, every little bit helps.

[baseline bum]

Yep.

And here some joker thinks I had no reason to buy CAS 9 RAM.

When memory starts constraining you, every little bit helps.

Sucks you need a Z series or X series board to run above 1600 MHz though (or 2133 MHz on Skylake). To be fair for years gaming benchmarks have shown RAM speed doesn't matter at all, and it's true for the vast majority of games which are going to be GPU bound unless you're running an extremely high end GPU, e.g., at least a $650 one, or a low end CPU. But almost no one benchmarks things like memory with low end CPUs to force a CPU bottleneck, so this performance gain when CPU bound has slipped through the cracks in gaming hardware sites.

[ElNono]

I would argue there's zero reason for Ryse Son of Rome to show a 50 fps drop... I suppose it has to do with physics, which is largely CPU bound... but still, sounds like a lack of optimization...

[baseline bum]

I would argue there's zero reason for Ryse Son of Rome to show a 50 fps drop... I suppose it has to do with physics, which is largely CPU bound... but still, sounds like a lack of optimization...

Ryse shows all kinds of weird . It actually drops 10% in framerate on a 980 at 1080p ultra when run at PCIE-2.0x16 vs PCIE-3.0x16 when no other major game shows anything more than maybe a 0.5% difference. I wonder if Star Citizen will be like that too, being another unoptimized CryTEK game.

[Wild Cobra]

I sure hope when KSP v 1.05 is released, they fix the slowing down because of the single core usage.

[Wild Cobra]

Here are my CPU Passmark results with hyperthreading on and off: