TR Forums

True, for pure gaming, 8GB is more than you'll need for now. Most games haven't even made the transition to 64-bit yet, so as I understand it they can't even address larger than 4GB as a single process. Unless you're planning to run two games at once, 8GB is fine.

Of course, the time will come when it isn't... you just have to judge if that time is likely to be soon enough to take the plunge now, or whether it's fine to put it off for a couple of years.

Luckily I don't have to consider it... I need a minimum of 32GB for my work, so games are automatically taken care of

I went corsair 16gig 1600 and I got 7950 vapour X.
Just got home with it all.

It is not that necessary to get a 16gb Ram. Know your storage need and you can decide. It might be bigger than 16gb if you are an user enthusiast.

sheesh I went down from 6 to 4 because of a bad stick or bad slot or something and I haven't noticed a difference at all.

I'm building new soon so I'll probably get 8, but anything above 8 is why bother territory for the next several years.

GrimDanfango wrote:

True, for pure gaming, 8GB is more than you'll need for now. Most games haven't even made the transition to 64-bit yet, so as I understand it they can't even address larger than 4GB as a single process. Unless you're planning to run two games at once, 8GB is fine.

64 bit gaming is due this year from EA I've heard, though I don't know if I believe it...we'll probably find out at E3 what's going on. Warframe and WoW have a 64 bit client...don't know what the actual usage pattern is for them though.

Won't see major uptake in 64 bit gaming on PC until the PS4 has been out a bit IMO.

At this point it seems like only 4GB is necessary in order to game. Obviously more is desirable but this seems to be what's needed based on what I've observed and reading through the recommendations for the latest games. At this point I would say the minimum anyone should get is 8GB because it allows for running other programs (browser, email, etc.) while gaming and it gives some cushion for new games that may use more memory. It also makes sense to get 4GB sticks instead of 2GB as that allows for more memory to be added in the future (assuming you have 4 slots for memory.)

As others have said anyone that has the money to spare should consider getting more memory as that allows for more caching even if the memory isn't needed for applications.

Savyg wrote:

64 bit gaming is due this year from EA I've heard, though I don't know if I believe it...we'll probably find out at E3 what's going on. Warframe and WoW have a 64 bit client...don't know what the actual usage pattern is for them though.

Won't see major uptake in 64 bit gaming on PC until the PS4 has been out a bit IMO.

I haven't seen Warframe use over ~3GB of RAM, oddly. Then again, the whole game is only like 4GB. Hehe.

32bit Windows apps give a 2GB/2GB app/kernel split to the mem map. That was one of the awesome things about going X64, you can give different "complete" 4GB memory maps to different apps.
Using >2GB for one app means you're on X64 or are running a bad and poorly documented Windows kernel option...

I had a point. I'll go find it.

Forge wrote:

32bit Windows apps give a 2GB/2GB app/kernel split to the mem map. That was one of the awesome things about going X64, you can give different "complete" 4GB memory maps to different apps.
Using >2GB for one app means you're on X64 or are running a bad and poorly documented Windows kernel option...

I had a point. I'll go find it.

I believe you are referring to the "Large address space aware" flag that you can pull on 32-bit applications in Windows x64. It will allow the 32-bit application to utilize up to 4GB per process instead of the normal 2GB limit.

Krogoth wrote:

Forge wrote:

Using >2GB for one app means you're on X64 or are running a bad and poorly documented Windows kernel option...

I believe you are referring to the "Large address space aware" flag that you can pull on 32-bit applications in Windows x64. It will allow the 32-bit application to utilize up to 4GB per process instead of the normal 2GB limit.

"Kernel option" makes me think he's referring to PAE, rather than LAA, although PAE is neither "bad" nor "poorly documented". PAE isn't that useful for a single app, though, since as you mentioned Krogoth, a single app can only use 3GB per process (with LAA; 4GB is only for 32-bit LAA apps on 64-bit OS.)

MadManOriginal wrote:

The 8GB I have now is that sweet Samsung low profile/low voltage that also overclocks like mad. The mobo is limited to DDR3-1600 though (at least without XMP) so I have it running DDR3-1600 CAS8 at 1.40V.

I love this RAM. I think more people should consider it instead of the "gaming" RAM even though it lacks ersatz heat sinks.

auxy wrote:

Krogoth wrote:

Forge wrote:

Using >2GB for one app means you're on X64 or are running a bad and poorly documented Windows kernel option...

I believe you are referring to the "Large address space aware" flag that you can pull on 32-bit applications in Windows x64. It will allow the 32-bit application to utilize up to 4GB per process instead of the normal 2GB limit.

"Kernel option" makes me think he's referring to PAE, rather than LAA, although PAE is neither "bad" nor "poorly documented". PAE isn't that useful for a single app, though, since as you mentioned Krogoth, a single app can only use 3GB per process (with LAA; 4GB is only for 32-bit LAA apps on 64-bit OS.)

Actually, I meant 3GB. http://goo.gl/7b3Bb PAE is wonderful, but Microsoft didn't want to support it on desktop OSes since it would cannibalize the push to x64. 3GB is poorly-documented and very-rarely-supported kernel switch that changes the default 2GB/2GB split on 32bit Windows to a 3GB/1GB one. It seems to be Server-only, too, which I had forgotten.

I was rather unclear, I apologize, I should know that forum posts and mobile devices don't usually mix well.

The forum kept mangling the URL, so I had to shorten it. The shortened link points here: http://technet.microsoft.com/en-us/libr ... v=exchg.65).aspx

Forge wrote:

PAE is wonderful, but Microsoft didn't want to support it on desktop OSes since it would cannibalize the push to x64.

As I recall it actually was a driver thing - some drivers did not like PAE at all and crashiness ensued. It wasn't an issue if you didn't have those drivers, but it would've impacted too many people to enable it.

(Not that XP was ever the pinnacle of stability heh.)

Plus it also limited the kernel space for drivers and OS processes to 1GB instead of 2 or something? I dunno.

PAE is dead get over it. It was a hack done out of necessity when 64-bit hardware didn't exist in x86 realm and high-end enterprise servers needed more than 4GB of memory. x86-64 made it completely obsolete for the x86 crowd. Microsoft and *nix crowd moved on with the times and dropped it once x86-64 hardware became commonplace on the enterprise-front.

PAE required drivers that were aware of it (limited to enterprise-level hardware that used it) and you took a performance penalty on its usage (5-15% CPU utilization back then, probably 1-2% on current CPUs). There's no point in using it these days. Get x64 hardware and OS if you want to use more than 3-4GB of memory. The fastest, purely 32-bit hardware for x86 was first-generation Prescotts (Pentium 4) and Barton-based Athlon XPs.

Krogoth wrote:

PAE is dead get over it.

Well, ironically, I think Win8s 32 bit ver uses it http://windows.microsoft.com/en-us/wind ... ae-nx-sse2

...I'm not actually sure.

Krogoth wrote:

PAE is dead get over it. It was a hack done out of necessity when 64-bit hardware didn't exist in x86 realm and high-end enterprise servers needed more than 4GB of memory. x86-64 made it completely obsolete for the x86 crowd. Microsoft and *nix crowd moved on with the times and dropped it once x86-64 hardware became commonplace on the enterprise-front.

You're only half right.

Linux distros (and Windows) are still available in 32-bit variants, and in many cases these kernels actually require PAE support from the CPU. Maybe you're thinking of Address Windowing Extensions, which was a software mechanism for exposing PAE at the application level?

Furthermore, the pre-existing PAE mechanism was used (with some extensions) under the hood to implement x86-64. So far from being obsolete, it is in fact more entrenched than ever (albeit better hidden).

In a nutshell, PAE lives on, and will probably continue to do so as long as x86-64 is around... but pretty much everyone who had a legitimate need for AWE has migrated to native 64-bit OSes by now.

Krogoth wrote:

PAE is dead get over it. It was a hack done out of necessity when 64-bit hardware didn't exist in x86 realm and high-end enterprise servers needed more than 4GB of memory. x86-64 made it completely obsolete for the x86 crowd. Microsoft and *nix crowd moved on with the times and dropped it once x86-64 hardware became commonplace on the enterprise-front.

PAE required drivers that were aware of it (limited to enterprise-level hardware that used it) and you took a performance penalty on its usage (5-15% CPU utilization back then, probably 1-2% on current CPUs). There's no point in using it these days. Get x64 hardware and OS if you want to use more than 3-4GB of memory. The fastest, purely 32-bit hardware for x86 was first-generation Prescotts (Pentium 4) and Barton-based Athlon XPs.

While I agree with your main statement ("x64 is the future"), I think that's the only thing you've said that I would agree with. As JBI points out, PAE provided a lot of the initial framework of AMD64, and it's alive and well in 32bit distros and 32bit versions of Windows. Saying it is "dead" is horribly uninformed at best.

MadManOriginal wrote:

The 8GB I have now is that sweet Samsung low profile/low voltage that also overclocks like mad. The mobo is limited to DDR3-1600 though (at least without XMP) :( so I have it running DDR3-1600 CAS8 at 1.40V.

Nice! I run G.SKILL ARES, which is cheap stuff rated at DDR3-1866 1.5v CL9; mine seems to run fine at 1.4v! I prefer to run it at 1.55v and CL9 1T tho.

kumori wrote:

I love this RAM. I think more people should consider it instead of the "gaming" RAM even though it lacks ersatz heat sinks.

Hehe, the ARES does have pretty heatsinks, but they're really small; they barely add to the size of the RAM itself. Mine are blue, but they come in a variety of colors. (The blue matches my Gigabyte mainboard nicely!)
_{I was going to make a snarky remark about mis-using the word ersatz, but it turns out I'm the one who didn't realize what it meant. "You keep using that word ... it doesn't mean what I think it means!"}

auxy wrote:

_{I was going to make a snarky remark about mis-using the word ersatz, but it turns out I'm the one who didn't realize what it meant. "You keep using that word ... it doesn't mean what I think it means!"}

Congratulations, you've satisfied your "try to learn something new each day" quota for today.

(Other than "Hey, 25 year old over-the-stove microwave ovens are mounted really different than newer ones!" I don't think I've satisfied my own quota yet...)

auxy wrote:

"You keep using that word ... it doesn't mean what I think it means!"[/sub]

You keep using that word... I do not think it means what you think it means. [/Inigo.Montoyo]

Forge wrote:

You keep using that word... I do not think it means what you think it means. [/Inigo.Montoyo]

thats_the_joke.jpg (*￣Oﾉ￣*)

Why not use the extra RAM to create a RAM disk? You can find an excellent RAM disk driver here:

http://www.softperfect.com/products/ramdisk/

It used to cost quite a lot but in the most recent version 3.3.3 they have made it freeware.

You can dump your virtual memory there and also your browser cache and other temporary files.

No more mysterious problems created by a corrupted <pagefile.sys> is one of the main benefits I have found.

Nec_V20 wrote:

You can dump your virtual memory there and also your browser cache and other temporary files.

No more mysterious problems created by a corrupted <pagefile.sys> is one of the main benefits I have found.

I don't think I've ever seen (that I'm aware of) "mysterious problems caused by a corrupted pagefile.sys".

IMO the only way it makes any sense to put your pagefile in a RAMdisk is if you're stuck running a 32-bit OS, and can put the RAMdisk in RAM above 4GB using PAE/AWE. On a 64-bit OS (or 32-bit OS with the RAMdisk located in the lower 4GB) you're better off having the extra RAM in the free pool instead of creating a RAMdisk out of it and putting the pagefile there.

I don't think I've ever seen (that I'm aware of) "mysterious problems caused by a corrupted pagefile.sys".

Ah my young Padawan, then met let me enlighten you.

I met an acquaintance of mine in the pub and he told me that the sound was gone on his laptop. He had reinstalled the drivers (and of course rebooted) and checked to make sure he had the latest drivers - so he had done the right things and yet he still had no sound. He had taken it to a computer shop and they had quoted him £40 just to look at it. I told him to get me a pint and I would have a look at it.

Of course I checked to see if he had the volume turned down to mute - which made him feel a bit silly because that was the one thing he had not checked. But the volume was turned up.

So I said to him, "OK, this is the first thing I always do, it doesn't always clear problems but often enough it does". So I went to the virtual memory, changed its settings (to create a fresh <pagefil.sys>), and rebooted. Windows started up again with the introductory sound.

The reason why this one sticks in my mind is because I said to him, "OK, let's see if the sound is working", and he replied "Of course it is", and I said, "How do you know?".

I've booted so many computers so many times I don't even hear that sound any more. A bit like the 24 needle printer (Star LC 24-10) I used to have where, after a few months of using it, I would not notice it when it was printing - loved that printer.

I've been a computer techie for 30 years and every day I find out something new that I have not seen or been aware of before, it's why I enjoy it. Of course because you have not heard of this before you will probably have spent a lot of time in the past going around in circles with regard to a problem where just eliminating and recreating the virtual memory file would have cured it.

All I can say to that is, just one more reason I'm glad I don't use Windows as my primary OS any more. In any sane OS, the state of the pagefile from a previous boot should have ZERO effect on system operation.

And I'm probably your age (if not older); been doing PC stuff since the Altair/IMSAI days.

Not sure why you're blaming the Windows for this and accusing the OS of not being sane.

Windows does not page drivers by default. The driver writer must choose to do that.

Neither you nor our story teller debugged the system in question. You can't even be sure mucking with the page file actually was the fix. It certainly could have been one big coincidence.

Next time some software misbehaves on Linux let me roll into the forums and accuse it of insanity.

OK, how about this instead: If the story is true, and some piece of leftover wreckage in the pagefile from a previous boot really did cause the soundcard to malfunction, then that's insane. Better?

Edit: Linux certainly has aspects that drive me nuts too. The whole desktop environment ecosystem for the past couple of years, for example (pretty sure I've ranted about this before). But it tends to get the underling infrastructure right.

just brew it! wrote:

Linux certainly has aspects that drive me nuts too. The whole desktop environment ecosystem for the past couple of years, for example (pretty sure I've ranted about this before). But it tends to get the underling infrastructure right.

OK. What is a desktop ecosystem? If you mean the various, what are they called, IDE ... hmmm integrated desktop environments, then they have very little to do with Linux. I imagine they are an effort to make windose users happier with Linux.

All I need from a Linux desktop is someplace to run whatever I might want in a window system. Long ago I fooled with Gnome and KDE and was mildly horrified by the concepts behind them as they had very little to do with the *nix way of doing things. Xfce and these days fluxbox do the job just fine. What could I possibly need an IDE for?

I guess it is cheap but it's not cheap enough for me to have 32BG's.

My Johnson wrote:

I guess it is cheap but it's not cheap enough for me to have 32BG's.

You can't budget 200$? Bummer. I did, and it's pretty awesome.