There is something amazing to be said about consumer-class 16GB high-performance modules. Unfortunately almost nothing in the consumer market currently needs more than 8GB of RAM...Windows system requirements have been on a decline every generation since Vista. Windows 8.1 runs on 1GB of RAM, a Pentium 3 and 915E graphics, and as long as you have an SSD, it's actually usable. Those are spec's from the 90's.
Virtual machines. Lots and lots of virtual machines. I have two VMs I keep up and running at all times on my main system, one running Ubuntu and one running OpenSUSE for secure web surfing. Each take 8GB of memory from the host OS. I run 32GB. I can run WoW on one screen while keeping the web surfing VMs up on the other.
The OS doesn't eat more memory; but applications can if you go beyond light usage.
I've regularly had 50-150 tabs open across two browsers for years; a half dozen years ago each browser regularly started around 500MB or ram and gradually fragmented it's heap to about 1GB when I restarted it. Now, they weigh in around 1GB each at startup and don't need restarted until they get above 3GB (where presumed heap fragmentation and/or frantic but futile garbage collection big them down). The memory footprint of Einstein@Home applications keeps going up as well (trading more memory for more complex calculations that increase the odds of a successful detection); they've gone from <100MB per core to ~200-400MB per core/gpu now (some are more resource intensive than others). Between those, the rest of the stuff my computer is running, and various games I started thumping into swap a few years ago when I had 12GB of ram on my i7-920; I dusted off my older set of 3x2GB dimms took the timing hit to plug them back in to get to 18GB. I haven't crossed that limit yet (or at least not by enough to cause noticeable UI lag); but it is a driving force behind my next box being specced out at 32GB.
You may want to invest in a 64-bit browser then. Physical memory does not get fragmented, since it is paged based, and thus can be allocated by a pool allocator on demand. The virtual address space, on the other hand, does get fragmented. On a 32-bit system you will definitely start seeing issues around 2-3GB of use. A 64-bit application should have close to no issues with this. If you are seeing any slowdown by approaching your physical commit limit (e.g. 3GB on a 4GB machine) you are likely seeing the effects of hitting page faults and paging data in from disk.
I have 32GB DDR4 and Windows 8.1 and even running 10 concurrent 1080P video encodes (using x264 on placebo mode), 30+ Chrome tabs, Steam, and Visual Studio 2013 I have never been able to hit >16GB RAM usage.
Other than VMs and astonishingly memory-hungry content creation apps -- both of which are used by a vanishingly small percentage of the PC market -- I don't know how much demand will exist for such large RAM sizes. There was a time that more RAM was always better because the need was just around the corner. While >8GB will eventually be required, the growth curve has been flat for as long as any time in history.
If Microsoft ever implemented OS level support for RAM disks and made it consumer accessible, that would be a pretty useful. Or find a way for the OS to more intelligently use the RAM as a disk cache to bypass disk access completely for commonly used applications, which I suppose could theoretically provide power savings on mobile devices.
RAMDisk is part of Windows since forever. You can add RAMDisk driver. It is though GUI-less and parameters re edited through registry. More info and source code (it is very simple driver): https://code.msdn.microsoft.com/windowshardware/RA...
Yeah, SkyLake is the first significant upgrade for Sandy Bridge users. My 2500k + GTX 670 FTW box can still manage just about everything I need at 1080p, really looking forward to the step up to PCIe SSDs and DDR4.
Agreed, a lot is riding on Skylake. I have an i7 2600 and very few things really peg the CPU, and the new equipment (outside of the iGPU which goes unused) is only better for very specific tasks such as encryption. I am to the point where I want to upgrade... but an upgrade would be largely useless at this point. Even my old-ish GTX570 can keep up with most things today (at least for 1080p) and is really only held back by only having 1GB of RAM on board rather than any sort of processing bottleneck. Skylake + PCIe SSDs + DDR4 + a newer GPU might actually make a case for an upgrade. Hopefully it will be faster, offer more cores, and have a significantly lower idle power use for when I am not using it but still have to have it on. But if the new chips offer the same thing the last few generations have offered with the same horsepower while using less wattage then I may just upgrade the GPU and be done with it.
According to the recent years Intel's "tradition", Skylake won't bring any impressive x86 performance improvements. Another 5%, I guess, not taking new specific Skylake intstructions into account (correct me if I'm wrong). Besides, Skylake is initially rumoured to be locked (non-overclockable). So, I have a big doubt it will be worth upgrading from i7-2600 (at least i7-2600K) from purely x86 performance point of view. I personally have Haswell i7-4790K and plan to use it for three years more or whenever the M/B dies of ageing in the future, because I believe an upgrade to a newer platform in the next 3 years won't be really worth the money.
If my budget allows it, I am considering Skylake, but honestly for my server, which is running a G1610 on a H67 chipset motherboard. More for the features than anything, but I am also hoping with a newer chipset and newer processor, that it might just be able to push down my idle power a couple more watts, while also leveraging a lot more "new stuff". For instance, having more than 2 6Gbps SATA ports would be nice, as I can't run my 2 disk RAID0 array and my boot SSD all on 6Gbps, which is an inconvenience (sure, the HDDs themselves can't saturate SATAII, but the DRAM caches on them CAN, so there could be a small gain if the array could live on the 6Gbps ports). Maybe/possible switching up that boot SATAIII 60GB SSD for something like a PCIe based m.2 60-64GB SSD could be nice too. Then 2x4GB of DDR4 or even 2x8GB of DDR4 for lower power and higher bandwidth would be nice.
Lastly, the low end of the core line has actually gotten some pretty reasonable improvements if you look at Sandy through Haswell and sadly Broadwell is getting skipped. So even if I stuck with a Celeron Skylake CPU, odds are decent I could get a reasonably good bump in CPU performance (maybe 10-15%) as well as a decent bump in GPU performance (which on a mostly headless server, isn't terribly important, but as the GPU DOES get leverage for some tasks, doesn't hurt). That and I may finally jump up to a higher level chip than a Celeron. Depending on price/performance/power consumption I may finally look at a low end i3...and who knows, maybe by Skylake we'll see quad core get pushed in to i3 territory with dual core being a Celeron and Pentium thing, with i5s gaining hyperthreading and i7's gaining hexacore. One can hope/dream.
It would be nice to see i3 having both hyperthreading AND turboboost, it would make my decision that much easier.
My desktop with its i5-3570@4GHz really doesn't leave too much on the table. I can't see upgrading that until maybe Broadwell-E or more likely Skylake-E hits the market, as I'd like a CLEAR increase in performance, though if the regular i7's gain hexacore with Skylake, that might be an excuse, I mean a reason, to upgrade. As it stands, I know new technologies come along all the time, but I am thinking/hoping for USB3.1 in the chipset natively the itteration after Skylake. It isn't crucial to have it native in my desktop (and I'd imagine by the time the 10 series chipsets hit, board manufacturers will have lots of add-on controllers for 2-4 ports of USB3.1), but it would be nice to have.
THAT's an interesting idea! One way Intel could try to juice up the relevance of Skylake in world where Sandy and successors are already pretty damned powerful and (for most purposes) low enough power might be to switch from the current 2 core/4 core devices for mobile/desktop to something like i3=3 cores, i5=4 cores, i7=6 cores... Marketing might love that as well, leaving 2 cores for the Celeron/Pentium really low-end stuff.
I'm running an almost 5 year old core i7-980x at 4.2ghz and i skipped upgrading to haswell-e in favor of getting the 2nd generation 55" oled from LG for 3000 dollars. Even after 5 years my pc is still not slow enough to upgrade. The only reason I have to upgrade is for additional features that weren't part of the x58 chipset. The 8 core would def boost my video encoding but is it really worth it to spend over 2000 on a new cpu mobo and expensive ddr4 ram just to shave a few minutes off encoding? Broadwell-e is going to be a boring 14nm shrink of haswell-e so i'll be sitting tight till skylake-e, by that time skylake will have made ddr4 more mainstream and prices should be a lot better for ddr4 and having pci-e 4.0 will be a nice reason to go for it and usb 3.1 should be in the chipset for skylake-e. Maybe all the pci-e 2.0 will get upgraded to 3.0 once 4.0 is out which would be a nice bonus too bumping sata express to 16gb/s instead of 10.
Because the standards everything will be running on are so different I will need new ssd's and new gpu so I will probably turn my current pc into a super powerful nas rather than re use the case and psu since i will have to buy new parts for everything else basically it wont save much money and will be more useful keeping it together.
5% is impressive for a yearly or even bi-yearly update, heck, even 2% is impressive. What you should do is compare 3 years worth of innovation looks like. Compare Skylake to an Intel CPU 3 years ago and you'll be impressed of how far they came.
We're never going to see any more than 3-5% for a single CPU generation again, we're reaching the limits of how much we can pull from the current technologies and it is going to take longer and longer to scale down the nm processes.
I know and I don't complain, actually. For example, with no more than 4 x86 instructions-wide decoder, found in all Intel Core designs, this situation of slow recent improvements with Intel's x86 performance per clock was unofficially named "the limit of Intel's superscalar" by. e.g., Boris Babaian, who was to do with Intel himself in the last 10 years. In three generations, however, 1.05^3 = 1.16, which is worth considering.
I have to disagree that 5% is in any way impressive. Back in the 90s if your computer was 2-3 years old it was probably half as fast as a new machine. Obviously those days are long behind us now, but even so, seeing a 20% improvement over the course of like 4 years is pretty lame.
The newer processors aren't fast enough for many of us to even justify upgrading, especially when games tend to rely so heavily on GPUs nowadays. I've been running my i5 2500K since the end of 2011. I ran it for two years at stock and then cranked up the overclock to 4.6 GHz, which put it close to being on par with newer processors.
I don't see myself upgrading my CPU for another two or three years. There's just no point. Who knows, maybe by the time I'm ready to upgrade AMD will be back in the game (yeah, I know, it's a long shot).
Core i7 3770K @4.2 GHZ when I bought it new. I've had the CPU pegged at 100% utilization since. I only have 16GB (2x 8GB I knew it wouldn't be enough when I bought it, so I was prepared for the upgrade) of memory and after 3 tabs in IE 9, 50-75 in Chrome, Vuze, Steam, World Community Grid and 1 modern AAA game I'm using somewhere between 12-15+ GB of memory. Sometimes I even hit 16GB plus and fell the computer paging things in and out on the HD. I'm hitting the ceiling now. I need 32GB right now, and I want 64GB. I'd like to keep a 4GB VM open all the time and have the option to run some content creation stuff without having to kill the game. There's a program available now that will load a game to a ram disk before play and unload it when you're done. X99 and the new processor architectures are appealing for the increased IPC, but what keeps me away from the upgrade are ram costs. I figure DDR4 needs another 18 months on the market for the prices to become reasonable.
Yeah at those costs I'm not upgrading any time soon, I recognize that the $60 or so I paid for my 16GB (4x4) of DDR3 bordered on ridiculous, but I haven't paid over half a grand for RAM in over a decade... I'm sure prices will come down, I'm just doubtful how fast that'll be with desktops in continued decline.
I am not sure what all you are listing as running concurrently but I routinely have 2 XP virtual machines (VMs) running using VMPlayer, Firefox with a 20-30 tabs open including Netflix screen and 4 browser games that appear to have problems with releasing back memory, Dungeon and Dragons Online running in ultra resolution (I stay logged in most of time and while waiting for quest party formation), Windows Media Player playing soft music playing, and either WINTV7 (Hauppauge software for the tv-tuner) or Windows Media Center (again for the tv-tuner, for some reason I cannot get channel 30.3 using WINTV7). I rarely exceed 8 GB used and only when I am running my big VM's or a lot of them or load DDO directory into a RAMdisk have I ever exceeded 18GB (DDO on a RAMdisk takes 12GB). Right now I have all those except DDO (so no RAMdisk either) and I am only using 6GB (2x1GB for each copy of XP in current execution).
I do not know what Steam or World Community Grid are but they must be extreme memory pigs.
I run only a 8MB pagefile.sys. That is only for those few programs that think they actually need to access the file. I have 32GB memory.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
31 Comments
Back to Article
Antronman - Tuesday, January 13, 2015 - link
16GB modules?Drool worthy.
imaheadcase - Tuesday, January 13, 2015 - link
More like Money draining worthy. heheSamus - Tuesday, January 13, 2015 - link
There is something amazing to be said about consumer-class 16GB high-performance modules. Unfortunately almost nothing in the consumer market currently needs more than 8GB of RAM...Windows system requirements have been on a decline every generation since Vista. Windows 8.1 runs on 1GB of RAM, a Pentium 3 and 915E graphics, and as long as you have an SSD, it's actually usable. Those are spec's from the 90's.TiGr1982 - Tuesday, January 13, 2015 - link
Not exactly - 1 GB of RAM became common a little bit later on desktops - around 2003 :)dgingeri - Tuesday, January 13, 2015 - link
Virtual machines. Lots and lots of virtual machines. I have two VMs I keep up and running at all times on my main system, one running Ubuntu and one running OpenSUSE for secure web surfing. Each take 8GB of memory from the host OS. I run 32GB. I can run WoW on one screen while keeping the web surfing VMs up on the other.TiGr1982 - Tuesday, January 13, 2015 - link
Not even only virtual machines; RAMdisks and "heavy" professional applications are another reasons for having plenty of RAM.DanNeely - Tuesday, January 13, 2015 - link
The OS doesn't eat more memory; but applications can if you go beyond light usage.I've regularly had 50-150 tabs open across two browsers for years; a half dozen years ago each browser regularly started around 500MB or ram and gradually fragmented it's heap to about 1GB when I restarted it. Now, they weigh in around 1GB each at startup and don't need restarted until they get above 3GB (where presumed heap fragmentation and/or frantic but futile garbage collection big them down). The memory footprint of Einstein@Home applications keeps going up as well (trading more memory for more complex calculations that increase the odds of a successful detection); they've gone from <100MB per core to ~200-400MB per core/gpu now (some are more resource intensive than others). Between those, the rest of the stuff my computer is running, and various games I started thumping into swap a few years ago when I had 12GB of ram on my i7-920; I dusted off my older set of 3x2GB dimms took the timing hit to plug them back in to get to 18GB. I haven't crossed that limit yet (or at least not by enough to cause noticeable UI lag); but it is a driving force behind my next box being specced out at 32GB.
inighthawki - Tuesday, January 13, 2015 - link
You may want to invest in a 64-bit browser then. Physical memory does not get fragmented, since it is paged based, and thus can be allocated by a pool allocator on demand. The virtual address space, on the other hand, does get fragmented. On a 32-bit system you will definitely start seeing issues around 2-3GB of use. A 64-bit application should have close to no issues with this. If you are seeing any slowdown by approaching your physical commit limit (e.g. 3GB on a 4GB machine) you are likely seeing the effects of hitting page faults and paging data in from disk.DanNeely - Wednesday, January 14, 2015 - link
If Mozilla ever gets around to releasing a win64 build...anonymous_user - Wednesday, January 14, 2015 - link
There are third-party builds if you want a 64-bit Firefox.Sivar - Tuesday, January 13, 2015 - link
I have 32GB DDR4 and Windows 8.1 and even running 10 concurrent 1080P video encodes (using x264 on placebo mode), 30+ Chrome tabs, Steam, and Visual Studio 2013 I have never been able to hit >16GB RAM usage.Other than VMs and astonishingly memory-hungry content creation apps -- both of which are used by a vanishingly small percentage of the PC market -- I don't know how much demand will exist for such large RAM sizes.
There was a time that more RAM was always better because the need was just around the corner. While >8GB will eventually be required, the growth curve has been flat for as long as any time in history.
inighthawki - Tuesday, January 13, 2015 - link
If Microsoft ever implemented OS level support for RAM disks and made it consumer accessible, that would be a pretty useful. Or find a way for the OS to more intelligently use the RAM as a disk cache to bypass disk access completely for commonly used applications, which I suppose could theoretically provide power savings on mobile devices.Klimax - Wednesday, January 14, 2015 - link
RAMDisk is part of Windows since forever. You can add RAMDisk driver. It is though GUI-less and parameters re edited through registry. More info and source code (it is very simple driver): https://code.msdn.microsoft.com/windowshardware/RA...inighthawki - Wednesday, January 14, 2015 - link
Very interesting, didn't know about this. Thanks for sharing!atticus14 - Wednesday, January 14, 2015 - link
So while I don't count as typical consumer the amount of Ram Chrome takes up...I think I can use a few of those.bitech - Wednesday, January 14, 2015 - link
Why would you want RAM requirements to rise? What good would it do to make consumers require more than 8gb RAM?Flunk - Tuesday, January 13, 2015 - link
Cool, looks like we're getting close to a reason to upgrade my 2500K. Sky Lake or AMD's next-gen chip seem interesting and they both use DDR4.foxtrot1_1 - Tuesday, January 13, 2015 - link
Yeah, SkyLake is the first significant upgrade for Sandy Bridge users. My 2500k + GTX 670 FTW box can still manage just about everything I need at 1080p, really looking forward to the step up to PCIe SSDs and DDR4.CaedenV - Tuesday, January 13, 2015 - link
Agreed, a lot is riding on Skylake. I have an i7 2600 and very few things really peg the CPU, and the new equipment (outside of the iGPU which goes unused) is only better for very specific tasks such as encryption. I am to the point where I want to upgrade... but an upgrade would be largely useless at this point.Even my old-ish GTX570 can keep up with most things today (at least for 1080p) and is really only held back by only having 1GB of RAM on board rather than any sort of processing bottleneck.
Skylake + PCIe SSDs + DDR4 + a newer GPU might actually make a case for an upgrade. Hopefully it will be faster, offer more cores, and have a significantly lower idle power use for when I am not using it but still have to have it on. But if the new chips offer the same thing the last few generations have offered with the same horsepower while using less wattage then I may just upgrade the GPU and be done with it.
TiGr1982 - Tuesday, January 13, 2015 - link
According to the recent years Intel's "tradition", Skylake won't bring any impressive x86 performance improvements. Another 5%, I guess, not taking new specific Skylake intstructions into account (correct me if I'm wrong).Besides, Skylake is initially rumoured to be locked (non-overclockable). So, I have a big doubt it will be worth upgrading from i7-2600 (at least i7-2600K) from purely x86 performance point of view.
I personally have Haswell i7-4790K and plan to use it for three years more or whenever the M/B dies of ageing in the future, because I believe an upgrade to a newer platform in the next 3 years won't be really worth the money.
azazel1024 - Tuesday, January 13, 2015 - link
If my budget allows it, I am considering Skylake, but honestly for my server, which is running a G1610 on a H67 chipset motherboard. More for the features than anything, but I am also hoping with a newer chipset and newer processor, that it might just be able to push down my idle power a couple more watts, while also leveraging a lot more "new stuff". For instance, having more than 2 6Gbps SATA ports would be nice, as I can't run my 2 disk RAID0 array and my boot SSD all on 6Gbps, which is an inconvenience (sure, the HDDs themselves can't saturate SATAII, but the DRAM caches on them CAN, so there could be a small gain if the array could live on the 6Gbps ports). Maybe/possible switching up that boot SATAIII 60GB SSD for something like a PCIe based m.2 60-64GB SSD could be nice too. Then 2x4GB of DDR4 or even 2x8GB of DDR4 for lower power and higher bandwidth would be nice.Lastly, the low end of the core line has actually gotten some pretty reasonable improvements if you look at Sandy through Haswell and sadly Broadwell is getting skipped. So even if I stuck with a Celeron Skylake CPU, odds are decent I could get a reasonably good bump in CPU performance (maybe 10-15%) as well as a decent bump in GPU performance (which on a mostly headless server, isn't terribly important, but as the GPU DOES get leverage for some tasks, doesn't hurt). That and I may finally jump up to a higher level chip than a Celeron. Depending on price/performance/power consumption I may finally look at a low end i3...and who knows, maybe by Skylake we'll see quad core get pushed in to i3 territory with dual core being a Celeron and Pentium thing, with i5s gaining hyperthreading and i7's gaining hexacore. One can hope/dream.
It would be nice to see i3 having both hyperthreading AND turboboost, it would make my decision that much easier.
My desktop with its i5-3570@4GHz really doesn't leave too much on the table. I can't see upgrading that until maybe Broadwell-E or more likely Skylake-E hits the market, as I'd like a CLEAR increase in performance, though if the regular i7's gain hexacore with Skylake, that might be an excuse, I mean a reason, to upgrade. As it stands, I know new technologies come along all the time, but I am thinking/hoping for USB3.1 in the chipset natively the itteration after Skylake. It isn't crucial to have it native in my desktop (and I'd imagine by the time the 10 series chipsets hit, board manufacturers will have lots of add-on controllers for 2-4 ports of USB3.1), but it would be nice to have.
name99 - Tuesday, January 13, 2015 - link
THAT's an interesting idea!One way Intel could try to juice up the relevance of Skylake in world where Sandy and successors are already pretty damned powerful and (for most purposes) low enough power might be to switch from the current 2 core/4 core devices for mobile/desktop to something like
i3=3 cores, i5=4 cores, i7=6 cores...
Marketing might love that as well, leaving 2 cores for the Celeron/Pentium really low-end stuff.
Laststop311 - Tuesday, January 13, 2015 - link
I'm running an almost 5 year old core i7-980x at 4.2ghz and i skipped upgrading to haswell-e in favor of getting the 2nd generation 55" oled from LG for 3000 dollars. Even after 5 years my pc is still not slow enough to upgrade. The only reason I have to upgrade is for additional features that weren't part of the x58 chipset. The 8 core would def boost my video encoding but is it really worth it to spend over 2000 on a new cpu mobo and expensive ddr4 ram just to shave a few minutes off encoding? Broadwell-e is going to be a boring 14nm shrink of haswell-e so i'll be sitting tight till skylake-e, by that time skylake will have made ddr4 more mainstream and prices should be a lot better for ddr4 and having pci-e 4.0 will be a nice reason to go for it and usb 3.1 should be in the chipset for skylake-e. Maybe all the pci-e 2.0 will get upgraded to 3.0 once 4.0 is out which would be a nice bonus too bumping sata express to 16gb/s instead of 10.Because the standards everything will be running on are so different I will need new ssd's and new gpu so I will probably turn my current pc into a super powerful nas rather than re use the case and psu since i will have to buy new parts for everything else basically it wont save much money and will be more useful keeping it together.
MikhailT - Tuesday, January 13, 2015 - link
5% is impressive for a yearly or even bi-yearly update, heck, even 2% is impressive. What you should do is compare 3 years worth of innovation looks like. Compare Skylake to an Intel CPU 3 years ago and you'll be impressed of how far they came.We're never going to see any more than 3-5% for a single CPU generation again, we're reaching the limits of how much we can pull from the current technologies and it is going to take longer and longer to scale down the nm processes.
TiGr1982 - Tuesday, January 13, 2015 - link
I know and I don't complain, actually. For example, with no more than 4 x86 instructions-wide decoder, found in all Intel Core designs, this situation of slow recent improvements with Intel's x86 performance per clock was unofficially named "the limit of Intel's superscalar" by. e.g., Boris Babaian, who was to do with Intel himself in the last 10 years.In three generations, however, 1.05^3 = 1.16, which is worth considering.
Nagorak - Friday, January 30, 2015 - link
A 16% improvement is next to nothing. Not even remotely worth considering upgrading over.Nagorak - Friday, January 30, 2015 - link
I have to disagree that 5% is in any way impressive. Back in the 90s if your computer was 2-3 years old it was probably half as fast as a new machine. Obviously those days are long behind us now, but even so, seeing a 20% improvement over the course of like 4 years is pretty lame.The newer processors aren't fast enough for many of us to even justify upgrading, especially when games tend to rely so heavily on GPUs nowadays. I've been running my i5 2500K since the end of 2011. I ran it for two years at stock and then cranked up the overclock to 4.6 GHz, which put it close to being on par with newer processors.
I don't see myself upgrading my CPU for another two or three years. There's just no point. Who knows, maybe by the time I'm ready to upgrade AMD will be back in the game (yeah, I know, it's a long shot).
Aslan7 - Wednesday, January 14, 2015 - link
Core i7 3770K @4.2 GHZ when I bought it new. I've had the CPU pegged at 100% utilization since. I only have 16GB (2x 8GB I knew it wouldn't be enough when I bought it, so I was prepared for the upgrade) of memory and after 3 tabs in IE 9, 50-75 in Chrome, Vuze, Steam, World Community Grid and 1 modern AAA game I'm using somewhere between 12-15+ GB of memory. Sometimes I even hit 16GB plus and fell the computer paging things in and out on the HD. I'm hitting the ceiling now. I need 32GB right now, and I want 64GB. I'd like to keep a 4GB VM open all the time and have the option to run some content creation stuff without having to kill the game. There's a program available now that will load a game to a ram disk before play and unload it when you're done. X99 and the new processor architectures are appealing for the increased IPC, but what keeps me away from the upgrade are ram costs. I figure DDR4 needs another 18 months on the market for the prices to become reasonable.Impulses - Wednesday, January 14, 2015 - link
Yeah at those costs I'm not upgrading any time soon, I recognize that the $60 or so I paid for my 16GB (4x4) of DDR3 bordered on ridiculous, but I haven't paid over half a grand for RAM in over a decade... I'm sure prices will come down, I'm just doubtful how fast that'll be with desktops in continued decline.TheinsanegamerN - Friday, January 16, 2015 - link
just wait until ddr4 comes to laptops. that'll be more than enough to push prices down.NatalieEGH - Thursday, March 12, 2015 - link
I am not sure what all you are listing as running concurrently but I routinely have 2 XP virtual machines (VMs) running using VMPlayer, Firefox with a 20-30 tabs open including Netflix screen and 4 browser games that appear to have problems with releasing back memory, Dungeon and Dragons Online running in ultra resolution (I stay logged in most of time and while waiting for quest party formation), Windows Media Player playing soft music playing, and either WINTV7 (Hauppauge software for the tv-tuner) or Windows Media Center (again for the tv-tuner, for some reason I cannot get channel 30.3 using WINTV7). I rarely exceed 8 GB used and only when I am running my big VM's or a lot of them or load DDO directory into a RAMdisk have I ever exceeded 18GB (DDO on a RAMdisk takes 12GB). Right now I have all those except DDO (so no RAMdisk either) and I am only using 6GB (2x1GB for each copy of XP in current execution).I do not know what Steam or World Community Grid are but they must be extreme memory pigs.
I run only a 8MB pagefile.sys. That is only for those few programs that think they actually need to access the file. I have 32GB memory.