Aces High Bulletin Board
General Forums => Hardware and Software => Topic started by: Max on March 16, 2017, 08:14:54 AM
-
I'm running Corsair Vengence 2X4 GB SDRAM DDR3 with Win 7 64 bit.
I've read conflicting reports saying 8 GB is a "sweet spot" for Win 7 while others suggest 16 GB is better suited, especially for gaming. I'm not aware of any problems as of now so the question is, does it make sense to spend $63 for the extra 2X4 GB sticks and if so, what would I gain?
Thanks
-
Just playing Aces High while browsing a few web pages won't put a strain on 8GB with Windows 7. Even with office open too.
Not much bang for your $63 bucks. I wouldn't bother. :cheers:
-
Please check your mainboard manual - with four memory sticks it may be possible it reduces the maximum permissible memory speed. If you upgrade to 16 GiB consider buying 2x 8GiB.
-
I have 16 in my machine and it seems even when editing videos I rarely get much more than 20% load. I don't think I've ever seen 30%. And that was the thinking when I went with 16 over 8. Could have saved a few bucks.
Sent from my XT1585 using Tapatalk
-
Please check your mainboard manual - with four memory sticks it may be possible it reduces the maximum permissible memory speed. If you upgrade to 16 GiB consider buying 2x 8GiB.
With out going and looking it up , isn't it not the other way around?
Say you are using 1333 MHz DDR3 or 1600 MHz DDR3 etc... 4 GB DIMMs (memory sticks) x 4...don't it slow the through put if you switched over to using (2) 8 GB DIMMs instead of using (4) 4 GB DIMMs?.....even though the system memory is of the same brand/speed/GB amount?
or am I thinking of something entirely different?
TC
-
Drano, getting into some heavy video transcoding, audio transcoding or something like using Photoshop for doing commercial graphics, using AutoCAD designing HVAC-R/plumbing, engineering blueprints etc... are some of the things that can and will happen to heavily tax a computer's system Ram....
Even more so in the old days before Windows 7...( WinXP, Win98, Win95, Win for WGs 3.0/3.11 etc ) where we used batch files and used the PAE, etc...
TC
-
As far as performance goes, there are two things to consider when thinking of system RAM.
1) Does the computer have enough RAM to run any given application?
2) Is performance the top priority?
Most of the time as you buy larger and larger memory modules, they tend to get slower and slower. Not talking about the rated clock rate. That is a marketing game. Check the specifications of the memory module and look for the number of wait states, as a start. Wait states can be inserted to help slow RAM to be marketed with fast numbers, but they actually run slower.
So there is some truth to adding more RAM can slow down a computer. I believe that statement has to do more with the memory module performance, than the actual size of the memory module.
TC, a system can perform better with 4) 4GB modules versus 2) 8GB modules, as the *GB modules usually contain much slower RAM chips. It also depends on the memory controller on the motherboard. Some of the older memory controllers were not installed correctly and had capacitance issues when that many memory modules were installed.
My approach has always been to determine the biggest memory hog I run and buy the RAM to meet that application need. No more than is necessary. My home computer has 16GB of system RAM and I do a lot of heavy editing, including 4K video. 16GB is the least amount I would have for my application and it is comprised of 4) 4GB modules.
-
I noticed a big difference when I upgraded from 4 to 8 GB. For everyday use and AH I believe 16 GB wouldn't make such a difference. Unless you're planning to do heavy photot/video editing I wouldn't bother.
-
8GB is more than enough for most people.
-
Haven't seen an issue with 32GB in either the laptops or the desktop :)
-
Haven't seen an issue with 32GB in either the laptops or the desktop :)
Unless you swap it with faster smaller modules, you will not see any difference.
-
Are you talking about modules with better timings, skuzzy?
-
Larger modules are not slower than smaller modules pending they have the same speed rating, timings and internal organization (Single-Rank vs Dual-Rank).
-
Larger modules are not slower than smaller modules pending they have the same speed rating, timings and internal organization (Single-Rank vs Dual-Rank).
They do not have to be, but many are. You can have the same speed rating (the DDR marketing number) and be slower.
You just have to do some math to see which is faster in the application as all the timings are measured in clock ticks. Just divide 1 by the rated speed of the memory, then multiply that time the various timings to get the actual time.
For instance, if the clock of the RAM is rated at 2400 and the CL is 5, versus another stick rated at 3200, with a CL of 10, the 2400 module is going to be faster.
-
They do not have to be, but many are. You can have the same speed rating (the DDR marketing number) and be slower.
You just have to do some math to see which is faster in the application as all the timings are measured in clock ticks. Just divide 1 by the rated speed of the memory, then multiply that time the various timings to get the actual time.
For instance, if the clock of the RAM is rated at 2400 and the CL is 5, versus another stick rated at 3200, with a CL of 10, the 2400 module is going to be faster.
Hi Skuzzy,
So to make sure that I got this, you divide the mem clock speed by the cas latency to arrive at the thruput over time....the higher the mem thruput over time (1 clock tick) the faster the mem module, right?
For example, 2400^5 = 480 per clock tick. 3200^10 = 320 per clock tick so the 2400 frequency w\ CL of 5 is the faster module. For the 3200 mem module to equal the 2400 one the CL has to be at 6.6 or lower than 6.6 to be faster than the 2400 module..........
Or am I missing something here?
Appreciate the info..............
:salute
-
Correct. Keep in mind that is just one value. Hold time is also a consideration. All the values matter in the overall scheme. I'll have to do timeline graph showing when each cycle starts and ends so you can get a better idea of the overall time it takes to read/write one bus width of data.
-
Correct. Keep in mind that is just one value. Hold time is also a consideration. All the values matter in the overall scheme. I'll have to do timeline graph showing when each cycle starts and ends so you can get a better idea of the overall time it takes to read/write one bus width of data.
Ok.
Yeah I was assuming that the clock tick time was parallel on the falling side as well as the rising side. Didn't even factor in the hold time or transition from rising to falling and vise-versa.
Thanks!
:salute
-
Just found a 16Gb 4x4Gb kit of Corsair Vengence LPX DDR4 3000 CL 15 w\ red heat spreaders that is faster than my current 2133 set at a sale price of $129.95 (cheapest price on Newegg....all other brands are more expensive & these are the fastest quad channel 16Gb set out there until you get past 3400 as long as the CL stays at 16 but the price on those is "WHEW!!!!"
I'm probably gonna spring for these......should squeeze a little more performance out of my little box....and give me a little color contrast as well on the ole Gigabyte X99M Gaming 5 mobo............
FYI.............
:salute
-
Anyone doing anything else while playing? If not then 8gb is enough.
-MX-
-
Now I'm wondering if getting 32gigs was to much? After running a 6gig machine for years I felt like more was better. :D
-
I've got 16 Gb DDR4 3200 and rarely exceed 20% usage. 8 Gb should be fine unless you have a lot of apps running at the same time or are into heavy video editing.
-
Now I'm wondering if getting 32gigs was to much? After running a 6gig machine for years I felt like more was better. :D
I was running 4gig which had to cut to 2gig when I was getting all kinds of BSODs.
I've just gone with 2x4gbs for my rebuild. With the intention of adding 2x8gb sticks when the price falls.
-
I was running 4gig which had to cut to 2gig when I was getting all kinds of BSODs.
I've just gone with 2x4gbs for my rebuild. With the intention of adding 2x8gb sticks when the price falls.
My old system was 4x2gig cards but I kept having a massive amount of BSODS, had to cut it down to 3x2gig which sucked for years. When I built this new machine I just decided to cram it to motherboard max of 4x8gig cards. So far I really don't notice and slow down but I'm sure there could be issues. For now I'm happy having 32 as I feel like I can plow through anything.
-
If four memory modules tend to be somewhat unstable it often works to reduce max speed by one grade. After ensuring the command rate is not set to 1T of course (seldomly works with four modules).
In todays gaming machines you want to go for 2x8GiB modules.
-
If four memory modules tend to be somewhat unstable it often works to reduce max speed by one grade. After ensuring the command rate is not set to 1T of course (seldomly works with four modules).
In todays gaming machines you want to go for 2x8GiB modules.
I've yet to run into any issues running (4) 8 GB DIMMs at 1T regardless if it is an Intel build or an AMD build...I can see where it is could be an issue though...if people are monkeying around with the timings/settings, pushing for too much, trying to overclock
TC
-
I just run 4gb and I am content
-
I've yet to run into any issues running (4) 8 GB DIMMs at 1T regardless if it is an Intel build or an AMD build...I can see where it is could be an issue though...if people are monkeying around with the timings/settings, pushing for too much, trying to overclock
TC
Some motherboards are prone to issues using 4 DIMMs. Always a good idea to check the compatibility list at the manufacturer's web-site.
-
Some motherboards are prone to issues using 4 DIMMs. Always a good idea to check the compatibility list at the manufacturer's web-site.
You know me, BE, I always do thorough research before the ordering and build, brother!
Edit: back when I was heavily donating parts, controllers, full builds, I did come across a Gigabyte MB that had issues with the 4th DIMM...
~§~
TC
-
I'm having some kind of issue with video spikes both in the game and watching streaming video and its giving me fits. With Comcast I believe its Flash related cause I have few when I unchecked Hardware acceleration in Flash controls it doesnt happen as much but the playback quality is lower. In Aces High it can only be my connection or my video card and I dont see why it should be my card.
I have an i7 950 at 3.07 so the processor is a little old and seems to stress when gaming at 4K. I have 12 gig RAM and considering upgrading that. I tried that Project Lasso software with mixed results, it prioritzes processes running in background. Actually I think Im having fewer spikes in game with it but still enough to bother me. Things is a few months ago I was running everything flawlessly.
I'm stumped with this. It can really be anything.
Operating System: Windows 10 Home 64-bit (10.0, Build 10586) (10586.th2_release.160906-1759)
Language: English (Regional Setting: English)
System Manufacturer: OEM
System Model: OEM
BIOS: Phoenix - AwardBIOS v6.00PG
Processor: Intel(R) Core(TM) i7 CPU 950 @ 3.07GHz (8 CPUs), ~3.1GHz
Memory: 12288MB RAM
Available OS Memory: 12280MB RAM
Page File: 3800MB used, 20766MB available
Windows Dir: C:\WINDOWS
DirectX Version: 12
DX Setup Parameters: Not found
User DPI Setting: 120 DPI (125 percent)
System DPI Setting: 96 DPI (100 percent)
DWM DPI Scaling: Disabled
Miracast: Not Available
Microsoft Graphics Hybrid: Not Supported
DxDiag Version: 10.00.10586.0000 64bit Unicode
Card name: NVIDIA GeForce GTX 780
Manufacturer: NVIDIA
Chip type: GeForce GTX 780
DAC type: Integrated RAMDAC
Device Type: Full Device
Device Key: Enum\PCI\VEN_10DE&DEV_1004&SUBSYS_27843842&REV_A1
Display Memory: 9179 MB
Dedicated Memory: 3039 MB
Shared Memory: 6139 MB
Current Mode: 2560 x 1440 (32 bit) (59Hz)
Monitor Name: Generic PnP Monitor
Monitor Model: ASUS PB278
Monitor Id: ACI27A3
Native Mode: 2560 x 1440(p) (59.951Hz)
Output Type: DVI
-
Try this: http://bbs.hitechcreations.com/smf/index.php/topic,71591.msg5129909.html#msg5129909
It does apply to your configuration.
-
Thanks. Last time I noticed a spike after a FR drop. I disabled Vsync and seemed to do better. Ive ran Trend Micro for years without issues I think I'll shut that off next.
-
Just found a 16Gb 4x4Gb kit of Corsair Vengence LPX DDR4 3000 CL 15 w\ red heat spreaders that is faster than my current 2133 set at a sale price of $129.95 (cheapest price on Newegg....all other brands are more expensive & these are the fastest quad channel 16Gb set out there until you get past 3400 as long as the CL stays at 16 but the price on those is "WHEW!!!!"
I'm probably gonna spring for these......should squeeze a little more performance out of my little box....and give me a little color contrast as well on the ole Gigabyte X99M Gaming 5 mobo............
FYI.............
:salute
Got these over the weekend and have them installed and all came up just fine.
The latency improvements w\ these 3000 mem modules vs my 2133 mem modules are very noticeable.....everything executes faster and AHIII runs much, much smoother now (MSI AB GPU frametime graphs bear this out).
Here's the math:
2133^13 = 164MHz per clock tick, 3000^15 = 200MHz per clock tick so the Corsair Vengence LPX 3000 DDR4 CL 15 4x4Gb mem kit is the faster kit performance-wise and the results that I've witnessed back this up.
From running the numbers I've noted that from 2133 to 2400 there isn't much of a change (164 vs 170) and between 3000 to 3200 there is no change at all (both @ 200) so the mem module performance threshold seems to be from 2133 to 3000 then from 3400 up....as long as the CL numbers hold.
FYI..........................
:salute
-
pudgie,
what uncore multiplier you using?
what qpi link freq?
older knowledge tells me that uncore freq should be double ram. me, since I use 3000MHz ram, thats 1500MHz /direction, so I set uncore to 3000MHz, x30.
I just read something about where cpu cache sizes work with the qpi freq.
ie: I have a 15MB cache cpu, so it can use 8 GT/s qpi link freq.
I bring this up cuz you talking about 3000MHz ram perf. Wondered about your bios choices for ram?
and I'm OCing cpu and stuffs being slowly remembered from my x58 OC........................
:salute
-
pudgie,
what uncore multiplier you using?
what qpi link freq?
older knowledge tells me that uncore freq should be double ram. me, since I use 3000MHz ram, thats 1500MHz /direction, so I set uncore to 3000MHz, x30.
I just read something about where cpu cache sizes work with the qpi freq.
ie: I have a 15MB cache cpu, so it can use 8 GT/s qpi link freq.
I bring this up cuz you talking about 3000MHz perf. Wondered about your bios choices for ram?
and I'm OCing cpu and stuffs being slowly remembered from my x58 OC........................
:salute
Hi MADe,
I didn't do anything but swap out the mem modules. Upon bootup, the Gigabyte F22 UEFI auto set the new mem modules thru the mem SPD to the exact uncore settings that you've posted along w\ the rated timings. This is the only thing that I checked the UEFI for. As far as the QPI freq, I really haven't checked this.....but my CPU (I7 5820K Haswell-E) also has a 15Mb L3 cache, running on an Intel X99 platform (which also has the exact same QPI vers & freq by specification), even though the mobo manuf's are different so at this time I'm assuming that this would also be the same.
Thanks for bringing this up.......got some checking to do. Hadn't really put any thought to any of this since the mem modules did set properly in the UEFI thru the SPD embedded in them.
Catching the flu over the weekend didn't help any either...................
The mem performance improvement was very noticeable from the get go.
:salute
-
Out of curiosity, how many people stress test their System Memory after they install/upgrade it? with something like memtest86 or memtest86+ or some other program...
I stress test all my computer builds and if upgrading certain components, stress test them as well...
I know several of y'all are diehard system tweakers, and am figuring that this is part of your guidelines
TC
-
This is a quick pass on this.......................
Here is what Intel has to say concerning the X99 chipset platform:
http://ark.intel.com/products/81761/Intel-X99-Chipset
Here is what Intel has to say concerning the I7 5820K CPU that I'm using:
http://ark.intel.com/products/82932/Intel-Core-i7-5820K-Processor-15M-Cache-up-to-3_60-GHz
Here is what Intel has to say concerning the I7 6850K CPU that you're using:
http://ark.intel.com/products/94188/Intel-Core-i7-6850K-Processor-15M-Cache-up-to-3_80-GHz
According to the X99 chipset data, there isn't a QPI present (as would be w\ a X58 platform), only a DMI 2.0 link rated at 5 GT\s (between the X99 chipset's I\O controller and the integrated Intel mem controller located on the mobo) so the CPU's would have to be linked to the same mem controller thru the on-die CPU FSB between the L3 cache and the integrated mem controller.....at least from my observation as I haven't found any documentation on this yet.
:salute
-
Here are block diagrams of a X58 & X99 platform:
:salute
-
Out of curiosity, how many people stress test their System Memory after they install/upgrade it? with something like memtest86 or memtest86+ or some other program...
I stress test all my computer builds and if upgrading certain components, stress test them as well...
I know several of y'all are diehard system tweakers, and am figuring that this is part of your guidelines
TC
Hi TC,
I usually run stress tests on my components as well to ensure fidelity & performance, but I haven't performed any mem module stress testing since I've upgraded to the X79 & X99 Intel platforms due to the copy of memtest86 that I had at the time wasn't compatible. I had already downloaded the correct versions of memtest86 & memtest86+ for these platforms but need to get an empty mem stick to set it up on so that I can use it.
But since using these particular Intel X series platforms I haven't had any issues that I could detect w\ mem modules not working OK so I've gotten a little lazy about finishing up my mem testing suite..............
:)
I wasn't having any issues w\ my Corsair Vengence LPX DDR4 2133 CL13 4x4Gb mem kit as all was working fine. I had come to a conclusion when I upgraded to this X99 platform to go w\ a 16Gb 4x4Gb quad-channel mem kit w\ the lowest CL number certified for this platform as this would represent the fastest mem modules from a latency standpoint & from using a 2133 frequency set of DDR3 mem modules on my X79 platform figured that 2133 would be enough. All was good...........until I read this thread & read Skuzzy's post which got me to rethink all this so I went on Newegg & started running performance numbers on all the different DDR4 mem modules & found what Skuzzy had said concerning using the frequency numbers alone was indeed misleading, but also using the CL numbers alone would cause 1 to leave some mem performance behind. I know that all this can be manipulated thru mem timings changes and such but I've given up on doing all this & just go w\ the mem's SPD. When I ran the numbers on a set of DDR4 3000 w\ CL 15 & found the MHz\clock tick results were large enough above my existing mem modules to get me to consider getting a set I stumbled across a set of the very same Corsair Vengence LPX DDR4 4x4Gb mem kit using 3000 frequency w\ CL 15 that were certified to work w\ my X99 platform on sale for $129.95....the cheapest of any comparable 16Gb quad channel mem module kits Newegg had as all others were more expensive....most were by a fair amount to boot. The only thing concerning these was that they had the red heatspreaders on them instead of the black ones & the exact same Corsair mem set that had the black heatspreaders weren't on sale at this time. But since I have a Gigabyte X99M Gaming 5 mobo w\ red\black color accents this wasn't a problem so I got these to see if these performance numbers would hold up & I could see\measure the difference.
All I did was pull my old set out, pop these new ones in, boot up into the UEFI to check for conformity to the SPD info....all had autoset as expected w\o any issue...all settings matched the SPD info....saved & rebooted & the rest is history.
I will be using this MHz\clock tick formula vs cost going forward to determine mem module performance capability. Will be waiting on Skuzzy to present his graphs to show the hold\transition time as well as the wait states to get a total idea of mem module performance. I've checked all over the Internet so far & haven't found any charts on this subject that were definitive enough to understand\use.
Now you know where I'm at.............
:salute
Provided below is a MSI AB graph of my box running AHIII w\ this new Corsair mem. Note the GPU frametime graph line.......this is the best my FuryX has ever done.....says something about the importance of mem module MHz\clock tick performance capability to adequately feed a big GPU to allow said GPU to maintain graphics frame timing as well as all other aspects (CPU, mem controller, DMA controller, etc). And you can clearly see a graphics screen pause captured as well.................
:salute
-
TC,
I have not memtested in awhile, my ram has always worked since going Kingston. I had some corsair highend sticks that were failing and memtest was invaluable.
Pudgie,
So ur just PnP things?
I was as well but since beginning the OC process, auto selections are bad news for moi. there are certain relationships between BCLK, Vcore, uncore and qpi freq's that cannot be pushed to far and I need specificity. X99 is different from the x58, terms changed, things became more isolated.
:salute
my x58 has been pushing a i7 920DO to 4GHz for 9years, a 3GHz cpu system is to slow for moi now. he he
wanna see some graph numbers change, OC that cpu.......................... ...
-
TC,
I have not memtested in awhile, my ram has always worked since going Kingston. I had some corsair highend sticks that were failing and memtest was invaluable.
Pudgie,
So ur just PnP things?
I was as well but since beginning the OC process, auto selections are bad news for moi. there are certain relationships between BCLK, Vcore, uncore and qpi freq's that cannot be pushed to far and I need specificity. X99 is different from the x58, terms changed, things became more isolated.
:salute
my x58 has been pushing a i7 920DO to 4GHz for 9years, a 3GHz cpu system is to slow for moi now. he he
wanna see some graph numbers change, OC that cpu.......................... ...
Yep, MADe pretty much all is PnP concerning the mem & my CPU w\ the exception of HT & CEIST being disabled on the CPU in the UEFI.
One of the side benefits of this new mem's SPD settings is that it upped my CPU's stock Turbo Boost clocks by 100 MHz....now clocking 3.7GHz instead of 3.6GHz by the mem SPD setting the uncore clock multi at 30x.
I ran a test on my prior mem & recorded the overlay readings in real time w\ my CPU set @ full stock TB clocks (3.6GHz) then ran another using CEIST enabled then going thru the Balanced power plan to set the Min Processor Power at 80% which set the lower CPU clock speed around 2.7GHz when CEIST detects a low CPU load\usage which clearly shows that AHIII doesn't need a lot of CPU clock speed as long as the necessary number of active CPU cores are used to maintain the necessary amount of CPU processing time to optimize the game threads to optimize the GPU performance (the CPU clock speeds stayed around the low power set @ 2.7GHz for the majority of the video run) which in both tests the GPU FPS was clocking at 90 FPS w\ MSI AB graphs showing the same similar, smooth frametime lines.
This is why I am staying at the stock settings as I just can't see any real benefit to OC the CPU to get any more performance from AHIII from the CPU side of things over what the stock Turbo Boost control can provide....outside of increasing CPU core temps & power usage. I may have been able to OC the 2133 mem DIMMS to run at 3000 w\ the relaxed timings to achieve the same results as noted w\ the new DIMMS but I don't believe they would have been stable in the process set up as Quad Channel.....would have been easier to do w\ 2 8Gb DIMMS set up in Dual Channel IMHO.....but the reason for buying this platform was to run the mem in quad channel for the data thruput capability..........
Over the years I have become convinced that w\ the advancements of hardware which have to a certain extent included some overclocking within the "stock settings" that I have no appetite to go past these ranges anymore........so in essence I have been "assimilated".
:D
:salute
-
I do not OC for AH, I OC to get the most out of my mid tier hardware choices. AH benefits as an unintended consequence.................. ... :cheesy:
i7 920 stock is 2.66GHz, but I have it at 4.0GHz. $200 price tag for $1000 performance.
i7 6850 is stock at 3.6GHz, I now have it at 4.5GHz. This is what some new new "lake" cpus can do.
Believe when I say you notice a 4.0GHz+ cpu envirionment. I say this cuz ultimately a faster cpu helps everything, high MHz ram to video and sound cards.
Also you seem to like the battery/power saving strats that are inheirant in all OS and mobos, why? if I may ask. Why care about battery powered unit practices on a desktop?
Same reasoning behind useing a wired pc connect over wifi connect, for me........................I turn it off, disable it, uninstall it, if it useless bloat or stuff I do not require. Making the pc handle gaming well ensures that it will work well for whatever I choose to do with it.
:salute
bit of a hijack but IMO its relevant to ram performance....
-
For those so inclined, here is a good spreadsheet to use to make runs on various mem modules to know which is the faster performing mem modules:
https://docs.google.com/spreadsheets/d/1NUcp5ICqiPHfr_uw1zmV_44QazuvsSjjTu-H4KyVfpg/edit#gid=2008389493
The 1st word results determine the faster mem modules....the lower this ns number is the faster the mem module will perform........
FYI.......................... ...
:salute
PS--I typed in the Corsair Vengence LPX 2133 DDR4 mem set that I had installed prior & the Corsair Vengence LPX 3000 DDR4 mem set I have installed now to provide a comparison...........
-
For plying new games today I'd say no less than 16 gigs. Playing Mass Effect Andromeda now and my system is using upwards of 12 gigs. I'm sure you can run it with less, but it clearly can use 12.
-
I doubt there's that much difference between 8 and 16 gigs of ram. There isn't that much difference between ram speeds either. You might be able to run some diagnostic and but frame rate....Not so much.
-
I ran memtest on my system when I put it together. I have 16GB on it, think the set was on sale at the time. Corsair vengeance. I'd also read on an overclocking site that clocking the ram on my 2600-k system would yield negligible results so I just left it at stock and clocked the CPU to 4.6k.
Sent from my XT1585 using Tapatalk