Overclock.net › Forums › Industry News › Hardware News › [TR] Exploring the impact of memory speed on Sandy Bridge performance
New Posts  All Forums:Forum Nav:

[TR] Exploring the impact of memory speed on Sandy Bridge performance - Page 8

post #71 of 77
Quote:
Originally Posted by Bwaaaa View Post
Same with Sandybridge and 1366 really. Actually all our builds are 99% E-Peen
Not really, some people can use a lot of the CPU speed/the PCIe lanes (People with dedicated RAID cards for example) on those things you mentioned, RAM speed only really helps video encoders and compression in any significant way.

Quote:
Originally Posted by Z Overlord View Post
I hear ram latencies matter more, can anyone explain this to me? And would actually decreasing the latency make a difference for me (gaming + novice photoshop user + some torrents/downloads)?

photoshop CS5 64 bit btw
Not sure about SB, but for Phenom II running 1333 CL5 is faster than 1600 CL6.


Quote:
Originally Posted by xd_1771 View Post
You guys have not saved anything. In fact, all of you have paid more by buying generic 1333 CL9 RAM. As I've said again, there is always 1600Mhz RAM for cheaper than generic kits now. You're going custom built, you have a choice. Get something worth it!
Not really, when I bought my RAM the only cheaper kit was still 1333Mhz and CL9.

Quote:
Originally Posted by beers View Post
Eh, difference has it's place.

From 2133 to 1333 in games such as Civ 5, 3 FPS equates to ~6.97% performance increase.
On this site people have done much more for much less of an increase..
It depends, time is something we all have, money to blow...Not so much.
I mean sure, I've probably spent about half a weeks worth of time in config files getting my Arch Linux install exactly where I want it to be, but right now I have heaps of spare time on my hands (Until uni starts up), whereas getting a decent 1600Mhz kit would mean I'd have to use generic TIM, I'd prefer the extra 100Mhz or so on my CPU, that probably made just as much if not more difference.

Quote:
Originally Posted by MushkinNick View Post
Plus you want to make sure your local memory reps are fed
Just saying, Love my Mushkin Silverlines, second or third cheapest kit I could get and they're working wonderfully.

Quote:
Originally Posted by Blameless View Post
Multi-tasking comparisons would be nice too.

The fact that memory performance is not a limiting factor in the vast majority of single apps is decade old news.

I run several VM concurrently, constantly access an 8GB RAM disk, often while playing games and encoding, all simultaneously. I'll have to knock my memory down a few multipliers and see if anything slows down noticeably.
True enough, someone needs to benchmark that.

Quote:
Originally Posted by staryoshi View Post
On a more related note, DDR3-1600 CL9 is the price/performance sweetspot as far as I'm concerned. You'll get solid synthetic performance and marginal real-world gains over DDR3-1333 CL9.

The reason I go with ECOs is because I'm a fan of low-voltage memory. I find that undervolting is more fulfilling than overvolting sometimes. (Please take note Zotac, I need a full range of voltage adjustment on my ITX motherboards - I won't buy one that sets a lower limit at 1.5v)

I have a set of OCZ low voltage DDR3-1600 CL9 waiting for my girl when I give her this CPU/motherboard combo
1600 CL9 for Intel, 1333 CL7 or CL8 for AMD as AMD prefers lower latency to higher clock speeds seems to be the sweet spots.

Quote:
Originally Posted by xxbassplayerxx View Post
I hope they've made significant improvements to the memory controller. AMD's memory bandwidth and controllers have really been lacking recently.
And the difference in the real world is none, I went from DDR2 to DDR3 (Not by choice either, my AM2+ board was on the way out) and got a decent increase in bandwidth according to memtest (Went from 36xxMB/s to 44xxMB/s when both were at final OC settings, 800Mhz CL4-4-4-12 1T and 1333Mhz CL8-8-8-24-1T) yet no applications really gained or lost speed.

Quote:
Originally Posted by Killam0n View Post
so then faster is better than quantity.. no?
Quantity is better than speed, as speed doesn't make much difference but having more RAM available can.

Quote:
Originally Posted by CorkyFan View Post
There is plenty of real world benefit in applications such as BOINC, Folding@Home, video and audio editing, rendering, and encoding.

Games typically see no benefit since they are almost always limited by the graphics card and nothing else.
I've never gotten an improvement in PPD from touching my RAM? Nor anything that would be like encoding (Compiling applications, which I do quite a bit)

And Games also get different FPS when you run them on a different CPU, even if the GPU is the limitation.
    
CPUMotherboardGraphicsRAM
Intel Core i5 3570k @ 4.5Ghz ASRock Z77 Pro3 Powercolor Radeon HD7950 3GB @ 1150/1350 4x4GB G.Skill Ares 2000Mhz CL9 
Hard DriveHard DriveHard DriveHard Drive
Samsung 840 250GB Western Digital Black 1TB WD1002FAEX Seagate Barracuda 3TB ST3000DM001 Samsung Spinpoint EcoGreen 2TB 
Optical DriveCoolingCoolingCooling
Pioneer DVR-220LBKS Noctua NH-D14 Scythe Gentle Typhoon 1850rpm Corsair AF140 Quiet Edition 
CoolingOSMonitorMonitor
Arcitc Cooling Acclero Twin Turbo II Arch Linux x86-64, amdgpu BenQ G2220HD BenQ G2020HD 
KeyboardPowerCaseMouse
Ducky Shine III Year of the Snake, Cherry Blue Silverstone Strider Plus 600w CoolerMaster CM690 II Black and White SteelSeries Sensei Professional 
Mouse PadAudioOther
Artisan Hien Mid Japan Black Large ASUS Xonar DX NZXT Sentry Mesh 30w Fan Controller 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Intel Core i5 3570k @ 4.5Ghz ASRock Z77 Pro3 Powercolor Radeon HD7950 3GB @ 1150/1350 4x4GB G.Skill Ares 2000Mhz CL9 
Hard DriveHard DriveHard DriveHard Drive
Samsung 840 250GB Western Digital Black 1TB WD1002FAEX Seagate Barracuda 3TB ST3000DM001 Samsung Spinpoint EcoGreen 2TB 
Optical DriveCoolingCoolingCooling
Pioneer DVR-220LBKS Noctua NH-D14 Scythe Gentle Typhoon 1850rpm Corsair AF140 Quiet Edition 
CoolingOSMonitorMonitor
Arcitc Cooling Acclero Twin Turbo II Arch Linux x86-64, amdgpu BenQ G2220HD BenQ G2020HD 
KeyboardPowerCaseMouse
Ducky Shine III Year of the Snake, Cherry Blue Silverstone Strider Plus 600w CoolerMaster CM690 II Black and White SteelSeries Sensei Professional 
Mouse PadAudioOther
Artisan Hien Mid Japan Black Large ASUS Xonar DX NZXT Sentry Mesh 30w Fan Controller 
  hide details  
Reply
post #72 of 77
Quote:
Originally Posted by CorkyFan View Post
There is plenty of real world benefit in applications such as BOINC, Folding@Home, video and audio editing, rendering, and encoding.

Games typically see no benefit since they are almost always limited by the graphics card and nothing else.
Agreeing with Brutuz here... I haven't really seen those benefits quantified, only that people claim them. Care to share benchmarks related to those specific activities?
Quote:
Originally Posted by CorkyFan View Post
Most CPU's and chipsets simply can't even take advantage of high bandwidth memory because they are just too slow and limiting. My memory gets limited until my Northbridge/IOC is over 3800MHz, then it levels off providing no further benefit. My bandwidth increases linearly well into the 4GHz range of my CPU, showing that even under 60% overclock the CPU still can't utilize the full performance of my RAM.
This is exactly our point. There are MANY other important bottlenecks in the typical computer system than ramspeed. I would argue though, that it's not most CPU's/chipsets - it's all of them.

Once someone shows me proof that an increase in performance from faster ram that is more than or equal to the same increase in performance for spending that extra money on a better CPU or GPU, then I'll shut my mouth. Really, all I've ever seen is people CLAIMING to have benefits, but have nothing to back it up except the most synthetic of benchmarks.
post #73 of 77
Quote:
Originally Posted by SgtSpike View Post
Once someone shows me proof that an increase in performance from faster ram that is more than or equal to the same increase in performance for spending that extra money on a better CPU or GPU, then I'll shut my mouth.
The proof doesn't exist afaik. RAM is the last item money should be spent on, especially if you're interest is in gaming/3D performance. The point is, after someone has the best CPU and motherboard, and overclock, the only way to increase performance from there is through memory.
post #74 of 77
I would have liked to see more professional applications benchmarked. Fast RAM certainly makes a difference in programs like Dynastrip, Quark, and Photoshop.
Bob Saget
(14 items)
 
  
CPUMotherboardGraphicsRAM
2500K Asrock Extreme4 Gen3 Integrated 8GB G.Skill Ripjaws CL9 1866Mhz 
Hard DriveOSMonitorKeyboard
60GB OCZ Vertex 2 + 1TB Samsung Spinpoint F3 (2x) Windows 7 x64 Samsung T240 + Acer X223Wbd ABS M1 
PowerCaseMouseAudio
Seasonic X-560 HAF X Logitech G5 X-Fi Xtremegamer 
  hide details  
Reply
Bob Saget
(14 items)
 
  
CPUMotherboardGraphicsRAM
2500K Asrock Extreme4 Gen3 Integrated 8GB G.Skill Ripjaws CL9 1866Mhz 
Hard DriveOSMonitorKeyboard
60GB OCZ Vertex 2 + 1TB Samsung Spinpoint F3 (2x) Windows 7 x64 Samsung T240 + Acer X223Wbd ABS M1 
PowerCaseMouseAudio
Seasonic X-560 HAF X Logitech G5 X-Fi Xtremegamer 
  hide details  
Reply
post #75 of 77
Quote:
Originally Posted by CorkyFan View Post
The proof doesn't exist afaik. RAM is the last item money should be spent on, especially if you're interest is in gaming/3D performance. The point is, after someone has the best CPU and motherboard, and overclock, the only way to increase performance from there is through memory.
Ok, I'll give you that. With an unlimited budget, there's only so much hardware you can buy.

Quote:
Originally Posted by ItsEggrollTime View Post
I would have liked to see more professional applications benchmarked. Fast RAM certainly makes a difference in programs like Dynastrip, Quark, and Photoshop.
I'd like to see that as well...
post #76 of 77
Quote:
Originally Posted by ehpexs View Post
I paid $100 for my memory, I had the choice of either 8GB of 1600 for $120 or this. Obviously 16GB is overkill for most users, but I'll still take double the ram over a slight performance increase and pocket the $20.
Thats a good deal but I would be surprised if you ever use more than 4GB mem loaded out. I would've spent that money elsewhere. This coming from a cheap sombiyatch such as myself
AMDouble-U
(17 items)
 
  
CPUMotherboardGraphicsRAM
Fx 6300 ASUS M5A99X EVO r2.0 MSI R9 270 Gskill 1866 red 
Hard DriveHard DriveHard DriveCooling
Western Digital Blue 500GB x2 raid 0 Samsung Western digital blue 1tb Custom Loop CPU/Stock GPU 
OSMonitorKeyboardPower
Windows 7 64bit Hanns G 28" 1920x1200 Thermaltake Poseidon Z Lepa MaxBron 800W 
CaseMouseAudioOther
Thermaltake Chaser A31  Roccat Savu Logitech THX cert. 2.1 LSI 3442 SAS Raid controller 
  hide details  
Reply
AMDouble-U
(17 items)
 
  
CPUMotherboardGraphicsRAM
Fx 6300 ASUS M5A99X EVO r2.0 MSI R9 270 Gskill 1866 red 
Hard DriveHard DriveHard DriveCooling
Western Digital Blue 500GB x2 raid 0 Samsung Western digital blue 1tb Custom Loop CPU/Stock GPU 
OSMonitorKeyboardPower
Windows 7 64bit Hanns G 28" 1920x1200 Thermaltake Poseidon Z Lepa MaxBron 800W 
CaseMouseAudioOther
Thermaltake Chaser A31  Roccat Savu Logitech THX cert. 2.1 LSI 3442 SAS Raid controller 
  hide details  
Reply
post #77 of 77
Did you guys forget this is an overclocking forum? Nanoseconds count.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [TR] Exploring the impact of memory speed on Sandy Bridge performance