Overclock.net › Forums › Components › Memory › Dominators > Team Extreem
New Posts  All Forums:Forum Nav:

Dominators > Team Extreem - Page 11

post #101 of 349
Those of you who believe that operating your System Memory within a 1:1 ratio within the Core Micro-Architecture is the best option please research the Term Von Neumann bottleneck and then come back to talking about Memory

Until then you will not understand why running your memory faster is the better alternative, regardless if you have even something like 20-20-20-90 timings (ns); (overexample, ~2300MHz would compensate).
post #102 of 349
So, basically The Manual, running an higher RAM than CPU divider is the best approach for Core 2 Duo?
The Shehanigator
(13 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core 2 Duo E6300 ASUS P5-B Deluxe ATI Radeon X1900GT 256MB DDR3 Crucial Ballistix (2x1GB) DDR2 1000 
Hard DriveOptical DriveOSMonitor
WD Caviar 250GB SATAII 3Gb/s 7200RPM SAMSUNG 18X DVD-RW Windows XP w/ SP2 Acer AL1706Ab Black 17" 8ms LCD Monitor 
KeyboardPowerCaseMouse
Logitech Media Keyboard 500W Silverstone "Element" Thermaltake Armor w/ 25CC Side Fan Logitech LX-3 Optical 
Mouse Pad
None 
  hide details  
Reply
The Shehanigator
(13 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core 2 Duo E6300 ASUS P5-B Deluxe ATI Radeon X1900GT 256MB DDR3 Crucial Ballistix (2x1GB) DDR2 1000 
Hard DriveOptical DriveOSMonitor
WD Caviar 250GB SATAII 3Gb/s 7200RPM SAMSUNG 18X DVD-RW Windows XP w/ SP2 Acer AL1706Ab Black 17" 8ms LCD Monitor 
KeyboardPowerCaseMouse
Logitech Media Keyboard 500W Silverstone "Element" Thermaltake Armor w/ 25CC Side Fan Logitech LX-3 Optical 
Mouse Pad
None 
  hide details  
Reply
post #103 of 349
In my opinion, and mathematically, it will always be better to run your System Memory speed greater than that of your System Bus on current Micro-Processors

The Data Transfer from your CPU is extremely high, far greater than what current Memory today is capable of transferring in the form of data (to 128 bits). Therefore increasing your memory speed as high as possible will generate the highest results because system bandwidth will be increased.

Normally the CPU has to spend a moderate proportion of it's time waiting for data from Main Memory to be received. With the increase in Bandwidth this problem will diminish, but not entirely.

Memory Timings matter much less within a Core 2 Duo computer, as Intel Enhanced Memory Timings are implemented that compensate for poor Memory Timings.

For comparison:
Increase your System Memory Clock speed greatly above you're system bus speed, and increase the latency timings. Then see how much faster SuperPI (or equivalent) completes it operations. It will be faster as the CPU can receive data faster, and has to wait less for data to be received.
The wait generated by the Memory Latency is much smaller than that of the CPU waiting for data
post #104 of 349
That's a very good and complete post. And very informative as well. Off we go to apply this process!

EDIT: Time to make this Crucial B 1000 rock!
The Shehanigator
(13 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core 2 Duo E6300 ASUS P5-B Deluxe ATI Radeon X1900GT 256MB DDR3 Crucial Ballistix (2x1GB) DDR2 1000 
Hard DriveOptical DriveOSMonitor
WD Caviar 250GB SATAII 3Gb/s 7200RPM SAMSUNG 18X DVD-RW Windows XP w/ SP2 Acer AL1706Ab Black 17" 8ms LCD Monitor 
KeyboardPowerCaseMouse
Logitech Media Keyboard 500W Silverstone "Element" Thermaltake Armor w/ 25CC Side Fan Logitech LX-3 Optical 
Mouse Pad
None 
  hide details  
Reply
The Shehanigator
(13 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core 2 Duo E6300 ASUS P5-B Deluxe ATI Radeon X1900GT 256MB DDR3 Crucial Ballistix (2x1GB) DDR2 1000 
Hard DriveOptical DriveOSMonitor
WD Caviar 250GB SATAII 3Gb/s 7200RPM SAMSUNG 18X DVD-RW Windows XP w/ SP2 Acer AL1706Ab Black 17" 8ms LCD Monitor 
KeyboardPowerCaseMouse
Logitech Media Keyboard 500W Silverstone "Element" Thermaltake Armor w/ 25CC Side Fan Logitech LX-3 Optical 
Mouse Pad
None 
  hide details  
Reply
post #105 of 349
Quote:
Originally Posted by The_Manual View Post
In my opinion, and mathematically, it will always be better to run your System Memory speed greater than that of your System Bus on current Micro-Processors

The Data Transfer from your CPU is extremely high, far greater than what current Memory today is capable of transferring in the form of data (to 128 bits). Therefore increasing your memory speed as high as possible will generate the highest results because system bandwidth will be increased.

Normally the CPU has to spend a moderate proportion of it's time waiting for data from Main Memory to be received. With the increase in Bandwidth this problem will diminish, but not entirely.

Memory Timings matter much less within a Core 2 Duo computer, as Intel Enhanced Memory Timings are implemented that compensate for poor Memory Timings.

For comparison:
Increase your System Memory Clock speed greatly above you're system bus speed, and increase the latency timings. Then see how much faster SuperPI (or equivalent) completes it operations. It will be faster as the CPU can receive data faster, and has to wait less for data to be received.
The wait generated by the Memory Latency is much smaller than that of the CPU waiting for data
I would like to see this explained mathematically. Math, as you know, doesn't lie.
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
post #106 of 349
Without going into extreme detail I will just give you a simple calculation for you to work with.

I will use my current Processor (including overclock as the example).

Current MT/s value = 1425QDR

Maximum 32 bit Integer Bandwidth: 1425 x 8 = 11400MB/s - (Don't forget this is only 32 bit, we can go much higher than this value in reality).

Any DDR/DDR2 Memory you know of that can deal with that? (GDDR is invalid).
post #107 of 349
Quote:
Originally Posted by The_Manual View Post
Without going into extreme detail I will just give you a simple calculation for you to work with.

I will use my current Processor (including overclock as the example).

Current MT/s value = 1425QDR

Maximum 32 bit Integer Bandwidth: 1425 x 8 = 11400MB/s

Any DDR/DDR2 Memory you know of that can deal with that? (GDDR is invalid).
So your processor can do 1.425 Billion transfers per second. (not sure what the QDR stands for). How did you arrive to this number? Is that your bus speed? Why did you multiply by 8? I want detail.
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
post #108 of 349
Thread Starter 
Quote:
Originally Posted by Intervention View Post
I still think you should try both my memory setting and your current settings as my ram will never do 1260Mhz, but am almost certain, you could achieve my settings. That would eliminate the mobo difference.
Well I have to go to work now. I will try to get the highest clock possible later.

I will bring my temp gun home to make sure things arent getting too hot.

1350 Here I come
MONOLITH
(13 items)
 
  
CPUMotherboardGraphicsRAM
E6600 L631A871 eVGA 780i Dual G92 GTS's Corsair Dominator 2GB 
Hard DriveOptical DriveOSMonitor
Samsung Spinpoint F1 1TB Lite-On LightScribe Win Vista SP1 BenQ FP241w 
KeyboardPowerCaseMouse
Saiktek Eclipse II Corsair HX620 Rocketfish Razer Copperhead 
Mouse Pad
Steelpad Qck Heavy 
  hide details  
Reply
MONOLITH
(13 items)
 
  
CPUMotherboardGraphicsRAM
E6600 L631A871 eVGA 780i Dual G92 GTS's Corsair Dominator 2GB 
Hard DriveOptical DriveOSMonitor
Samsung Spinpoint F1 1TB Lite-On LightScribe Win Vista SP1 BenQ FP241w 
KeyboardPowerCaseMouse
Saiktek Eclipse II Corsair HX620 Rocketfish Razer Copperhead 
Mouse Pad
Steelpad Qck Heavy 
  hide details  
Reply
post #109 of 349
Quote:
Originally Posted by The_Manual View Post
In my opinion, and mathematically, it will always be better to run your System Memory speed greater than that of your System Bus on current Micro-Processors

The Data Transfer from your CPU is extremely high, far greater than what current Memory today is capable of transferring in the form of data (to 128 bits). Therefore increasing your memory speed as high as possible will generate the highest results because system bandwidth will be increased.

Normally the CPU has to spend a moderate proportion of it's time waiting for data from Main Memory to be received. With the increase in Bandwidth this problem will diminish, but not entirely.

Memory Timings matter much less within a Core 2 Duo computer, as Intel Enhanced Memory Timings are implemented that compensate for poor Memory Timings.

For comparison:
Increase your System Memory Clock speed greatly above you're system bus speed, and increase the latency timings. Then see how much faster SuperPI (or equivalent) completes it operations. It will be faster as the CPU can receive data faster, and has to wait less for data to be received.
The wait generated by the Memory Latency is much smaller than that of the CPU waiting for data
I wouldn't go as far as all that! I mean, I'll agree with you, timings are a little less important on the C2D processors than on the AMDs. The reason for this being the memory controller on AMDs is on the chip.

But the point that I make is, is that SuperPi is a synthetic benchmark. It tests theoretical values. It uses the RAM as repositories during its test. In no way does it test real world performance. If you really want to see the difference between running a 1:1 ratio on tight timings versus a divider on looser timings, do what Ropey says.

Take a digital stopwatch, load a game like F.E.A.R. or Oblivion and time which one is faster (leaving everything else constant of course). He has tried this on his C2D rig and I have tried it on my AM2 rig. We have the same exact RAM and both got the same results. That while the increased bandwidth and looser timings outperformed in SYNTHETIC benchmarks, in real world apps (ie loading a game) the 1:1 ratio and tightest timings clearly were more efficient.

People can believe whatever they want. But I'm telling you I have tested it both ways. I run my rig 24/7 on 1:1 ratio. When I benchmark, here comes the bandwidth.
2000s
(14 items)
 
Tel'aran'rhiod
(14 items)
 
Laptop
(13 items)
 
CPUMotherboardGraphicsRAM
Q6700 G0 @ 3.6GHz 24/7 1.47v ASUS P5K Deluxe BFG 8800GT OC 512MB 2x2GB G.Skill DDR2-1000 (1:1 4-4-4-12) 
Hard DriveOptical DriveOSMonitor
2x 150GB WD VelocoRaptor RAID 0 & 1TB Seagate Samsung SATA DVD+/-RW Windows 7 Pro. 64 bit 25.5" ASUS VW266H 
KeyboardPowerCaseMouse
Red Edition Saitek Eclipse Corsair TX750 CM 590 Razor Diamondback 
Mouse PadAudio
14" x 18" Gaming pad X-Fi Xtreme Gamer 
CPUMotherboardGraphicsRAM
i7-2700K  ASRock P67 Extreme6 (B3) LGA 1155  Sapphire HD 6970 2GB Dual BIOS/Fan G.SKILL Ripjaws X Series 16GB (4 x 4GB) DDR3-16... 
Hard DriveOptical DriveCoolingOS
2x 150GB WD VelociRaptor Samsung SATA DVD+/-RW Corsair H50 Yate Loon Push-Pull Windows 7 Professional 64bit 
MonitorKeyboardPowerCase
ASUS VW266H 25.5" LCD 2ms 1920 x 1200 Blue Edition Saitek Eclipse Corsair TX750 Antec 900 
MouseAudio
Logitech G500 X-Fi Xtreme Gamer 
CPUGraphicsRAMHard Drive
i7-720QM 1.6Ghz 1GB Nvidia GeForce GT 230M 4GB DDR3 320GB 7200RPM SATA 
Optical DriveOSMonitorCase
Lightscribe Blu-Ray ROM w/ DVD+/-R/RW DL Windows 7 Prof. x64 18.4" 1920 x 1080 HP 
  hide details  
Reply
2000s
(14 items)
 
Tel'aran'rhiod
(14 items)
 
Laptop
(13 items)
 
CPUMotherboardGraphicsRAM
Q6700 G0 @ 3.6GHz 24/7 1.47v ASUS P5K Deluxe BFG 8800GT OC 512MB 2x2GB G.Skill DDR2-1000 (1:1 4-4-4-12) 
Hard DriveOptical DriveOSMonitor
2x 150GB WD VelocoRaptor RAID 0 & 1TB Seagate Samsung SATA DVD+/-RW Windows 7 Pro. 64 bit 25.5" ASUS VW266H 
KeyboardPowerCaseMouse
Red Edition Saitek Eclipse Corsair TX750 CM 590 Razor Diamondback 
Mouse PadAudio
14" x 18" Gaming pad X-Fi Xtreme Gamer 
CPUMotherboardGraphicsRAM
i7-2700K  ASRock P67 Extreme6 (B3) LGA 1155  Sapphire HD 6970 2GB Dual BIOS/Fan G.SKILL Ripjaws X Series 16GB (4 x 4GB) DDR3-16... 
Hard DriveOptical DriveCoolingOS
2x 150GB WD VelociRaptor Samsung SATA DVD+/-RW Corsair H50 Yate Loon Push-Pull Windows 7 Professional 64bit 
MonitorKeyboardPowerCase
ASUS VW266H 25.5" LCD 2ms 1920 x 1200 Blue Edition Saitek Eclipse Corsair TX750 Antec 900 
MouseAudio
Logitech G500 X-Fi Xtreme Gamer 
CPUGraphicsRAMHard Drive
i7-720QM 1.6Ghz 1GB Nvidia GeForce GT 230M 4GB DDR3 320GB 7200RPM SATA 
Optical DriveOSMonitorCase
Lightscribe Blu-Ray ROM w/ DVD+/-R/RW DL Windows 7 Prof. x64 18.4" 1920 x 1080 HP 
  hide details  
Reply
post #110 of 349
First QDR stands for Quad Dual Rate (Intel Quad Bus Design)

To calculate the Transfers per Second (MT/s) you are required to multiply you're system bus speed by the Number of Transfers per Clock Cycle.

My bus speed = 356MHz.

Intel Core 2 Duo Micro-Processors are capable of 4 Transfers per clock (following design, and will almost always accomplish this value).

Therefore 356 x 4 = ~1425MHz QDR (What you would call FSB speed).

The "FSB Speed" is the Number of Million Transfers a second within a Computer Micro-Processor.

The reason why you are to multiply this value by the Integer eight is to get the value for Million Transfers a second into the term "bytes (Megabytes in this case)

@ CWell1337. Testing Memory within 3D applications is illogical. System Memory is not dramatically "paged" for bandwidth within three dimensional objects, and systematic rendering. If you wish to ascertain what I have stated I suggest you get your computer working on a task that requires extreme data transfers. For this I suggest you have a look at some Computation Algorithms.
If you want to play games you can leave memory out of the equation (to a degree) as it does not influence the system as much as the GPU and CPU (and equivalent components). Changing the Memory Config will cause slight changes in Frame Rates, but small in comparison to other changes you could make.
Wait a second, you are testing an AMD processor vs. an Intel processor? Well that really changes the meaning of "a fair test" in the Memory Departement
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Memory
Overclock.net › Forums › Components › Memory › Dominators > Team Extreem