Shortcuts

DDR3 Round-Up: Memory Performance With The Core

rated by 0 users
This post has 21 Replies | 7 Followers

Top 10 Contributor
Posts 26,171
Points 1,185,690
Joined: Sep 2007
ForumsAdministrator
News Posted: Mon, Feb 23 2009 4:48 PM

There's quite a lot of variance in DDR3 modules today, which can certainly be overwhelming for potential new buyers. While the Core i7 processor is officially rated to run at only DDR3-1066 speeds, modules that can run at DDR3-2000 or even higher are available for the Core i7, and many new motherboards support speeds even greater than these. There is also the introduction of huge 12 GB (6 x 2 GB modules) capacity kits to consider, and we have latencies in the CAS 7 to CAS 9 range, all of which affect pricing of these various modules kits dramatically.

Today, we're going to look at some of these new kits and see if we can break down what memory-related aspects users should look for when buying an X58 / Core i7 platform. We've got kits from big names like Corsair, Kingston, and OCZ in the labs. Let's try to clear up some questions, shall we?

DDR3 Round-Up: Memory Performance With The Core i7




  • | Post Points: 95
Top 25 Contributor
Posts 3,476
Points 47,080
Joined: Nov 2005
Location: Metropolis
ForumsAdministrator
Moderator
Super Dave replied on Mon, Feb 23 2009 10:43 PM

What a great story...thanks for the comparo! DDR3 is so much more expensive than DDR2 right now (I have read that the price difference between the two will be much closer at the end of this year), but Core i7 builders don't have a choice. If a Core i7 builder tries to save some money and only installs two sticks of memory, will it run in dual-channel or single-channel mode?

 SPAM-posters beware! ®

  • | Post Points: 35
Top 50 Contributor
Posts 3,017
Points 43,210
Joined: May 2005
Location: United States, Virginia
Moderator
nelsoncp21 replied on Mon, Feb 23 2009 11:10 PM

awesome article guys. Could have used that a couple weeks ago thoughStick out tongue. I would have went with the OCZ modules more then likely. Although I have not yet messed around with the dominator 1600 set that I got. I have a feeling that if I tweak it I can probaly get some better timings out of it. Default settings on my board have it running at 1333. What did you guys use when you tested the rampage 2 board?

  • | Post Points: 5
Top 75 Contributor
Posts 1,792
Points 28,435
Joined: Feb 2008
Location: South Carolina

Im thinking the right ddr3 and the rampage 2 boards are going to sport some very high marks and cant wait to see the numbers the the lucky guys here that alredy have access to these setups!Make you lealous and slobber!

 

 

 

 

  • | Post Points: 5
Top 25 Contributor
Posts 3,590
Points 54,905
Joined: Jul 2004
Location: United States, Massachusetts
ForumsAdministrator
MembershipAdministrator
Dave_HH replied on Mon, Feb 23 2009 11:57 PM

Glad you liked this one, gents. Chris did a good job with it for sure. It shows some interesting results actually.

For the RE 2 board, I believe Rob used some Qmonda RAM.

Editor In Chief
http://hothardware.com


  • | Post Points: 50
Top 150 Contributor
Posts 835
Points 11,255
Joined: Nov 2008
Location: New Zealand
tanka12345 replied on Tue, Feb 24 2009 12:38 AM

Great review. I too hope that DDR3 prices go down, so I'll be able to use my Eclipse.

Well, maybe I could 'borrow' a review sample. Wink

*New KILLER Rig!!* - Windows 7 Ultimate 64 Bit Build 7100 (RC)

Intel Core i7 920 OCed to 3.2GHz | CoolerMaster V8 | Patriot Viper 3x2GB DDR3-1333MHz | MSI Eclipse X58 | LG W2261V 22" 16:9 | Western Digital Caviar Black 640GB | MSI ATI 4870 1GB | Enermax Revolution 85+ 1050W PSU | CoolerMaster HAF 932 | ASUS DRW-22B1LT Lightscribe | Creative SB X-Fi Xtreme | Logitech X-540 |  Razer Carcharias | Razer Lachesis | Razer Destructor/ Razer Goliathus Speed/ Fnatic SteelSeries Qck+ XL | OCZ Alchemy Elixir |

*My OLD Killer PC* - Windows Xp Home Edition 32 Bit

Intel Pentium 4 2.6Ghz -1GB (2x512MB) Hynix DDR-400Mhz SDRAM - Micro ATX ECS L4S5MG/651+ - Acer AL506 15" Monitor - Seagate 2.5" ST380012A IDE 5200RPM 80GB - ASUS nVidia TNT2 M64 32MB AGP x4 - Hyena 300W PSU - *CoolerMaster Storm Scout*

  • | Post Points: 5
Not Ranked
Posts 1
Points 20
Joined: Feb 2009
kato_cmd replied on Tue, Feb 24 2009 12:47 AM

Nice article,  I do have a question about your comparison of memory bandwidth testing between 1,2 and 3 channel modes.  I've noticed that the amount of ram changes with each test.  While this isn't a problem on synthetic tests that actually check the memory bandwidth it will skew real world testing as things like photoshop tend to gain performance when the ram capacity increases.

Have you guys thought of keeping the ram capacity the same in single, dual and tri channel modes?

This would give you an exact indication of whether the extra channels are giving a benefit or if the previous result was due to extra capacity.

  • | Post Points: 20
Top 50 Contributor
Posts 3,017
Points 43,210
Joined: May 2005
Location: United States, Virginia
Moderator

Dave_HH:

Glad you liked this one, gents. Chris did a good job with it for sure. It shows some interesting results actually.

For the RE 2 board, I believe Rob used some Qmonda RAM.

Qmonda RAM? Is that like an abreviation for something? I was more curious about the speeds and how he might have adjusted timings if he did any of that when it was overclocked. Seems to have locked in settings for the timings when adjusting the speed of the memory. I have't messed around with the gazillion settings yet but thought since we were on the topic of memory I would ask.

  • | Post Points: 20
Top 150 Contributor
Posts 835
Points 11,255
Joined: Nov 2008
Location: New Zealand
tanka12345 replied on Tue, Feb 24 2009 10:14 PM

Don't you mean Qimonda Dave?

*New KILLER Rig!!* - Windows 7 Ultimate 64 Bit Build 7100 (RC)

Intel Core i7 920 OCed to 3.2GHz | CoolerMaster V8 | Patriot Viper 3x2GB DDR3-1333MHz | MSI Eclipse X58 | LG W2261V 22" 16:9 | Western Digital Caviar Black 640GB | MSI ATI 4870 1GB | Enermax Revolution 85+ 1050W PSU | CoolerMaster HAF 932 | ASUS DRW-22B1LT Lightscribe | Creative SB X-Fi Xtreme | Logitech X-540 |  Razer Carcharias | Razer Lachesis | Razer Destructor/ Razer Goliathus Speed/ Fnatic SteelSeries Qck+ XL | OCZ Alchemy Elixir |

*My OLD Killer PC* - Windows Xp Home Edition 32 Bit

Intel Pentium 4 2.6Ghz -1GB (2x512MB) Hynix DDR-400Mhz SDRAM - Micro ATX ECS L4S5MG/651+ - Acer AL506 15" Monitor - Seagate 2.5" ST380012A IDE 5200RPM 80GB - ASUS nVidia TNT2 M64 32MB AGP x4 - Hyena 300W PSU - *CoolerMaster Storm Scout*

  • | Post Points: 5
Not Ranked
Posts 3
Points 25
Joined: Feb 2009

I don't understand how (in Article Index: Clock Speed") the Crysis benchmark graph shows

  1866 MHz = 73.76 FPS

   800 MHz = 66.36 FPS

difference =  7.40 FPS

but the article text in that section claims

"Moving from DDR3-800 to DDR3-1866 bumped up Crysis by nearly 10 FPS..."

Please explain.

  • Filed under:
  • | Post Points: 20
Top 10 Contributor
Posts 4,830
Points 45,790
Joined: Feb 2008
Location: Kennesaw
rapid1 replied on Wed, Feb 25 2009 11:35 AM

No Nelson thats not an abbreviation actually. Qmonda is the company who actually made the actual DDR5 on ATI video cards. They also made the DDR4 on the 3870's as well I believe. So they are about the fastest ram maker in the world at least on a developmental aspect. I know other people are making DDR5 now but they had it developed and rolled out for the 4870's release which was what mid July last year.

OS:Win 7 Ultimate 64-bit
MB:ASUS Z87C
CPU:Intel(R) Core(TM) i7 4770 ***
GPU:Geforce GTX 770 4GB
Mem:***ingston 16384MB RAM
  • | Post Points: 5
Top 10 Contributor
Posts 4,830
Points 45,790
Joined: Feb 2008
Location: Kennesaw
rapid1 replied on Wed, Feb 25 2009 11:40 AM

No Nelson thats (Qmonda) not an abbreviation actually. Qmonda is the company who actually made the actual DDR5 on ATI video cards. They also made the DDR4 on the 3870's as well I believe. So they are about the fastest ram maker in the world at least on a developmental aspect. I know other people are making DDR5 now but they had it developed and rolled out for the 4870's release which was what mid July last year.

Oh and referring to 7.40 FPS being credited as 10 FPS I think they were using an average if it is over 5 FPS it is relatively 10 FPS inrease. Whereas, if it was 4.40 fps it would have been credited with no increase in FPS in Crysis at all.

OS:Win 7 Ultimate 64-bit
MB:ASUS Z87C
CPU:Intel(R) Core(TM) i7 4770 ***
GPU:Geforce GTX 770 4GB
Mem:***ingston 16384MB RAM
  • | Post Points: 20
Not Ranked
Posts 3
Points 25
Joined: Feb 2009

rapid1:
Oh and referring to 7.40 FPS being credited ...

Hmm...

 

  • | Post Points: 5
Not Ranked
Posts 9
Points 135
Joined: Feb 2009
Location: MI, USA
Modred189 replied on Wed, Feb 25 2009 5:25 PM

OK, so let me get this straight, you tested 6gb ram kits on a 32-bit system that, if you just include the video card, can only address 3.5gb of it? How does that make any sense?

Main Rig: Q9700@3.15ghz under a Scythe NinjaCu, Asus P5BDeluxeWifiAP, 8gb DDR2800, OCZ, BFG GTX260+ Maxcore, RaptorX150, WD Black 1TB, BFG GS650, Vista Ultimate 64-bit

HTPC: e7400, Intel DG33TL, 3GBram, Seagate 1TB, WDGreen 1TB, ATi4650, Logitech DiNovo Mini, Vista HP 32bit

Netbook: Acer Aspire One, Atom, 1.5gb ram, 120gb HD, Vista HP 32bit

  • | Post Points: 35
Top 25 Contributor
Posts 3,590
Points 54,905
Joined: Jul 2004
Location: United States, Massachusetts
ForumsAdministrator
MembershipAdministrator
Dave_HH replied on Fri, Feb 27 2009 12:18 AM

All test systems were configured with Vista 64-bit. That was a typo in the spec list and has been corrected.

Editor In Chief
http://hothardware.com


  • | Post Points: 5
Not Ranked
Posts 8
Points 85
Joined: Feb 2008

Dual channel mode.  Raw bandwidth takes a hit, but real world performance impact is nil.

  • | Post Points: 5
Not Ranked
Posts 8
Points 85
Joined: Feb 2008

This was a consideration, but keeping memory capacity at an identical level while swapping out modules for this specific test proved to be near impossible given the modules we had on hand.     Once you get above 2GB, capacity doesn't make a huge difference in overall system performance, so we went ahead with it.   Capacity is certainly a factor to keep in mind with these tests though.

  • | Post Points: 5
Not Ranked
Posts 8
Points 85
Joined: Feb 2008

These tests were indeed done with Vista x64 Edition.   All of the tests, when possible, were done in 64-bit mode.    The OS being listed as x32 was a typo which has since been corrected.  Thank you for pointing it out to us.

  • | Post Points: 20
Not Ranked
Posts 9
Points 135
Joined: Feb 2009
Location: MI, USA
Modred189 replied on Fri, Feb 27 2009 8:19 AM

Thank YOU for the response and correction!

ON a side note, I would also read these two articles by Extremetech, they claim that the amount of memory in a system is far more important than the speed:

http://www.extremetech.com/article2/0,2845,2328804,00.asp

http://www.extremetech.com/article2/0,2845,2333466,00.asp

Main Rig: Q9700@3.15ghz under a Scythe NinjaCu, Asus P5BDeluxeWifiAP, 8gb DDR2800, OCZ, BFG GTX260+ Maxcore, RaptorX150, WD Black 1TB, BFG GS650, Vista Ultimate 64-bit

HTPC: e7400, Intel DG33TL, 3GBram, Seagate 1TB, WDGreen 1TB, ATi4650, Logitech DiNovo Mini, Vista HP 32bit

Netbook: Acer Aspire One, Atom, 1.5gb ram, 120gb HD, Vista HP 32bit

  • | Post Points: 20
Top 25 Contributor
Posts 3,590
Points 54,905
Joined: Jul 2004
Location: United States, Massachusetts
ForumsAdministrator
MembershipAdministrator
Dave_HH replied on Fri, Feb 27 2009 9:21 AM

Modred189:

Thank YOU for the response and correction!

ON a side note, I would also read these two articles by Extremetech, they claim that the amount of memory in a system is far more important than the speed:

http://www.extremetech.com/article2/0,2845,2328804,00.asp

http://www.extremetech.com/article2/0,2845,2333466,00.asp

 


Heya Modred,

First, thanks for your inputs here.  Good stuff and good to have you with us. 

I wouldn't directly relate the findings in those ET articles as all that relevant to the tests we ran on a Core i7 test bed.  In fact, if you look at the performance numbers we came up with, the 3GB kits performed about on par with the 6GB kits and were significantly more affected by latency and clock speed.  The fact of the matter is, with Core i7, the game has changed a bit.  Now, instead of a 2GB versus 4GB density spread with a Core 2 dual-channel system, you have a 3GB versus 6GB spread with Core i7.

See here:
http://hothardware.com/Articles/Triple-Channel-DDR3-with-the-Core-i7/?page=11
http://hothardware.com/Articles/Triple-Channel-DDR3-with-the-Core-i7/?page=12

Remember, both Kingston kits are 3GB and the OCZ and Corsair kits are 6GB.

It was an interesting look that Chris put together for us and offered quite a few surprises along the way!

Thanks,
Dave Altavilla
Editor In Chief
HotHardware.com
 

Editor In Chief
http://hothardware.com


Not Ranked
Posts 9
Points 135
Joined: Feb 2009
Location: MI, USA
Modred189 replied on Fri, Feb 27 2009 11:06 AM

AH, I see. FOr some reason I was thinking that they were all 6's. I wonder where the difference comes in? Maybe the increased bandwidth or fewer steps between ram and cpu allow the speed to be more of a factor...

 

BTW, I'm glad I followed the endgadget link here, I really have enjopyed some of teh articles, particularly the Sus nettop vid. +1 reader!

Main Rig: Q9700@3.15ghz under a Scythe NinjaCu, Asus P5BDeluxeWifiAP, 8gb DDR2800, OCZ, BFG GTX260+ Maxcore, RaptorX150, WD Black 1TB, BFG GS650, Vista Ultimate 64-bit

HTPC: e7400, Intel DG33TL, 3GBram, Seagate 1TB, WDGreen 1TB, ATi4650, Logitech DiNovo Mini, Vista HP 32bit

Netbook: Acer Aspire One, Atom, 1.5gb ram, 120gb HD, Vista HP 32bit

  • | Post Points: 20
Top 25 Contributor
Posts 3,590
Points 54,905
Joined: Jul 2004
Location: United States, Massachusetts
ForumsAdministrator
MembershipAdministrator
Dave_HH replied on Fri, Feb 27 2009 11:11 AM

Welcome aboard Modred189! Big Smile

Editor In Chief
http://hothardware.com


  • | Post Points: 5
Page 1 of 1 (22 items) | RSS