Take a look at our
ThinkPads.com HOME PAGE
For those who might want to contribute to the blog, start here: Editors Alley Topic
Then contact Bill with a Private Message

4GB RAM on T60P

T60/T61 series specific matters only
Message
Author
akhavan
thinkpads.com customer
thinkpads.com customer
Posts: 278
Joined: Fri Apr 30, 2004 5:21 am
Location: Dubai

4GB RAM on T60P

#1 Post by akhavan » Fri Jan 12, 2007 1:19 am

Hi,

Is anyone running 4GB RAM on T60p? If yes how is the performance compare to 3GB/2GB RAM?

Sepehr

ronan_zj
Junior Member
Junior Member
Posts: 311
Joined: Sat Oct 16, 2004 2:02 am
Location: San Francisco/UC Berkeley/UC Davis

#2 Post by ronan_zj » Fri Jan 12, 2007 2:04 am

i dont even see the difference between 1G and 2G on XP.
maybe its noticeable on Vista.

freakwave
Freshman Member
Posts: 118
Joined: Wed May 10, 2006 1:00 pm

#3 Post by freakwave » Fri Jan 12, 2007 2:50 am

Hi,

as always it depends. I am running 4GB. Windows reports about 3.1 as available (famous PCI hole).
I need this memory because I am running vmware where I need alone 1.7Gig.
If you do not need the memory from an application standpoint, you will not feel any difference in speed.

Regards,

Wolfgang

akhavan
thinkpads.com customer
thinkpads.com customer
Posts: 278
Joined: Fri Apr 30, 2004 5:21 am
Location: Dubai

#4 Post by akhavan » Fri Jan 12, 2007 3:17 am

Thanks guys

Wolfgang I thought window XP Pro was able to recognize 4GB ram and utilize it?

Sepehr

ronan_zj
Junior Member
Junior Member
Posts: 311
Joined: Sat Oct 16, 2004 2:02 am
Location: San Francisco/UC Berkeley/UC Davis

#5 Post by ronan_zj » Fri Jan 12, 2007 4:08 am

akhavan wrote:Thanks guys

Wolfgang I thought window XP Pro was able to recognize 4GB ram and utilize it?

Sepehr
no, windows XP pro only recognize memory up to 3GB.

freakwave
Freshman Member
Posts: 118
Joined: Wed May 10, 2006 1:00 pm

#6 Post by freakwave » Fri Jan 12, 2007 5:30 am

Hi Sepher,

no operating system will be able to address this PCI hole.

Windows does not see the full 4GB since the chipset and BIOS use the addresspace below 4GB for PCI devices.

Since Windows XP 32 can only address 4GB the PCI address space has to be within that region, thats why you loose physical memory.

If all is 64bit (processor/os) and chipset/bios supports it, you could move this pci address space to a higher address region (above 4GB) , and you would see the full physical 4GBytes.
But I do not know if this would even be theoretically possible with the chipsets/BIOS we have in the thinkpad right now.

So I would not assume that you will see that 1GByte that you loose right now in the future.
Even with 64bit OS it is very likely it is not usable.

Regards,

Wolfgang

akhavan
thinkpads.com customer
thinkpads.com customer
Posts: 278
Joined: Fri Apr 30, 2004 5:21 am
Location: Dubai

#7 Post by akhavan » Fri Jan 12, 2007 8:04 am

Hi Wolfgang,

Thank you for your explanation. If I understand it right then for Duo processor the best would be 2GB since with 3GB it would not process in duo mode. Am I right?

thanks

Sepehr

christopher_wolf
Special Member
Posts: 5741
Joined: Sat Oct 08, 2005 1:24 pm
Location: UC Berkeley, California
Contact:

#8 Post by christopher_wolf » Fri Jan 12, 2007 2:38 pm

The CPUs themselves, due to PAE, have had the ability to address far more than 4GB for quite awhile now, certainly through the P-M line. The problem lies with not only the OS you use, and there are tricks that 32-bit OSes can use to address more memory, but the chipset hardware itself. The CPU may support it, but will the BIOS/Chipset/PCIs.

In any case, even if you have 4GB of RAM in your T60 C2D, you will still hit the limit of 3.24GB in Windows XP Pro; I haven't, however, tested a 64-bit OS out on the T60 C2Ds with 4GB to see if nearly 4GB can be physically addressed. :)
IBM ThinkPad T43 Model 2668-72U 14.1" SXGA+ 1GB |IBM 701c

~o/
I met someone who looks a lot like you.
She does the things you do.
But she is an IBM.
/~o ---ELO from "Yours Truly 2059"

pianowizard
Senior ThinkPadder
Senior ThinkPadder
Posts: 8506
Joined: Tue Jun 28, 2005 5:07 am
Location: Ann Arbor, MI
Contact:

#9 Post by pianowizard » Fri Jan 12, 2007 3:40 pm

freakwave wrote:as always it depends. I am running 4GB. Windows reports about 3.1 as available (famous PCI hole).
Then why didn't you just get 1GB + 2GB and save yourself $300US?
HP Pavilion x2 12 (Core M3-6Y30, 3.14lb); Microsoft Surface 3 (Atom x7-Z8700, 2.00lb)
Dell OptiPlex 5040 SFF (Core i5-6600); HP 8300 Elite tower (Core i7-3770)
Acer ET322QK, T272HUL; Bolva 40BL00H7; Crossover 404K; Dell 3008WFP, D2015Hf; Monoprice 10734; QNIX QHD2410R; Seiki Pro SM40UNP

Austin_Goh
User with bad email address, PLEASE fix!
Posts: 18
Joined: Sun Dec 17, 2006 9:50 am
Location: Skudai, Johor, Malaysia
Contact:

#10 Post by Austin_Goh » Sun Jan 14, 2007 10:37 am

I myself using 4GB on T60 (no 'p' here :()

My review:
http://forum.thinkpads.com/viewtopic.php?t=35638

yes there is only 3GB of addressable memory, tried on Vista 32bit, XP Pro 32/64bit, Server 2003 Enterprise Edition etc.

If you want to buy 4GB RAM for T60, do PM me, I have 4 more pieces 2GB modules for sale at USD700 per pair. :(

brentpresley
ThinkPadder
ThinkPadder
Posts: 1434
Joined: Fri Dec 23, 2005 4:19 pm
Location: Dallas, TX
Contact:

#11 Post by brentpresley » Sun Jan 14, 2007 11:03 am

Austin_Goh wrote:I myself using 4GB on T60 (no 'p' here :()

My review:
http://forum.thinkpads.com/viewtopic.php?t=35638

yes there is only 3GB of addressable memory, tried on Vista 32bit, XP Pro 32/64bit, Server 2003 Enterprise Edition etc.

If you want to buy 4GB RAM for T60, do PM me, I have 4 more pieces 2GB modules for sale at USD700 per pair. :(
Just a word to the wise, these are the slower 533MHz sticks, not the 667 good stuff (PC2-5300).
Custom T60p
2.33GHz 4MB 667MHz Core 2 Duo
4GB PC2-5300 DDR SDRAM
Bluetooth / Atheros ABGN
200GB 7k200 7200RPM Hard Drive
8X DVD Multiburner
15" UXGA - ATI FireGL V5250 (256MB)

http://www.xcpus.com

pianowizard
Senior ThinkPadder
Senior ThinkPadder
Posts: 8506
Joined: Tue Jun 28, 2005 5:07 am
Location: Ann Arbor, MI
Contact:

#12 Post by pianowizard » Sun Jan 14, 2007 12:37 pm

brentpresley wrote:Just a word to the wise, these are the slower 533MHz sticks, not the 667 good stuff (PC2-5300).
If you use 533MHz memory, does that cause the 667MHz processor to run at 533MHz as well?
HP Pavilion x2 12 (Core M3-6Y30, 3.14lb); Microsoft Surface 3 (Atom x7-Z8700, 2.00lb)
Dell OptiPlex 5040 SFF (Core i5-6600); HP 8300 Elite tower (Core i7-3770)
Acer ET322QK, T272HUL; Bolva 40BL00H7; Crossover 404K; Dell 3008WFP, D2015Hf; Monoprice 10734; QNIX QHD2410R; Seiki Pro SM40UNP

brentpresley
ThinkPadder
ThinkPadder
Posts: 1434
Joined: Fri Dec 23, 2005 4:19 pm
Location: Dallas, TX
Contact:

#13 Post by brentpresley » Sun Jan 14, 2007 12:38 pm

pianowizard wrote:
brentpresley wrote:Just a word to the wise, these are the slower 533MHz sticks, not the 667 good stuff (PC2-5300).
If you use 533MHz memory, does that cause the 667MHz processor to run at 533MHz as well?
No. It just cuts the available memory bandwidth. The CPU bus and memory bus have independent dividers.
Custom T60p
2.33GHz 4MB 667MHz Core 2 Duo
4GB PC2-5300 DDR SDRAM
Bluetooth / Atheros ABGN
200GB 7k200 7200RPM Hard Drive
8X DVD Multiburner
15" UXGA - ATI FireGL V5250 (256MB)

http://www.xcpus.com

Austin_Goh
User with bad email address, PLEASE fix!
Posts: 18
Joined: Sun Dec 17, 2006 9:50 am
Location: Skudai, Johor, Malaysia
Contact:

#14 Post by Austin_Goh » Sun Jan 14, 2007 8:49 pm

brentpresley wrote:
Austin_Goh wrote:I myself using 4GB on T60 (no 'p' here :()

My review:
http://forum.thinkpads.com/viewtopic.php?t=35638

yes there is only 3GB of addressable memory, tried on Vista 32bit, XP Pro 32/64bit, Server 2003 Enterprise Edition etc.

If you want to buy 4GB RAM for T60, do PM me, I have 4 more pieces 2GB modules for sale at USD700 per pair. :(
Just a word to the wise, these are the slower 533MHz sticks, not the 667 good stuff (PC2-5300).
Your word was not wise, these are lower clocked memory modules, however doesn't mean slower.
Overall performance of Dual Channel 533MHz Latency @ CL4 is on par if not faster than, Dual Channel 667MHz Latency @ CL5.

pianowizard
Senior ThinkPadder
Senior ThinkPadder
Posts: 8506
Joined: Tue Jun 28, 2005 5:07 am
Location: Ann Arbor, MI
Contact:

#15 Post by pianowizard » Sun Jan 14, 2007 9:13 pm

Austin_Goh wrote:Overall performance of Dual Channel 533MHz Latency @ CL4 is on par if not faster than, Dual Channel 667MHz Latency @ CL5.
We had the exact same discussion just last week. 533MHz Latency @ CL4 and 667MHz Latency @ CL5 both have a latency of 7.5 ns. I don't mean to burst your bubbe, but you really should have gotten 667MHz memory for your T60.
HP Pavilion x2 12 (Core M3-6Y30, 3.14lb); Microsoft Surface 3 (Atom x7-Z8700, 2.00lb)
Dell OptiPlex 5040 SFF (Core i5-6600); HP 8300 Elite tower (Core i7-3770)
Acer ET322QK, T272HUL; Bolva 40BL00H7; Crossover 404K; Dell 3008WFP, D2015Hf; Monoprice 10734; QNIX QHD2410R; Seiki Pro SM40UNP

blackwood
Posts: 37
Joined: Tue Aug 15, 2006 4:11 pm
Location: Redondo Beach, CA

#16 Post by blackwood » Sun Jan 14, 2007 9:21 pm

freakwave wrote: So I would not assume that you will see that 1GByte that you loose right now in the future.
Even with 64bit OS it is very likely it is not usable.

Regards,

Wolfgang
I've seen a single process (64-bit CATIA V5) actually using a total of 12GB (8 physical) on my Xeon XPx64 workstation (at the office).

Image
T60 2613EAU
gallery.mblackwood.com

Austin_Goh
User with bad email address, PLEASE fix!
Posts: 18
Joined: Sun Dec 17, 2006 9:50 am
Location: Skudai, Johor, Malaysia
Contact:

#17 Post by Austin_Goh » Sun Jan 14, 2007 9:43 pm

pianowizard wrote:
Austin_Goh wrote:Overall performance of Dual Channel 533MHz Latency @ CL4 is on par if not faster than, Dual Channel 667MHz Latency @ CL5.
We had the exact same discussion just last week. 533MHz Latency @ CL4 and 667MHz Latency @ CL5 both have a latency of 7.5 ns. I don't mean to burst your bubbe, but you really should have gotten 667MHz memory for your T60.
Prove me wrong then, please use DUAL CHANNEL, not single channel.
And also please do it on Mobile Merom 667MHz FSB Platform, don't use Desktop Conroe 1066FSB to test.

Prove to me 2 screenshots with Sisoft Sandra Memory Bandwidth Test. 1 with 2x2GB 533, 1 with 2x2GB 667.

Furthermore, would you pay USD589 x2 for 667MHz Dual Channel? or USD700 (Not meant to promote my things but it is merely a comparison) for Dual Channel 533MHz?

pianowizard
Senior ThinkPadder
Senior ThinkPadder
Posts: 8506
Joined: Tue Jun 28, 2005 5:07 am
Location: Ann Arbor, MI
Contact:

#18 Post by pianowizard » Sun Jan 14, 2007 9:57 pm

Austin_Goh wrote:Prove me wrong then
There is no need to prove anything. Just look at the numbers: 667MHz is higher than 533MHz. What's your proof that 533MHz is faster than 667MHz?
Austin_Goh wrote:Furthermore, would you pay USD589 x2 for 667MHz Dual Channel? or USD700 (Not meant to promote my things but it is merely a comparison) for Dual Channel 533MHz?
I would not even buy any 2GB modules at this point because I know very well that I would not need that much memory, and that these sticks (both the 533MHz and the 667MHz modules) are currently way over-priced. I would wait for a year or two before buying them. But if I were forced (e.g. if someone put a gun on my head) to buy two 2GB modules for my X60s (which has a 667MHz FSB), I would rather pay more and get two 667MHz modules.

In fact, if the 533MHz sticks were really faster than the 667MHz sticks, why would Lenovo and other sellers have the guts to sell the latter at higher prices?
HP Pavilion x2 12 (Core M3-6Y30, 3.14lb); Microsoft Surface 3 (Atom x7-Z8700, 2.00lb)
Dell OptiPlex 5040 SFF (Core i5-6600); HP 8300 Elite tower (Core i7-3770)
Acer ET322QK, T272HUL; Bolva 40BL00H7; Crossover 404K; Dell 3008WFP, D2015Hf; Monoprice 10734; QNIX QHD2410R; Seiki Pro SM40UNP

Austin_Goh
User with bad email address, PLEASE fix!
Posts: 18
Joined: Sun Dec 17, 2006 9:50 am
Location: Skudai, Johor, Malaysia
Contact:

#19 Post by Austin_Goh » Sun Jan 14, 2007 10:02 pm

pianowizard wrote:
Austin_Goh wrote:Prove me wrong then
There is no need to prove anything. Just look at the numbers: 667MHz is higher than 533MHz. What's your proof that 533MHz is faster than 667MHz?
Austin_Goh wrote:Furthermore, would you pay USD589 x2 for 667MHz Dual Channel? or USD700 (Not meant to promote my things but it is merely a comparison) for Dual Channel 533MHz?
I would not even buy any 2GB modules at this point because I know very well that I would not need that much memory, and these sticks (both the 533MHz and the 667MHz modules) are currently way over-priced. I would wait for a year or two before buying them. But if I were forced (e.g. if someone put a gun on my head) to buy two 2GB modules for my X60s (which has a 667MHz FSB), I would rather pay more and get two 667MHz modules.
Then read here carefully.

Bottleneck of Core 2 Duo Merom Mobile Platform, is on Processor Front Side Bus.
667MHz only.

Anything Dual, no matter Dual 533 or Dual 667 Memory Bandwidth is more than the processor FSB already.

CPU FSB - Chipset - Memory Bus

When CPU FSB is 667MHz,

Memory Bandwidth @ Single Channel 533MHz < CPU FSB
Memory Bandwidth @ Single Channel 667MHz = CPU FSB
Memory Bandwidth @ Dual Channel 533MHz > CPU FSB
Memory Bandwidth @ Dual Channel 667MHz > CPU FSB

The extra 100% memory bandwidth provided by Dual Channel technology will never improve CPU FSB by 100%, get yourself a copy of Sisoft Sandra to have a test.

The minority performance gain at less than 5% is due to the CPU FSB is limited at 667MHz only.

Your theory could be right when ONLY the CPU FSB on Merom is 1066FSB above. Sadly it isn't, and Santa Rosa Platform will be 800MHz FSB limited too.

So, lower latency is slightly faster here, although in real life the difference is too small.

On the other hand, IF single channel, 667MHz will be faster because merely single channel 533MHz couldn't fulfill the processor FSB Limit @ 667MHz.

GO and read again Intel's articles on why was there a need for old Pentium 4 C type processor @ 800FSB needs Dual Channel DDR400 to provide equivalent bandwidth to feed the processor FSB.

And also why a over-supplied PC133 memory bandwidth doesn't help in performance of old Intel Celeron 667MHz with 67MHz FSB.
Last edited by Austin_Goh on Sun Jan 14, 2007 10:21 pm, edited 1 time in total.

pianowizard
Senior ThinkPadder
Senior ThinkPadder
Posts: 8506
Joined: Tue Jun 28, 2005 5:07 am
Location: Ann Arbor, MI
Contact:

#20 Post by pianowizard » Sun Jan 14, 2007 10:12 pm

Austin_Goh wrote:Bottleneck of Core 2 Duo Merom Mobile Platform, is on Processor Front Side Bus.
667MHz only.

Anything Dual, no matter Dual 533 or Dual 667 Memory Bandwidth is more than the processor FSB already.
Even with dual channel, 533MHz is slower than 667MHz. The dual channel mode increases the speed by no more than 10%, some even say no more than 5%. Increasing 533MHz by 5% gives you 560MHz.
Austin_Goh wrote:So, lower latency is slightly faster here, although in real life the difference is too small.
How many times do I have to tell you that 533MHz @ CL4 and 667MHz @ CL 5 both correspond to the same latency, i.e. 7.5 ns?
HP Pavilion x2 12 (Core M3-6Y30, 3.14lb); Microsoft Surface 3 (Atom x7-Z8700, 2.00lb)
Dell OptiPlex 5040 SFF (Core i5-6600); HP 8300 Elite tower (Core i7-3770)
Acer ET322QK, T272HUL; Bolva 40BL00H7; Crossover 404K; Dell 3008WFP, D2015Hf; Monoprice 10734; QNIX QHD2410R; Seiki Pro SM40UNP

Austin_Goh
User with bad email address, PLEASE fix!
Posts: 18
Joined: Sun Dec 17, 2006 9:50 am
Location: Skudai, Johor, Malaysia
Contact:

#21 Post by Austin_Goh » Sun Jan 14, 2007 10:29 pm

pianowizard wrote:
Austin_Goh wrote:Bottleneck of Core 2 Duo Merom Mobile Platform, is on Processor Front Side Bus.
667MHz only.

Anything Dual, no matter Dual 533 or Dual 667 Memory Bandwidth is more than the processor FSB already.
Even with dual channel, 533MHz is slower than 667MHz. The dual channel mode increases the speed by no more than 10%, some even say no more than 5%. Increasing 533MHz by 5% gives you 560MHz.
Austin_Goh wrote:So, lower latency is slightly faster here, although in real life the difference is too small.
How many times do I have to tell you that 533MHz @ CL4 and 667MHz @ CL 5 both correspond to the same latency, i.e. 7.5 ns?
7.5ns? That's latency found on DDR266 or PC133 RAM.
Who told you these?
Which latency value are you refering to?

Use your calculator, 1000/133 = ?

Don't mislead people here, talk is cheap, provide some proof please.
Last edited by Austin_Goh on Mon Jan 15, 2007 11:29 am, edited 1 time in total.

brentpresley
ThinkPadder
ThinkPadder
Posts: 1434
Joined: Fri Dec 23, 2005 4:19 pm
Location: Dallas, TX
Contact:

#22 Post by brentpresley » Sun Jan 14, 2007 10:57 pm

Austin_Goh wrote: Don't mislead people here, talk is cheap, provide some proof please.
Say the SAME thing but much more politely.

LET me be CRYSTAL CLEAR here. This is not [H], Anandtech, XS, or some other forum where you can come in and drop an insult on a senior member and it be ignored.

BE POLITE or get out. Plain and simple.



Your ns calculation is correct, but pianowizard is correct, in more general terms. His calculation was just incomplete.

The ELAPSED time in CL4 @ 533MHz and CL5 @ 667MHz is identical, even though the CYCLES are incremented by one.
Last edited by brentpresley on Sun Jan 14, 2007 10:59 pm, edited 3 times in total.
Custom T60p
2.33GHz 4MB 667MHz Core 2 Duo
4GB PC2-5300 DDR SDRAM
Bluetooth / Atheros ABGN
200GB 7k200 7200RPM Hard Drive
8X DVD Multiburner
15" UXGA - ATI FireGL V5250 (256MB)

http://www.xcpus.com

pianowizard
Senior ThinkPadder
Senior ThinkPadder
Posts: 8506
Joined: Tue Jun 28, 2005 5:07 am
Location: Ann Arbor, MI
Contact:

#23 Post by pianowizard » Sun Jan 14, 2007 10:57 pm

Austin_Goh wrote:7.5ns? That's CAS latency found on DDR266 or PC133 RAM.
Who told you these?
Which latency value are you refering to?

Use your calculator, 1000/133 = ?

Don't mislead people here, talk is cheap, provide some proof please.
Do you know what "CL4" and "CL5" mean? They mean 4 and 5 clock cycles, respectively. At 533MHz, each cycle = 1 sec divided by 533,000,000 = 1.876 ns, and at 667MHz, each cycle = 1 sec divided by 667,000,000 = 1.499 ns. 4 x 1.876 ns = 7.5 ns, and 5 x 1.499 ns = 7.5 ns.

Simple math.
HP Pavilion x2 12 (Core M3-6Y30, 3.14lb); Microsoft Surface 3 (Atom x7-Z8700, 2.00lb)
Dell OptiPlex 5040 SFF (Core i5-6600); HP 8300 Elite tower (Core i7-3770)
Acer ET322QK, T272HUL; Bolva 40BL00H7; Crossover 404K; Dell 3008WFP, D2015Hf; Monoprice 10734; QNIX QHD2410R; Seiki Pro SM40UNP

pianowizard
Senior ThinkPadder
Senior ThinkPadder
Posts: 8506
Joined: Tue Jun 28, 2005 5:07 am
Location: Ann Arbor, MI
Contact:

#24 Post by pianowizard » Sun Jan 14, 2007 10:59 pm

brentpresley wrote:Your ns calculation is correct, but pianowizard is correct, in more general terms. His calculation was just incorrect.
If it's not 7.5 ns, what is it?
HP Pavilion x2 12 (Core M3-6Y30, 3.14lb); Microsoft Surface 3 (Atom x7-Z8700, 2.00lb)
Dell OptiPlex 5040 SFF (Core i5-6600); HP 8300 Elite tower (Core i7-3770)
Acer ET322QK, T272HUL; Bolva 40BL00H7; Crossover 404K; Dell 3008WFP, D2015Hf; Monoprice 10734; QNIX QHD2410R; Seiki Pro SM40UNP

brentpresley
ThinkPadder
ThinkPadder
Posts: 1434
Joined: Fri Dec 23, 2005 4:19 pm
Location: Dallas, TX
Contact:

#25 Post by brentpresley » Sun Jan 14, 2007 11:01 pm

pianowizard wrote:
brentpresley wrote:Your ns calculation is correct, but pianowizard is correct, in more general terms. His calculation was just incorrect.
If it's not 7.5 ns, what is it?
Re-check my post. I meant to say incomplete, not incorrect. :wink:
Custom T60p
2.33GHz 4MB 667MHz Core 2 Duo
4GB PC2-5300 DDR SDRAM
Bluetooth / Atheros ABGN
200GB 7k200 7200RPM Hard Drive
8X DVD Multiburner
15" UXGA - ATI FireGL V5250 (256MB)

http://www.xcpus.com

pianowizard
Senior ThinkPadder
Senior ThinkPadder
Posts: 8506
Joined: Tue Jun 28, 2005 5:07 am
Location: Ann Arbor, MI
Contact:

#26 Post by pianowizard » Sun Jan 14, 2007 11:09 pm

brentpresley wrote:This is not [H], Anandtech, XS, or some other forum where you can come in and drop an insult on a senior member and it be ignored.
But to be fair to Austin_Goh, senior members shouldn't be rude to new members either. I have been a little cranky in the last few days due to my upcoming birthday, so my last few messages addressed to him did have an unpleasant tone. If Austin_Goh felt offended, I apologize.
HP Pavilion x2 12 (Core M3-6Y30, 3.14lb); Microsoft Surface 3 (Atom x7-Z8700, 2.00lb)
Dell OptiPlex 5040 SFF (Core i5-6600); HP 8300 Elite tower (Core i7-3770)
Acer ET322QK, T272HUL; Bolva 40BL00H7; Crossover 404K; Dell 3008WFP, D2015Hf; Monoprice 10734; QNIX QHD2410R; Seiki Pro SM40UNP

Austin_Goh
User with bad email address, PLEASE fix!
Posts: 18
Joined: Sun Dec 17, 2006 9:50 am
Location: Skudai, Johor, Malaysia
Contact:

#27 Post by Austin_Goh » Sun Jan 14, 2007 11:12 pm

pianowizard wrote:
Austin_Goh wrote:7.5ns? That's CAS latency found on DDR266 or PC133 RAM.
Who told you these?
Which latency value are you refering to?

Use your calculator, 1000/133 = ?

Don't mislead people here, talk is cheap, provide some proof please.
Do you know what "CL4" and "CL5" mean? They mean 4 and 5 clock cycles, respectively. At 533MHz, each cycle = 1 sec divided by 533,000,000 = 1.876 ns, and at 667MHz, each cycle = 1 sec divided by 667,000,000 = 1.499 ns. 4 x 1.876 ns = 7.5 ns, and 5 x 1.499 ns = 7.5 ns.

Simple math.
Then again 667MHz RAM is only faster in real-life IF:

- Single Channel applied
- CPU FSB > Memory FSB.

And tell us why those overclockers need lower latency RAM to achieve better performance.

brentpresley
ThinkPadder
ThinkPadder
Posts: 1434
Joined: Fri Dec 23, 2005 4:19 pm
Location: Dallas, TX
Contact:

#28 Post by brentpresley » Mon Jan 15, 2007 8:51 am

Austin_Goh wrote:
pianowizard wrote: Do you know what "CL4" and "CL5" mean? They mean 4 and 5 clock cycles, respectively. At 533MHz, each cycle = 1 sec divided by 533,000,000 = 1.876 ns, and at 667MHz, each cycle = 1 sec divided by 667,000,000 = 1.499 ns. 4 x 1.876 ns = 7.5 ns, and 5 x 1.499 ns = 7.5 ns.

Simple math.
Then again 667MHz RAM is only faster in real-life IF:

- Single Channel applied
- CPU FSB > Memory FSB.

And tell us why those overclockers need lower latency RAM to achieve better performance.
I DON'T need lower latency RAM with C2D to get good performance. To be honest, except in memory benchmarks I can't tell the difference in CL3 and CL8 RAM (in my desktop E6600).

Why? The caches for Core 2 Duo are so freaking big and the memory prefetch so good in the CPU that it really doesn't have a problem keeping the CPU fed. This is why Intel can release a quad core CPU on the same socket, using the same amount of available memory bandwidth, without any significant loss of performance on an IPC level.

Only on AMD systems do you need low latency RAM to really shine. And that is because the caches are much smaller and the system relies on the integrated memory controller instead (kinda like a giant L3 cache).
Custom T60p
2.33GHz 4MB 667MHz Core 2 Duo
4GB PC2-5300 DDR SDRAM
Bluetooth / Atheros ABGN
200GB 7k200 7200RPM Hard Drive
8X DVD Multiburner
15" UXGA - ATI FireGL V5250 (256MB)

http://www.xcpus.com

pianowizard
Senior ThinkPadder
Senior ThinkPadder
Posts: 8506
Joined: Tue Jun 28, 2005 5:07 am
Location: Ann Arbor, MI
Contact:

#29 Post by pianowizard » Mon Jan 15, 2007 9:49 am

Austin_Goh wrote:Then again 667MHz RAM is only faster in real-life IF:

- Single Channel applied
- CPU FSB > Memory FSB.
Please see what I wrote above about Dual Channel, i.e. even with Dual Channel a 533MHz RAM still doesn't exceed the CPU's 667MHz FSB.
Austin_Goh wrote:And tell us why those overclockers need lower latency RAM to achieve better performance.
If you compare two modules of the same speed, shorter latency is always better. For example, PC2700 @ CL2 is better than PC2700 @ CL2.5. When comparing modules of different speeds, it's more complicated, as shown in my calculations.

I suggest that you read this article to get a better understanding of what we've been talking about.

BTW, I just checked the prices of 2GB PC2-4200 vs. PC2-5300 SODIMMs on shop.kingston.com and the difference is pretty small: $589 vs. $640.
HP Pavilion x2 12 (Core M3-6Y30, 3.14lb); Microsoft Surface 3 (Atom x7-Z8700, 2.00lb)
Dell OptiPlex 5040 SFF (Core i5-6600); HP 8300 Elite tower (Core i7-3770)
Acer ET322QK, T272HUL; Bolva 40BL00H7; Crossover 404K; Dell 3008WFP, D2015Hf; Monoprice 10734; QNIX QHD2410R; Seiki Pro SM40UNP

Austin_Goh
User with bad email address, PLEASE fix!
Posts: 18
Joined: Sun Dec 17, 2006 9:50 am
Location: Skudai, Johor, Malaysia
Contact:

#30 Post by Austin_Goh » Mon Jan 15, 2007 11:16 am

Did I say I bought them @ USD1178? I have told many times the 2x 2GB DDR2 533 were bought at USD700. I know the market price well and there is no reason to go for USD1280 for little performance gain (If there is any) on DDR2 667MHz Dual on merely 667MHz CPU FSB.

I even shared this info the thread starter since I can get 2 more pairs if he really wants.

When we are talking about dual channel here, it represents theoretically 8400MB/s ( 2*533MHz*64bit/8 ) or
10600MB/s ( 2*667MHz*64bit/8 ).

Your all are right ONLY IF the Core 2 Duo on mobile has 1066MHz or 1333MHz Front Side Bus.

Bottleneck on Current Merom and Upcoming Santa Rosa Core 2 Duo is the CPU Front Side Bus, that's why no matter is 400/533/667/800 it is, as long as dual channel applied, the effective memory bandwidth obtained is about the same, at the same time lower latency RAM would provide performance gain in real life applications.

CPU to Chipset Front Side Bus Speed is 667MHz.
The Sisoft Sandra I tested:

Single Channel 533MHz ~3330MB/s, by theory should be 4200MB/s
Single Channel 667MHz ~3400MB/s, by theory should be 5300MB/s

Dual Channel 533MHz ~3620MB/s, by theory should be 8400MB/s
Dual Channel 667MHz ~3620MB/s, by theory should be 10600MB/s

The actual bandwidth loss mostly due to:
- Limited CPU FSB @ Merely 667MHz
- Higher Latency on high clocked 667MHz RAM


The actual reason to get 4GB is to build a RAM Drive as a medium to install frequently used programs on it to have a real boost over the performance, bear in mind that HDD is the slowest part on any laptop, the memory frequency or latency we all talking about here cannot solve this bottleneck.

Those involved in argument with me, tell me, did you really try 4GB before? did you use VMWare or RAM Drive? if the answer is no, you don't need to reply me any further as it makes this thread out of topic. I won't reply further too as I sense strong culture in marginalizing junior member in discussion here.

Post Reply
  • Similar Topics
    Replies
    Views
    Last post

Return to “ThinkPad T6x Series”

Who is online

Users browsing this forum: Bing [Bot] and 18 guests