DDR vs DDR2 memory debate

[systemBuilder]

In 2006 (when the T43 was released), I looked at graphics benchmarks for ALL of the T4x series of laptops, and determined that the ATI 9600 of the T42 (SXGA+) was the fastest for 3-D games, by a small margin, over the FireGL GPUs. This led to my decision to purchase a T42 SXGA+ model for my wife (2373-H6U).

The memory bus on the T43 is a DOWNGRADE to the memory bus on the T42, the move to DDR2 memory was done as a cost-saving measure, so generally the T42's had smallest von-neumann bottleneck of the T4x CPUs. The memory for the T43 is way cheaper, but the random access performance is less (the -2 on DDR-2 memory means that each fetch pumps out two rows of RAM, not just one row of RAM, so on paper memory throughput is higher, but in practice the random access performance is lower.)

[RealBlackStuff]

Would you please stop dredging old cows out of the ditch?
Several threads you are 'replying' to are many, many years old!
Please look at the DATE of the last post in a thread, that should give you an indication if this subject is still 'alive'.
I'd suggest: if older than 2-3 months, don't bother...

[systemBuilder]

Would you please stop dredging old cows out of the ditch?
Several threads you are 'replying' to are many, many years old!
Please look at the DATE of the last post in a thread, that should give you an indication if this subject is still 'alive'.
I'd suggest: if older than 2-3 months, don't bother...

So if there is false information in the archives, you prefer to keep the mistakes alive ?? Makes no sense, imho.

[ajkula66]

So if there is false information in the archives, you prefer to keep the mistakes alive ?? Makes no sense, imho.

The info you've posted is debatable at best.

Lot of presumptions on your part and no hard data to back it up.

Any benchmark test of 9600 vs. V3200 will have latter as the clear winner. Period.

DDR2 wasn't cheap at the time (03/2005) when T43 was released.

And so forth and so on...

[systemBuilder]

Nonsense, there are 3 disadvantages of the T43, and only one advantage, and these were widely recognized when that machine was foisted onto the world by IBM:

(a) The machine runs far hotter and noiser than the earlier T42, T41, and T40 machines.
(b) The machines incorporate hard-drive identification checks in the BIOS, making 3rd-party drives difficult to install (must flash a jailbroken BIOS)
(c) The memory bus architecture (DDR2) is a cost savings measure and the performance at equivalent clock speeds is less than T42 and other similar Dothan thinkpads.

The one and only advantage of the T43 is that the CPU revision includes the XD (execute disable) bit, and therefore, they can run Windows-8. You might consider that a disadvantage, considering the popularity of Windows-8, and considering the quote from an MIT professor, "Windows-8 makes the perfect Christmas gift for anyone that you HATE."

[RealBlackStuff]

@systembuilder:
why don't you stay with repairing bikes, rather than write your nonsense here all the time?

[ajkula66]

Nonsense, there are 3 disadvantages of the T43, and only one advantage, and these were widely recognized when that machine was foisted onto the world by IBM:

(a) The machine runs far hotter and noiser than the earlier T42, T41, and T40 machines.

That was true when the platform was released. EC version 1.05 and above have taken care of this. Installing a M10 fan from the T42p instead of the original M24 also helps, if one is *really* sensitive to fan noise.

(b) The machines incorporate hard-drive identification checks in the BIOS, making 3rd-party drives difficult to install (must flash a jailbroken BIOS)

Absolutely NOT.

I've installed anything from XP to Vista to W7 and W8 - not to mention various Linux distros - on "non-approved" hard drives. Yes, the machine gives a warning but one either presses "esc" key and continues (on BIOS versions older than 1.24 I believe) or set the BIOS up (on newer versions) to automatically continue booting after the warning has been displayed. Modded BIOS is only required if one wants to get rid of the warning altogether and/or wants to install a third-party wireless card.

(c) The memory bus architecture (DDR2) is a cost savings measure and the performance at equivalent clock speeds is less than T42 and other similar Dothan thinkpads.

No.

The necessity of enabling T43/R52 to run in a dual-channel mode was the reason for this change, and they will outperform any prior ThinkPad, no matter how you set the meters.

The real question is how much any of that matters nowadays, since the final examples of the T4x series rolled off the assembly line at the end of 2006 and are nearing obsolescence whether we like it or not....

The one and only advantage of the T43 is that the CPU revision includes the XD (execute disable) bit, and therefore, they can run Windows-8. You might consider that a disadvantage, considering the popularity of Windows-8, and considering the quote from an MIT professor, "Windows-8 makes the perfect Christmas gift for anyone that you HATE."

While I'm not going to go into the whole W8 debate (no I don't like the OS in question but that's beside the point of this thread) the fact of the matter is that newer Ubuntu-based distros also have a PAE requirement, and will not run on pre-R52/T43/X41 ThinkPads.

The fact of the matter also is that a PM780 outperforms PM765 and still ranks marginally higher on the CPU lists.

The fact of the matter also is that the failure rate of T43/p platform is *significantly* lower when compared to any of the previous T4x models.

My advice to the OP: get a M10 fan and enjoy the T43. No big investment - or hard labour in replacing it - is involved.

[systemBuilder]

@systembuilder:
why don't you stay with repairing bikes, rather than write your nonsense here all the time?

Haha RealBlackStuff, do you have a problem with a type of information known as "FACTS" ???

If you were around in 2005/2006 when the T43's were being released, I think you would remember the hue and cry from the disadvantages of these machines. That's a type of information known as a "Fact" and you can search the archives to confirm it as a "Fact".

I stand by my assertion. A T43 is nothing but a hot noisey T42. In benchmarks, the graphics cards of all T43's are beaten by the ATI 9600 on the T42. That is why i paid $2200 to get my wife a T42 in 2006.

[ajkula66]

If you were around in 2005/2006 when the T43's were being released, I think you would remember the hue and cry from the disadvantages of these machines. That's a type of information known as a "Fact" and you can search the archives to confirm it as a "Fact".

The problem with the stated facts is that they date back to 2006, and a lot has changed since then. We're in 2013 right now. BIOS/EC versions have been updated a lot since the initial release of T43 platform, and most of the initial teething problems were successfully addressed since...

I stand by my assertion. A T43 is nothing but a hot noisey T42. In benchmarks, the graphics cards of all T43's are beaten by the ATI 9600 on the T42. That is why i paid $2200 to get my wife a T42 in 2006.

What benchmarks would these be, since I'd love to run them? And I am in possession of a very nice T42 with Radeon 9600 right now...

Freestone, which is one of the rare serious benchmarks that will run on cards of this era, shows T43p's V3200 as clearly superior to T42p's Fire GL T2, as I've proven here:

http://www.thinkpads.com/forum/viewtopi ... 12&t=90221

It even beat the X1300 seen in T60 series...of course, throw in a V5250 and it will beat the hell out of V3200, no questions asked...

[RealBlackStuff]

Haha RealBlackStuff, do you have a problem with a type of information known as "FACTS" ???

If you were around in 2005/2006 when the T43's were being released, I think you would remember the hue and cry from the disadvantages of these machines. That's a type of information known as a "Fact" and you can search the archives to confirm it as a "Fact".

I stand by my assertion. A T43 is nothing but a hot noisey T42. In benchmarks, the graphics cards of all T43's are beaten by the ATI 9600 on the T42. That is why i paid $2200 to get my wife a T42 in 2006.

I'm sure I was already around before you were even born!
And I have been working on and with computers since 1968.
I don't mind facts, if that's what they really are, but obviously YOU are still living in 2006.
The absolute rubbish you post...

[systemBuilder]

The memory bus on the T43 is a DOWNGRADE to the memory bus on the T42, the move to DDR2 memory was done as a cost-saving measure, so generally the T42's had smallest von-neumann bottleneck of the T4x CPUs. The memory for the T43 is way cheaper, but the random access performance is less (the -2 on DDR-2 memory means that each fetch pumps out two rows of RAM, not just one row of RAM, so on paper memory throughput is higher, but in practice the random access performance is lower.)

Just so you know that the T43 thinkpad is a cheaper, slower, cost-saving machine compared to a T42 thinkpad, here are the two web pages that make my point perfectly clear :

http://en.wikipedia.org/wiki/DDR_SDRAM
http://en.wikipedia.org/wiki/DDR2_SDRAM

DDR SDRAM used in a T42 (PC-2700) is DDR-333 and has a cycle time of 6 ns and a peak data rate of 2666 MB/s. DDR2 SDRAM (PC2-4200) used in a T43 has a cycle time of 7.5 ns. What that means is that that the random access performance is 13% slower on a T43 than on a T42. There's a reason why it's called RANDOM access memory - because the RANDOM access speed is the main determinant of performance. DDR2 is a cheaper, slower type of RAM. To make up for a loss of RANDOM access performance, the RAM tries to compensate by transfering 60% more data (at peak transfer rate). However, in most benchmarks a DDR2 machine @ 2 Ghz and a DDR machine @ 2 Ghz, the DDR machine will beat the DDR2 machine.

In memory performance, cycle time is king. Cycle time is the time it takes to recharge the row and column sense amps so that a new cell of RAM can be read.

-

[systemBuilder]

(c) The memory bus architecture (DDR2) is a cost savings measure and the performance at equivalent clock speeds is less than T42 and other similar Dothan thinkpads.

No.

The necessity of enabling T43/R52 to run in a dual-channel mode was the reason for this change, and they will outperform any prior ThinkPad, no matter how you set the meters.

The real question is how much any of that matters nowadays, since the final examples of the T4x series rolled off the assembly line at the end of 2006 and are nearing obsolescence whether we like it or not....

The one and only advantage of the T43 is that the CPU revision includes the XD (execute disable) bit, and therefore, they can run Windows-8. You might consider that a disadvantage, considering the popularity of Windows-8, and considering the quote from an MIT professor, "Windows-8 makes the perfect Christmas gift for anyone that you HATE."

While I'm not going to go into the whole W8 debate (no I don't like the OS in question but that's beside the point of this thread) the fact of the matter is that newer Ubuntu-based distros also have a PAE requirement, and will not run on pre-R52/T43/X41 ThinkPads.

The fact of the matter also is that a PM780 outperforms PM765 and still ranks marginally higher on the CPU lists.

The fact of the matter also is that the failure rate of T43/p platform is *significantly* lower when compared to any of the previous T4x models.

My advice to the OP: get a M10 fan and enjoy the T43. No big investment - or hard labour in replacing it - is involved.

Just so you know that the T43 thinkpad is a cheaper, slower, cost-saving machine compared to a T42 thinkpad, here are the two web pages that make my point perfectly clear :

http://en.wikipedia.org/wiki/DDR_SDRAM
http://en.wikipedia.org/wiki/DDR2_SDRAM

DDR SDRAM used in a T42 (PC-2700) is DDR-333 and has a cycle time of 6 ns and a peak data rate of 2666 MB/s. DDR2 SDRAM (PC2-4200) used in a T43 has a cycle time of 7.5 ns. What that means is that that the random access performance is 13% slower on a T43 than on a T42. There's a reason why it's called RANDOM access memory - because the RANDOM access speed is the main determinant of performance. DDR2 is a cheaper, slower type of RAM. To make up for a loss of RANDOM access performance, the RAM tries to compensate by transfering 60% more data (at peak transfer rate). However, in most benchmarks a DDR2 machine @ 2 Ghz and a DDR machine @ 2 Ghz, the DDR machine will beat the DDR2 machine.

In memory performance, cycle time is king. Cycle time is the time it takes to recharge the row and column sense amps so that a new cell of RAM can be read.

[ajkula66]

However, in most benchmarks a DDR2 machine @ 2 Ghz and a DDR machine @ 2 Ghz, the DDR machine will beat the DDR2 machine.

Care to post any benchmarks to prove this point? We're still waiting for the ones for the graphics card as well...

[ajkula66]

However, in most benchmarks a DDR2 machine @ 2 Ghz and a DDR machine @ 2 Ghz, the DDR machine will beat the DDR2 machine.

Benchmarks, please...

[Adda]

DDR SDRAM used in a T42 (PC-2700) is DDR-333 and has a cycle time of 6 ns and a peak data rate of 2666 MB/s. DDR2 SDRAM (PC2-4200) used in a T43 has a cycle time of 7.5 ns. What that means is that that the random access performance is 13% slower on a T43 than on a T42.

In memory performance, cycle time is king. Cycle time is the time it takes to recharge the row and column sense amps so that a new cell of RAM can be read.

-

So you claim DDR-333 has a peak data rate of 2666MB/s, sounds reasonable as Everest Home edition claims a Pentium M 1.5GHz with DDR333 has a bandwidth of 2470MB/s.
My T43p has a memory bandwidth of 3184MB/s@ 2.13GHz, I can't run it at 1.5, so lets give the old platform the benefit of the doubt and run my T43p at 1466Mhz, this gives a bandwidth of 2981MB/s...

Alright you said at 2Ghz DDR333 should be faster, so get Everest Home Edition 2.20 and test for your self.
At 2GHz my T43p scores 3122MB/s, how about your T42's?

But bandwidth isn't everything, latency also has a small impact on performance, Everest claims a 1.5GHz Pentium M with DDR-333 CL2.5-3-3-7 has a latency of 101ns, at 1466Mhz with DDR2-533 CL4-4-4-12 has a latency of 107.4ns, at 2Ghz 105.6ns.

So SystemBuilder, get Everest here:
http://www.softpedia.com/get/System/Sys ... tion.shtml

And back up your claims.

Video performance for various ThinkPads, including R9600 FGL T2 and V3200, is covered here:
http://forum.thinkpads.com/viewtopic.php?f=12&t=90221

[Cigarguy]

Memory, it and of itself, do not work in isolation. Bus speed, CPU speed, software, chipset, HDD/SSD, etc will affect the overall machine. My road bicycle wheels is capable of 80 km/h, lighter than and have less drag then my car wheels. However, taken as a whole, my car is significantly faster.

[RealBlackStuff]

This systemBuilder (yeah right...) has nothing better to do than reply to old posts, that nobody is interested in any more anyway.
When does this guy stop spreading nonsense and FUD?

[systemBuilder]

"DDR SRAM is no slower than DDR2 SRAM, but DDR2 memory is cheaper" - Toms Hardware.

I reset my case. The rest of you whiners who post no sources, read MINE (my SECOND source) and weep ...

http://www.tomshardware.com/reviews/gen ... 63-11.html

[systemBuilder]

Toms hardware backs me up ... DDR2 is simply a "cost reduced, slower-cycletime" memory compared to DDR ram (see wikipedia to learn how much slower the cycle time is in DDR2 memory).

And you are backed up by ?? Nobody. Simply, Nobody.

http://www.tomshardware.com/reviews/gen ... 63-11.html

[Cigarguy]

^^^LOL. What a joke! Yes DDR2 is cheaper than DDR but get this, DDR3 is cheapest yet.

You call that proof? A single early comparison of DDR vs DDR2 on a obscure MB running DDR2 desktop memory that is running "pseudo-synchronously"? Based on this you extrapolate that memory in a T42 is faster than T43?

Since we are in the realm of extrapolation here's one for you. My bicycle is faster than a Toyota Camry. This is not a false statement. I did pass a Camry the other day on my bicycle who was going about 10 km/h.....because he had a flat tire.

[ajkula66]

"DDR SRAM is no slower than DDR2 SRAM, but DDR2 memory is cheaper" - Toms Hardware.

I reset my case. The rest of you whiners who post no sources, read MINE (my SECOND source) and weep ...

http://www.tomshardware.com/reviews/gen ... 63-11.html

Oh my...

AMD wrote:

Why use DDR2?

There are several reasons why DDR2 should be considered. First, the required voltage is lower and reduces the system power demanded by the memory. Second, the increased pre-fetch produces a reduced core speed dependency for better yields. Third, DDR2 allows migration of higher bus speeds. Fourth, its write latency provides an improvement in command bus frequency. Finally, DDR2 provides additional features that improve the overall performance and effectiveness of the memory.

source:http://www.techsupportforum.com/forums/ ... 32275.html

Crucial wrote:

The Advantages of DDR2 Memory


DDR2 is a new memory architecture that supports improvements in chipsets and other system components.

DDR2 maximizes system performance and can operate at data rates of 667MHz and beyond.

DDR2-based dual-channel systems process memory more efficiently and reduce system latency time.

DDR2 better handles the memory-intensive applications used on today's faster, feature-packed systems.

Lower power consumption improves server performance and prolongs notebook battery life.

Lower operating voltage of 1.8 volts (compared to 2.5 volts for DDR and 3.3 volts for SDRAM).

Lower operating temperatures help notebooks and servers with cooling and required airflow.

source: http://www.crucial.com/kb/answer.aspx?qid=4081

And some early, real-life benchmarks here:

http://www.legitreviews.com/article/212/3/

Enough already.

[Adda]

"DDR SRAM is no slower than DDR2 SRAM, but DDR2 memory is cheaper" - Toms Hardware.

I reset my case. The rest of you whiners who post no sources, read MINE (my SECOND source) and weep ...

http://www.tomshardware.com/reviews/gen ... 63-11.html

In all seriousness though, some reviewers statement does not falsify actual data, you can say what ever you like, you can cherry pick articles, but that doesn't change the numbers.

[Adda]

Toms hardware backs me up ... DDR2 is simply a "cost reduced, slower-cycletime" memory compared to DDR ram (see wikipedia to learn how much slower the cycle time is in DDR2 memory).

And you are backed up by ?? Nobody. Simply, Nobody.

http://www.tomshardware.com/reviews/gen ... 63-11.html

I back him up, so does the numbers.

The reason why CL are higher on DDR2, is because is runs at higher frequencies, do you think DDR3 is slow due to CL latencies in double digits or what?
The higher CL latency is compensated for by a higher clock rate, and is a non issue, and the chipset has a higher impact on latencies the CL does.

Look at it like this, if I pick an apple every two clock cycles at a frequency of 100, I pick 50 apples.

If I pick an apple every four clock cycles at a frequency of 200, I still pick 50 apples.

[rkawakami]

...(the -2 on DDR-2 memory means that each fetch pumps out two rows of RAM, not just one row of RAM, so on paper memory throughput is higher, but in practice the random access performance is lower.)

Sorry, but you may be confused about this. DDR2, like DDR before it, transfers data at both the rising and falling edges of the system clock. DDR means "Double Data Rate". This is in contrast to the earlier SDR, or Synchronous DRAM, which only transferred data at the rising edge of the clock. Additionally, data access in synchronous (or burst) mode is along the same row, which means that data is coming from consecutive columns in sequential mode or staggered columns in interleaved mode. For any synchronous DRAM, the memory is presented with a single ROW address, a starting COLUMN address and then data is written or read out according to the burst mode that been programmed by the MRS (Mode Register Set) commands.

The main differences between DDR and DDR2 have to do with operating voltages and internal architecture changes to the memory to allow for increased bandwidth (internal clock runs at half data clock).

While it may be true that DDR2 has a longer latency time over DDR when single word, random data access is considered, that is, a burst length (BL) of 1, the most effective and typical uses of DDR memories is when they use BL of 4 or 8.

[axur-delmeria]

My 2 cents:

The lower latency of DDR1 is a factor, but only when compared to DDR2 at the same frequency: e.g. comparing DDR1-400 (PC3200) and DDR2-400. In this case, DDR1 has a slight edge.

The T43 uses DDR2-533 while the T42 has DDR1-333.
In this case, the much higher bandwidth of DDR2 generally offsets any latency advantage of DDR1.

[rkawakami]

Combined off-topic posts from "Possible to swap T42 & T43 LCD assembly?" thread in T4x forum into this thread.

@systembuilder: Keep posts on topic and don't dredge up old threads in an attempt to re-state your opinion concerning this DDR vs. DDR2 debate. I'd also suggest that you provide links to those relevant benchmarks you've quoted (and have been asked several times now to produce).

@everybody: No personal attacks; keep it civil.

[Temetka]

Sorry, but you may be confused about this. DDR2, like DDR before it, transfers data at both the rising and falling edges of the system clock. DDR means "Double Data Rate". This is in contrast to the earlier SDR, or Synchronous DRAM, which only transferred data at the rising edge of the clock. Additionally, data access in synchronous (or burst) mode is along the same row, which means that data is coming from consecutive columns in sequential mode or staggered columns in interleaved mode. For any synchronous DRAM, the memory is presented with a single ROW address, a starting COLUMN address and then data is written or read out according to the burst mode that been programmed by the MRS (Mode Register Set) commands.

The main differences between DDR and DDR2 have to do with operating voltages and internal architecture changes to the memory to allow for increased bandwidth (internal clock runs at half data clock).

While it may be true that DDR2 has a longer latency time over DDR when single word, random data access is considered, that is, a burst length (BL) of 1, the most effective and typical uses of DDR memories is when they use BL of 4 or 8.

And now we get into the nuts and bolts of it all. I came in here to post pretty much what you just said in regards to data transfer on both leading and trailing edges of the clock cycle.