Why does Dell put slower memory in their systems ?

Status
Not open for further replies.
Y

youdontwannaknow

Joined
Dec 18, 2006
Messages
588
Location
CA
I don't know if all manufacturers do this or not but I have noticed that the memory that Dell provides in their desktops and laptops is usually slower than the FSB of the supplied cpu.

for example if the max FSB is 400Mhz they will ship the 333Mhz memory with it. They will always ship memory that is a step or two slower than the FSB. Why in the world would they do such a thing (other than a sly cunning way of saving cost) ?

Wouldn't this mean they are cheating the customer ? I mean how is the 'average' unsuspecting customer to know, and that too with cryptic terminology (pc-xxx, pc2-xxx, ddr, ddr2, ddr3, double, quad, etc.etc..) what they are really getting ?

I haven't done any research to see if other leading manufacturers do this or not but I am sure they must be.
 
Usually OEMs stick with the CPU FSB to Memory speed ratio at 1/1. Although this isn't always the case. In many cases the rated memory speed is often higher than the actual FSB.

P4/Core2, etc (basically all of Intels processors except Nehalem) used whats called a "quad pumped" fsb. To get the actual FSB you divide by 4.
DDR2 (DOUBLE data rate) advertised memory speed is actually 2x the "real" speed. So to get the real memory speed of the memory in your computer you divide by 2.

So for example an Core 2 E8500 is rated at 1333mhz fsb. The actual fsb is 1333/4 = 333.25mhz.

Most OEMs are probably shipping 800 or 667mhz DDR2.
800mhz is /2 =400mhz
667mhz is /2 =333.5mhz

In both of the scenarios above your memory is not "slower" than your CPU FSB.

Its all advertising.
 
Last edited:
Originally Posted By: astraelraen

So for example an Core 2 E8500 is rated at 1333mhz fsb. The actual fsb is 1333/4 = 333.25mhz.

Most OEMs are probably shipping 800 or 667mhz DDR2.
800mhz is /2 =400mhz
667mhz is /2 =333.5mhz

In both of the scenarios above your memory is not "slower" than your CPU FSB.

Exactly. The Dell Vostro I bought last year came with Core2Duo E8200 (1333 FSB) and DDR 667MHz memory, so it's a match, IMO. Since Dell makes it practically impossible to overclock your CPU, getting faster RAM would be a waste.
 
The OP mentioned 400FSB, so it is already talking about the actual speed that is de-quad pumped (quarter pump?) from the CPU FSB, but ship a memory that is running at a lower speed than that (333 instead of 400).

I think the problem is that CPU come with the factory with a particular speed and FSB/multiplier. It is not that DELL can change this configuration by running a 3.2GHZ CPU at 333 instead of 400. These multipliers are locked and you are only going to slow the CPU by running it at a lower FSB.

Why don't they ship with 400 (1600) memory instead? Have you check how expensive are those? Those are the performance/gaming memory for people that overclock, and benchmark shows that it doesn't affect performance by more than 1-2%. FYI you need at least 5-10% to feel a difference, and people usually don't upgrade until the difference is at least 100-200% more than the old system.

The only people who buy these exotic memory are the ones that overclock and have reach the maximum and is limited by the RAM. It is sort of the equivalence of an all metal radiator, stronger valve springs, larger fuel injectors, etc. It doesn't do anything to a stock car, but would bring out the potential of a modified race car.
 
It's not just Dell, they all do that unless you specify faster RAM at purchase/build time.

I have countless HP & Compaq "business" desktops where they use the cheapest & slowest memory possible.
 
Originally Posted By: Dan55
The computer is only as fast as it's slowest component, so it might not matter.


+1 on that, too.
whistle.gif
 
Originally Posted By: uc50ic4more
Originally Posted By: Dan55
The computer is only as fast as it's slowest component, so it might not matter.


+1 on that, too.
whistle.gif



No. If you have been looking at the benchmark, the memory speed is not that big of a factor anymore. These processors have large caches that are optimized for slower memory.

I'd say the difference in performance between 1333 and 1600 memory is only 2% max, between 800 and 1333 is probably 2-3%, Way lower than the difference in CPU speed.
 
sorry, by "fsb" I meant the actual dual/quad data rate. So if the FSB of a cpu is 200Mhz then wouldn't the optimum memory to go with it would be a DDR3200(400Mhz) ? And if you use, say 333Mhz, then I suppose it could hurt performance in one of the following ways ? -

If the multiplier is fixed you not only get lower cpumemory data rate but lower actual cpu speed as well. Double whammy.

If the cpu/bios automatically adjusts the multiplier to work at a fixed fsb (and cpu frequency) you are still loosing the cpumemory speed.

The 'slowest component' theory does not apply in this case. What component, let alone slowest component, affects cpumemory data rate ? I also dont understand the 2-3% part. Wouldn't the drop in bandwidth would be about ~16% between 333Mhz and 400Mhz ?
 
Last edited:
The Hard Drive is by far the largest bottleneck in a computer.

Build your own computer....problem solved.

I remember ordering an HP desktop a few years ago and it came with a 5400rpm hard drive instead of a 7200rpm. They also used (2) 256mb instead of (1) 512mb stick.
 
Originally Posted By: GMFan
I remember ordering an HP desktop a few years ago and it came with a 5400rpm hard drive instead of a 7200rpm.

Those older 5400 rpm drives also had relatively small cache (2MB?). The 7200 rpm drive you replaced it with probably had at least 8 MB - that makes a noticeable difference in and by itself.

Quote:

They also used (2) 256mb instead of (1) 512mb stick.

Does that really make a lot of difference in performance? I always wondered about that...
 
QP, for DDR RAM, yes, it makes a difference. Our informal testing showed about 25% difference in speed with copying system images down to a computer when we built them. After we saw that, we insisted that any system with DDR had matched pairs.

Any truth that newer computers are faster with two sets of matched pair sticks rather than a single set? We have been ordering them with 4x1GB modules rather than 2x2GB sticks due to price. Just wondering if there's any performance increase by having 4 RAM modules compared to two...
 
IF the memory controller on your motherboard has "Dual Channel" capability then yes using two sticks of ram instead of one is much better. Theoritically double the bandwidth.

GMFan, its called thread-jacking. Can we get back on topic please ? And we are talking about the optimal cpumemory performance, and not that of the whole "computer".
 
Last edited:
Originally Posted By: youdontwannaknow
If the multiplier is fixed you not only get lower cpumemory data rate but lower actual cpu speed as well. Double whammy.


Originally Posted By: youdontwannaknow
If the cpu/bios automatically adjusts the multiplier to work at a fixed fsb (and cpu frequency) you are still loosing the cpumemory speed.


No, there is a 5:4 or 4:3 ratio option, it is asynchronous and will lose some performance (~1% or so due to latency) compare to sync, but not by much.

Originally Posted By: youdontwannaknow
The 'slowest component' theory does not apply in this case. What component, let alone slowest component, affects cpumemory data rate ? I also dont understand the 2-3% part. Wouldn't the drop in bandwidth would be about ~16% between 333Mhz and 400Mhz ?

You are assuming that every CPU calculation needs to go to the RAM for data IO. But CPU is running at much higher speed and no matter how fast your DRAM is, it is still too slow. CPU has cache internally that kept 90%+ of the most recently used data/instruction so it only goes out for new data when it has never been read in the recent past. This is the main and most important difference between the celeron, pentium, core series of CPU. The execution unit is the same but the cache size is different, and that makes the performance difference 20% or more, much bigger than the speed of your RAM, for the same CPU speed it is run on.

It takes something like 120+ CPU cycle to read the DRAM in one cycle, and you have something like 1MB (Pentium E) to 6MB (Core2Dual) of cache on the CPU vs 2GB on your DRAM.
 
Originally Posted By: PandaBear
No, there is a 5:4 or 4:3 ratio option, it is asynchronous and will lose some performance (~1% or so due to latency) compare to sync, but not by much.


Sure, but what about the inherent reduction in the data rate (400 v/s 333) to begin with ? What makes up for that ?
Looks like you are saying that the more deeper the L1 data cache and L2 caches are, less is the difference in ‘effective’ bandwidth of different memory speeds ?
But if you don’t have a decent sized cache then you’re screwed, right ? and secondly what is the optimal cache size then that works effectively for all sorts of programs(memory requests and operations) ?
 
Last edited:
Originally Posted By: youdontwannaknow
Sure, but what about the inherent reduction in the data rate (400 v/s 333) to begin with ? What makes up for that ?
Looks like you are saying that the more deeper the L1 data cache and L2 caches are, less is the difference in ‘effective’ bandwidth of different memory speeds ?
But if you don’t have a decent sized cache then you’re screwed, right ? and secondly what is the optimal cache size then that works effectively for all sorts of programs(memory requests and operations) ?


The more cache the better, if you don't want to pay and get the loss leader celeron and pentium E, that's what you have to expect. Nobody forces you to buy 333 memory, you can always upgrade to 400 when you order or make your purchases, just pay for it. IMO that's splitting hair, but if you are that irritated about small performance difference, PAY FOR IT.
 
Last edited:
http://www.xbitlabs.com/articles/memory/display/core2duo-memory-guide_12.html#sect0

Conclusion

We've already given our general recommendations about choosing system memory for Core 2 Duo systems in the course of this review. So, you should be aware by now what characteristics of DDR2 SDRAM modules are to be taken into consideration when you are building your computer with a new Intel processor with the Core micro-architecture. In case you've missed something, here are the main points.

First of all we have to acknowledge the high efficiency of data pre-fetch algorithms implemented in Core 2 Duo processors. It is thanks to them that platforms with such processors have data access latency comparable to that of Socket AM2 Athlon 64 systems, which feature an integrated memory controller. However, notwithstanding the impressive achievement of Intel's engineers, the memory subsystem of Core 2 Duo systems with an external memory controller, located in the chipset's North Bridge, cannot match the memory subsystem of Socket AM2 systems in overall efficiency. To be exact, the platforms with new Intel processors cannot provide as high memory bandwidth as the competing platforms do.

The memory bandwidth on Core 2 Duo systems is limited not by the characteristics of DDR2 SDRAM modules but by the bandwidth of the bus that connects the CPU with the chipset's North Bridge. This is why your changing the memory frequency or timings is going to have a small effect on performance of a non-overclocked Core 2 Duo system (but the frequency influences performance somewhat more than the timings do).

More interesting are the results of the overclocked platform. In this case, there is more sense in using fast memory and the optimal memory frequency divisor is 1:1 (FSB:DRAM) as has been shown in our tests. In other words, you can achieve maximum performance by using memory with lowest possible timings in synchronous mode. It means that if you overclock the FSB to 400MHz, DDR2-800 SDRAM with low timings is the optimal choice. If the FSB is overclocked more, DDR2-1000 or DDR2-1067 SDRAM is the best option. An additional argument in favor of using memory and the FSB in synchronous mode at overclocking is that the 1:1 divisor is the most stable one on a majority of mainboards.
 
Last edited:
Status
Not open for further replies.

Similar threads

buster
  • Locked
Engine Break-In - Found the answer
3 4 5
Replies
96
Views
19K
ChristianReske
K
  • Locked
HP vs DELL vs MAC - my recent experience
2 3
Replies
43
Views
16K
PandaBear
B
  • Locked
Economy Laptops: Dell Inspiron 1545 vs Lenovo G550
2
Replies
25
Views
14K
crinkles
Back
Top