Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
667mhz RAM with 800mhz FSB
- Thread starter prism
- Start date
- Sort by reaction score
FSB is between the CPU and the rest of the system.
Santa Rosa only has a 667mhz memory controller.
Well, according to the IBM T61 buyers their machines DO support 800MHz DDR2 (as in the system gets faster aswell), I guess it might be that no 800MHz ram was available when Intel wrote their spec pages.
Could anyone try with 800MHz ram?
Are this a fact proven beyond all doubt?
The fact is simple: Intel placed a 667MHz memory controller into the current SR chipset. This means that even if the chipset supports 800MHz RAM, your RAM will still be limited to 667MHz throughput, so your question is irrelevant. You will just be wasting your money. This is nothing new with Intel. Hopefully the SR revision that coincides with the initial release of Penryn will boost the memory controller to 800 MHz.
PS--This was also covered in other threads (e.g. MBP predictions and New MBP threads).
The FSB is the CPUs connection to the outside world. The memory is only a part of the outside world. On top of that, SR has a dual channel memory controller, which means, that both SODIMMS are access simultaniously.
If you count that together, you'll see that the 667MHz should be ok.
Jochen
If you count that together, you'll see that the 667MHz should be ok.
Jochen
No, I see the OP's point. After all some body has managed to do just that with a Thinkpad. Not only that, it is completely unsubstantiated with any forum post or benchmark showing the result. And we as always are supposed to believe Intel on this one.
How do we know Intel did indeed put in a 667mhz memory controller? They could have put in a 800mhz one but due to constraints in the 800mhz memory market they hide the fact to prevent slowdowns in their sales.
Goodness! Conspiracy!
That does not explain why it would be faster.
Because some anonymous and unnamed person said it is indeed faster, it must be true!
If the IBM T61 uses the mobile 965 chipset, the memory will be limited to 667MHz tops. That is Intel's specs for the chipset. Maybe someone, even IBM themselves, have hacked it but I certainly wouldn't trust running a hacked computer like that.
Have you not noticed that you are trusting Intel's spec sheet? Who is to said that they actually *gasp* lied?
My take:
Granted, 800MHz memory, assuming the MCH supported it, which I'm not convinced of given I've never seen anything from Intel suggesting it would be true, would be better, since even single-channel mem accesses would be at "full rate", and of course, synchronous memory is always more efficient as you never have a wait-state while you wait for the next cycle...
Having said that, I suspect the increased power drain isn't worth it. DDR3 will "fix" that by using lower voltages for the RAM so it'll all be moot in 6-8 months time anyway -- once Montevina launches...
I've read that ThinkPad site but I'm just not convinced. He claimed Z-CPU (a third party Windows system config tool) doesn't lie but there's no screenies etc. -- and it's pretty easy to mis-read it, since it would display the SPD data showing the DIMMs were 800MHz, even if they were running at 667MHz...
I'm not saying SR definitely doesn't support 800MHz though. I've been bitten that way before
Having said that, the improvement would be pretty damn small, especially given the difference in cost.
- Intel says the MCH (memory hub controller) only runs at 667MHz
- 800MHz RAM would work fine on a controller that accepts 667MHz. It'll run at 667MHz, and the latency should be better than regular 667MHz RAM.
- The FSB is between the CPU and the northridge, the mem speed is the bus between the memory and the MCH (which is on the northbridge), ie CPU <===> Northbridge <===> DIMMs (much simplified)
- Assuming "matched" DIMMs, the MCH (memory controller hub) is dual channel, so it has the potential to run at 2x667MHz
- 2*667MHz > 800MHz, so there is a gain with the faster FSB when needed... and when not, it clocks down to 400MHz to save power
- I guess overclocking is also possible, however unlikely
Granted, 800MHz memory, assuming the MCH supported it, which I'm not convinced of given I've never seen anything from Intel suggesting it would be true, would be better, since even single-channel mem accesses would be at "full rate", and of course, synchronous memory is always more efficient as you never have a wait-state while you wait for the next cycle...
Having said that, I suspect the increased power drain isn't worth it. DDR3 will "fix" that by using lower voltages for the RAM so it'll all be moot in 6-8 months time anyway -- once Montevina launches...
I've read that ThinkPad site but I'm just not convinced. He claimed Z-CPU (a third party Windows system config tool) doesn't lie but there's no screenies etc. -- and it's pretty easy to mis-read it, since it would display the SPD data showing the DIMMs were 800MHz, even if they were running at 667MHz...
I'm not saying SR definitely doesn't support 800MHz though. I've been bitten that way before
Having said that, the improvement would be pretty damn small, especially given the difference in cost.
If that is indeed the case why do ram manufacturers still bother to make 800mhz sodimms? Nobody is going to buy it since... 1) laptops are not overclockable by and large 2) nobody has the hardware to exploit it since the upcoming chipsets will use DDR3
Yes, the fact that according to aliquis-, there exists one Thinkpad user who got improved performance when using 800mhz ram modules!
The 800MHz DIMMs were available *before* SR was.
They're probably using chips that have also been used in non-SO-DIMMs.
You can often buy DIMMs that far exceed the specs of current mem controllers.
There's a fairly large market of people who will pay top dollar for memory etc. that is "high performance" for whatever definition you prefer. Ever looked at the audio cable market? Snakeoil doesn't begin to describe it...
I hadn't missed that -- I did note that 800MHz DIMMs should exhibit better latency than 667MHz ones, and that is by any definition better performance, especially if the latency dif is enough to stop data regularly causing a wait-state to occur.
However, I'm not convinced by a single anecdotal report.
I don't see why Intel wouldn't claim their controller supported 800MHz memory if it did. I just don't see what's in it for them. It's possible it works but it's not considered "in spec" of course... but that would suggest that it has the potential for failing.
E6600's (Conroe) overclock by 50% too ...
The 800MHz DIMMs were available *before* SR was.
They're probably using chips that have also been used in non-SO-DIMMs.
You can often buy DIMMs that far exceed the specs of current mem controllers.
There's a fairly large market of people who will pay top dollar for memory etc. that is "high performance" for whatever definition you prefer. Ever looked at the audio cable market? Snakeoil doesn't begin to describe it...
I hadn't missed that -- I did note that 800MHz DIMMs should exhibit better latency than 667MHz ones, and that is by any definition better performance, especially if the latency dif is enough to stop data regularly causing a wait-state to occur.
However, I'm not convinced by a single anecdotal report.
I don't see why Intel wouldn't claim their controller supported 800MHz memory if it did. I just don't see what's in it for them. It's possible it works but it's not considered "in spec" of course... but that would suggest that it has the potential for failing.
E6600's (Conroe) overclock by 50% too ...
Thanks for the insight, I was contemplating buying 4gb worth of memory from crucial myself, but for a while there decided to hold back a bit to see if there are indeed 800mhz modules coming soon. Screw it, I will just get the 667mhz ones
Thanks for the insight, I was contemplating buying 4gb worth of memory from crucial myself, but for a while there decided to hold back a bit to see if there are indeed 800mhz modules coming soon. Screw it, I will just get the 667mhz ones
Should 800MHz bus turn out to be supported, the difference is still likely to be negligable. The large L2 cache hides most of it anyway...
A particularly latency-sensitive app might see 10% at best... more likely is you'd see 2-3% -- if it supports 800MHz and if you want to pay the (currently significant) premium... and at the cost of decreased battery life...
Personally, I'd just go for the 667MHz DIMMs
Well, the RAM and the battery are user accessible. I think anything else, including the hard drive, isn't.
The only parts considered user-accessible by Apple (for warranty matters) is the RAM and the battery. Of course, if you do damage your computer hardware while replacing or working with one of these components they'll just file it under "accidental damage" and tell you the warranty doesn't cover it.
That said, if you're determined (and possibly foolhardy) enough any part is technically "user-accessible". You'll find many, many threads on this forum regarding the MBP's hard drive being user-replaceable or not.
Well, I have a screwdriver...
I agree though, it'd be much better if the MBP were like the MacBook...