Intel Core i7: Dual-Channel vs. Triple-Channel Memory Mode

[Parad0x]

Conclusion

After all these tests you need to ask yourself: What the heck is that triple channel mode useful for anyway, when it is obvious that is doesn’t provide performance boost that is expected? Maybe in the future and with some new versions of memory controllers, triple channel mode will provide better performances so transition to this platform will make more sense. Until that moment triple channel mode will be just another nice sticker on LGA 1366 motherboards and memory packages. Of course memory manufacturers already have triple channel DDR3 kits that are intended to be used on Intel Core i7 platform. Since these are triple channel kits they have three DDR3 modules with total capacity of 3GB or 6GB. Since prices for DDR3 memory are still pretty high it is obvious who will profit the most from this situation where (uninformed) users will go for triple channel kits with intention to use up potentials of their new platform.

http://www.insidehw.com/Reviews/Memo...mory-Mode.html

[geo]

who cares about i7 swindle. release the i5 already

[largon]

DDR3-1066 (CL 8-8-8-20)

I'm truly surprised they managed to find such positively ancient DDR3 modules.

[mikeyakame]

Whether it's faster or not I7 uses it and that is all that it comes down to. Such a ridiculous article, I don't know why websites even bother publishing this kind of thing.

If we can run more memory with minimal performance hit, that is good enough for me. Triple channel will always have slightly more latency because round-trip turnaround for 3 channels is always going to be longer than for 2.

What Intel have done is made the difference minimal at best and optimized the system for it. I commend them for that.

[mstp2009]

Old news. We covered this in a review months ago.

[BenchZowner]

That's a lame-ass review.
Insert seal of approval here

[Parad0x]

We have seen today that dual channel memory on the Core i7 is more than enough to keep it fed with bandwidth, especially since the CPU and memory are communicating with each other directly. There is very little lost bandwidth and it makes better use of all of the dual channel memory than the triple channel memory. However, with Intel’s plans to increase speeds and allow heavier memory usage, especially with IGP based systems, the extra bandwidth will end up being welcomed.

Our moral to this story is if you’re planning on a Core i7 system, don’t just go out and buy triple channel memory straight away. If your budget allows for it, then by all means more memory will help, especially in Vista. However, with today’s results you can see that dual channel memory is more than capable of keeping up with the Core i7.

http://www.tweaktown.com/articles/16..._it/index.html

[BenchZowner]

Nobody seems to have done some extensive testing ( 20+ benchmarks from various software categories ).
I'm not talking about having a different outcome, but it would be nice to see more tests.
And no! Don't look at me! I already have lots of things to do

[mikeyakame]

Nobody seems to have done some extensive testing ( 20+ benchmarks from various software categories ).
I'm not talking about having a different outcome, but it would be nice to see more tests.
And no! Don't look at me! I already have lots of things to do

Says the man who has time to suggest such a test

[BenchZowner]

Says the man who has time to suggest such a test

The man in question has done such things a lot of times already.
And is writing 30 page mobo reviews including 20+ benchmarks each

[mikeyakame]

The man in question has done such things a lot of times already.
And is writing 30 page mobo reviews including 20+ benchmarks each

*cracks the whip* reviews don't write themselves

I think it'd be hard to show the kind of results that would be of interest to the general i7 owning market.

I believe Intel added triple channel IMC for server market uses, I doubt there will be much to show in single socket processors as the gains will probably only be visible in 2+ socket configs, where memory bandwidth can be better utilized for heavy computational tasks.

[STaRGaZeR]

What the heck is that triple channel mode useful for anyway, when it is obvious that is doesn’t provide performance boost that is expected? Maybe in the future and with some new versions of memory controllers, triple channel mode will provide better performances so transition to this platform will make more sense. Until that moment triple channel mode will be just another nice sticker on LGA 1366 motherboards and memory packages.

I wonder what this guy was expecting

[RPGWiZaRD]

DDR3's gotten cheap nowadays, 6GB DDR3-1600 CL8 for 125 EUR and even highend performance kits like 6GB DDR3-2000 CL9 for just under 250 EUR. That's more important for me than tripple channel not providing any realworld benefit. I mean it wasn't that long ago 6GB DDR3-16000 kits were like 250 - 300 EUR. If we look back DDR3's price has dropped way faster than DDR2 did. It was the DDR3 prices that's put me off upgrading to an i7 setup knowing DDR3 prices would drop a lot in matter of a half year already, now it starts making more sense switching.

Hope i5 will make demand for DDR3 grow further so prices can continue drop even more.

[highoctane]

Triple channel is simply offers more bandwidth for what a vast majority of apps will use/need.

A couple other conclusions on the matter:

http://www.anandtech.com/cpuchipsets...spx?i=3448&p=5

http://xtreview.com/review232.htm

[Rammsteiner]

Ugh, if someone doesnt want to waste money on triple channel he/she could just get some 3x DDR1066 kit, it's not like you'll pay a whole lot more for that.

My kit was only 200 Euro's (actually 300, but 100 Euro discount).

No one is forcing you to go a) with a triple channel kit or b) to get a high-end triple channel kit like those DDR2K 7-8-7 kits. AFAIK DDR1066 vs DDR1600 didnt do much in games either, but wtf do I care, those black heatspreaders looked cool for my mob... O wait

[Tonucci]

It may become more usefull when 6 or 8 core CPU's are released.

[Loque]

lol.. I got more than twice the speed he got on everest with 12gbs @ 1700 cr1

[Hornet331]

I may become more usefull when 6 or 8 core CPU's are released.

No not even then, desktops apps arn't mem bandwidth limited, period.

And apps that are HPC, Database apps already gain from every MB/s more bandwidth.

[Glow9]

who cares about i7 swindle. release the i5 already

Why? Isnt it slower than i7, so it will be like their current c2d offering with ddr3 yay? Doesnt seem like much of an upgrade surprised intel would even release it with chip sales as they are.

[ColonelCain]

DDR3's gotten cheap nowadays, 6GB DDR3-1600 CL8 for 125 EUR and even highend performance kits like 6GB DDR3-2000 CL9 for just under 250 EUR. That's more important for me than tripple channel not providing any realworld benefit. I mean it wasn't that long ago 6GB DDR3-16000 kits were like 250 - 300 EUR. If we look back DDR3's price has dropped way faster than DDR2 did. It was the DDR3 prices that's put me off upgrading to an i7 setup knowing DDR3 prices would drop a lot in matter of a half year already, now it starts making more sense switching.

Hope i5 will make demand for DDR3 grow further so prices can continue drop even more.

Agreed.

Why go with dual-channel, when for very little more you can go triple?

Also, why spend the time picking through the masses of dual-channel DDR3 chips to find ones that use low volts?

[mstp2009]

6GB of OCZ is not $63 after rebate at newegg. I think prices like that pretty much makes this debate moot.

[highoctane]

Hell I spent over $200 for my 6gb but money wasn't the issue, it was simply a matter of using what I paid for with triple channel. I mean really how many folks are going to fork over the money for i7 and not bite the bullet for a 3x kit especially with as cheap as memory is getting now.

I think the debated issue is triple channel simply doesn't offer much if any tangible gains over dual channel. More or less it seems allot of folks seem to think performance will be degraded if you don't run triple channel and thats simply not the case i7.

You will basically get the same level of performance whether you run i7 dual or triple channel.

Even with i5 and dual channel config memory bandwidth will not be the biggest factor for performance differences to i7.

[JumpingJack]

It may become more usefull when 6 or 8 core CPU's are released.

Desktop applications are not memory BW hungry, any advantage of memory BW in benchmarks only show up on the mem BW tests for the synthetics (Everest, Sandra, Winrar's internal bench utility).

The vast majority of workloads on single user applications for desktop computing are invariant to memory BW past a certain point. Memory BW is like a water pipe, a certain diameter of pipe will flow x gallons/min at a given pressure. If demand never exceeds that flow rate then there will be no disruption in supply.

C2D still relies on on off-die memory controller shuttling data through a low BW FSB, yet has great performance -- because caching efficiency and workload demands make that poor FSB performance irrelevant in the computational equation.

[Parad0x]

Why? Isnt it slower than i7, so it will be like their current c2d offering with ddr3 yay? Doesnt seem like much of an upgrade surprised intel would even release it with chip sales as they are.

From my point of view i5 will be the "mainstream" simplified & cooler DC IMC version of the i7. The performance difference regarding the IMC imho will be marginal. My main worry is the Ibex Peak (P55) platform limitation on the PCI-E lanes (single card @ x16 but SLI or CF will have to do with only x8 per card). It'll be damn nice to have a simplified motherboard coz of the elimination of the NB & SB and going for the PCH solution. The cpu should also be a lil' cooler with only two mem channels to take care of.

[Slovnaft]

ah yes, buy the cheapest hardware available, run a suite of benchmarks unrelated to the disputed topic and call yourself a review site.
...