Q6600 1066FSB is slower ram bottleneck?
I just had a question about the Q6600. If it is rated to use 1066 Front Side Bus wouldn't it be logical to only get PC2-8500 RAM and a motherboard that supports PC2-8500? I have read articles on memory timings and RAM with higher frequencies but when they overclock and tune the timings it really doesn't amount to much. So if you get say DDR2-800 which is really cheap right now is it really going to make that much difference? If so...what is affected? Bandwidth? The ability to easily overclock with faster RAM?
FSB is quad pumped so to match it with memory you have to divide it by 4 which equals to 266(.5). If you want to synchronize your ram (1:1) you'll be ok with just 533MHz.
Personally, I have 800Mhz RAM at 533 with 3-3-3-10 timings stock volt and value RAM. (the cheap stuff, no Corsair/OCZ)
If you don't mind me posting a question here, I was wondering if it's better to have 800Mhz not synchronized & 5-5-5-15 timings or have 533Mhz synchronized & 3-3-3-10 timings. Thanks.
a higher memory frequency gives greater bandwidth, and will be useful in a core 2 duo/quad system, seen as low latencies hardly give any performance boost in this platform (way less than a higher frequency).
fast ram (e.g. pc2-1066) means you can set the ram speed, low at first, then overclock the fsb (which will increase the ram frequency) e.g
you set the ram to 533 (i.e. 1:1 with fsb) and overclock the fsb to 400, the ram will now be running at 800mhz etc. the key being the ram doesn't exceed it's rated speed and limit your overclock.
personally in a intel system at the moment high frequency ram is nice, but i'm not sure the benefits above 800mhz ram are that great (couple of fps in games at most). hope this helps a bit
The Q6600 & Q6700 has an effective FSB of 1066 but as Chasos stated it is quad pumped (266 * 4 = 1064). I am pretty sure quad pumped has nothing to do with the fact that we are talking about a quad processor but more with the core arctitecture. Memory is only the FSB * 2 (266 * 2 = 533) so 533MHz memory is sufficient to run the memory to fsb ratio at (1:1). Most motherboards let you change the memory to fsb ratio manually or they will adjust it automatically based on what memory is detected. PC2 4200 = 533, PC2 5300 = 667, PC2 6400 = 800 and PC2 8500 = 1066. PC2(DDR2) 6400 is pretty much the standard nowdays and the most cost effective memory. If you run DDR2 6400 (800) the motherboard will adjust the ratio to something like (4:5) to run at 800MHz. Use CPUid to see what speed your memory is running at.
Memory speed (533 vs 800), as well as memory timings, don't effect performance that much. Run some benchmarks and you will see. To answer Chaos-Energy's question, 3-3-3-10 timings are better than 5-5-5-15 but I think the higher MHz of 800 (vs 533) might yield better performance even with higher timings. Also check the memory specs to see if you actually have 5-5-5-15 memory. My motherboard wants to run my 4-4-4-12 at 5-5-5-15 when set to Auto.
1066 RAM will run at 533 FSB. When 800 RAM will run at 400. 400 times a multiplier of 9 gives 3.6Ghz... which is hard to hit on air cooling with a Q6600. This is why I got the 800 RAM, 400 FSB x 8 multiplier will give 3.2Ghz. Hopefully I can get that to run stable.
I really don't see the point to 1066 memory unless you are watercooled and looking to exceed 3.6Ghz.
I just bought a Q6600 from Dell, and have some spare pc-4200 ram strips. I'm guessing from checking out wikipedia and the piece at http://www.madshrimps.be/?action=getarticle&number=4&artpage=1965&articID=472 that I'd be OK using this memory in that board, even though it normally "calls for" pc-5300 or better.
Am I on track here? Any advice would be appreciated. TIA.
Ok, I need to say one thing.
The MINIMUM for an Intel Processor is 1:1. The argument has always been that the system is most EFFICIENT at 1:1 and not that it is PERFORMING better.
While it might not show a REAL world improvement in all situations I can guarentee that running faster RAM *DOES* improve performance and memory read/writes. Why would we be using DDR3 if this were not the case?
I don't have access to them now, but when I get home I will post my performances numbers for:
1066 FSB - 800 MHz RAM. (CAS 4)
1600 FSB - 800 MHz RAM (1:1). (CAS 4)
1600 FSB - 1200 MHz RAM. (CAS 5)
Edit: Again, as far as timings go, it's all relative. Timings are cycles before the memory does X. Now I won't use actually MHz (as the numbers are way to small for me to feel like dealing with), you can use it as Hertz.
Relative times (Using Hertz, not MHz).
667 CAS 3 - 0.0045
800 CAS 4 - 0.005
1066 CAS 5 - 0.0047
1200 CAS 6 - 0.005
1600 CAS 7 - 0.0044
These just consider the latencies, the other thing to consider is the relative bandwidth. Faster RAM will transfer the data faster.
PlasticSashimi said:I run my RAM unlinked....
also I noticed on my Quad rig...
5-5-5-15 @ 960mhz
benched lower in games and synthetic
than 4-4-4-12 @ 900mhz
It was my mistake when I said Bandwidth was always better, 801 MHz CAS 5 is going to perform worse than 800 MHz CAS 4. It has to be enough of a hike in MHz to offset the fact that it will take one extra cycle.
Why aren't we all using DDR400 CAS 2.5 on 200 MHz Bus E4600s if that were the case?