Eh, those results are a bit hard to believe and I won't till TAT or CoreTemp is used. SpeedFan had/has a well known issue with reading Core 2 based processor temperatures either way to low or just simply wrong, and on a consistent basis as well. I want to see more "legitimate" program and then I'll believe the temperatures are that low.
Core 2 Quad Q6600 - Is it Gamer Choice
- Thread starter Shtal
- Start date
Eh, those results are a bit hard to believe and I won't till TAT or CoreTemp is used. SpeedFan had/has a well known issue with reading Core 2 based processor temperatures either way to low or just simply wrong, and on a consistent basis as well. I want to see more "legitimate" program and then I'll believe the temperatures are that low.
Unknown Soldier
Veteran
Hmm.. wierd that he doesn't run it at 400Mhz x9 instead of 450 x 8 .. both will hit 3600 but the ration for x9 will be 1:1
Maybe he was testing potential of the FSB. I'd prefer to run it at x9 though.
US
I was under the impression that the multiplers were locked on normal consumer parts.
upward only. speedstep uses the unlocked downward multi to throttle the CPU and lighten power draw when the computers idle. You can force a lower multi operation through bios though.
Unknown Soldier
Veteran
Nope .. maybe you got confused with fully unlockable .. like the QX6x00's they can be unlocked at the multiplier at any multiple I believe. The Q6600 though can only use 6, 7, 8 and 9(default).
QX6x00 can do 10 and 11
US
QX6x00 can do 10 and 11
US
More memory bandwidth, 1800 FSB (450x4) vs 1600 FSB (400x4). Remember that it's easier for a CPU to reach higher FSB at a lower multiplier (for instance, just because a CPU can do 450 FSB with an 8x multiplier, it doesn't mean it can do 450 FSB with a 9x multiplier). I personally would rather have 450x8 than 400x9.
Besides FSB:Memory ratios haven't really matter for some time:
HardForum's memory guide said:
I'll assume this still applies to C2D chips too (please correct if I'm wrong).
Unknown Soldier
Veteran
Sorry, I think 400x9 would be better than 450x8. Better timings etc. Only thing why going x8 would be better is because you can most probably clock it higher. I've noticed that using x9 the memory clocks cannot get as high as with x8.
As for why I say x9 is better. I did a test and found that at 356x9 I get a better 3DMark06 score than at 400x8 both which translate to 3.2Ghz. My memory timings for both was 356Mhz 4.4.4.11 while 400Mhz had 5.5.5.15.
US
As for why I say x9 is better. I did a test and found that at 356x9 I get a better 3DMark06 score than at 400x8 both which translate to 3.2Ghz. My memory timings for both was 356Mhz 4.4.4.11 while 400Mhz had 5.5.5.15.
US
Buy better ram.
Throwing ram timings into the equation isn't what we were talking about. We were talking about why someone would prefer 450x8 over 400x9. Obviously if your ram is on the lower quality side, you won't be able to obtain higher FSB speeds (at least not at the desired timings). That has no reflection on the 450x8 vs 400x9 debate. Perhaps 356x9 may translate to better for performance for you, but for those who have high quality ram, they will see better performance at 400x8. It's not a big difference, but it's hard to argue that having a higher a multiplier with lower FSB would be more beneficial than a lower multiplier with a higher FSB (assuming one's ram is capable at higher FSB speeds).
What about something like my RAM, Corsair 8500C5D which is rated 4.4.4.12 @ 800Mhz or 5.5.5.15 @ 1066Mhz. I haven't really done any benchmarks comparing the 2, but I'm currently running around 850ish at 4.4.4.12 and not really sure which would be the better way to go. Pushing as much Mhz out of it @ 5.5.5.15 or lower Mhz @ 4.4.4.12?
Sorry for derailing the thread again, but to continue where Shtal and I left off...
Nehalem IMC for Server/Workstation & Extreme Edition only. Granted it's L'inq, but it's better than Fuad...
Nehalem IMC for Server/Workstation & Extreme Edition only. Granted it's L'inq, but it's better than Fuad...
Considering the Fact that AMD only has 2MB L2 cache for Athlon-X2/FX CPU’s compare to Intel Core 2 4MB L2. AMD CPU’s don’t rely as much as Intel CPU’s to have larger L2 cache (To boost performance) because of AMD IMC and Intel does not. Nehalem will be interesting CPU’s “If combine larger L2 cache + IMC” what kind impact of performance it will have?
IMCs + serial links greatly reduce latency for CPU-RAM communication when you don't already have an effective latency-reduction implementation in-place, i.e. Intel's large caches + aggressive pre-fetch algorithms on shared-FSB + discrete shared-MC approach. Shared-FSBs are the main bottlenecks for throughput (and to a lesser extent latency), as they are more parallel than the serial link approach used by AMD (so more complex) and thus can only be scaled to a certain clockspeed, generally quite a bit lower. Additionally, because of the nature of shared-FSBs there is a much heavier load placed upon them, which also contributes to a lower overall clockspeed (and ultimately transfer rate) compared to serial links. Shared (off-die) MCs are the main bottleneck for latency because of the additional distance the information has to travel, and the extra "hop" going from one chip to another and then to memory (and back).
These bottlenecks become especially apparent in multi-socket systems. However, even the IMC approach is not the perfect latency-hiding solution, since >4S systems have an additional hop in the worst-case scenario (i.e. when a CPU not directly linked to another needs to access the other's memory).
Speaking of latency!
If you ever looked at AMD IMC between DDR1-S939 vs. DDR2-AM2 socket.
Using IMC for DDR1 you get a lot lower latency access time vs. going to DDR2, but using DDR2 results higher memory bandwidth available for Athlon X2/FX CPU's which resulted that AMD CPU's are not limited by bandwidth.
But going to DDR3 using IMC, not sure of any advantage.
If you ever looked at AMD IMC between DDR1-S939 vs. DDR2-AM2 socket.
Using IMC for DDR1 you get a lot lower latency access time vs. going to DDR2, but using DDR2 results higher memory bandwidth available for Athlon X2/FX CPU's which resulted that AMD CPU's are not limited by bandwidth.
But going to DDR3 using IMC, not sure of any advantage.
Must resist urge to buy...Penryn relatively near...must fight against the craving
I was very close myself to buy (Intel core2 Q6600 quad) but I have to wait for 45 nanometer Penryn do to it will have SSE4 support which I really need....
I was very close myself to buy (Intel core2 Q6600 quad) but I have to wait for 45 nanometer Penryn do to it will have SSE4 support which I really need....
You're missing out man; I just got mine today!
But the rest of the computer is coming Wednesday...
...so I can only stare at my Q6600 for the next few days...
I was very close myself to buy (Intel core2 Q6600 quad) but I have to wait for 45 nanometer Penryn do to it will have SSE4 support which I really need....
Buy now, then upgrade and sell later.
I don't see the value in depriving yourself from enjoyment now.