Help with configuring Core 2 Quad E9300
12 Dec 2008 12:03:16 UTC
Topic 194080
(moderation:
Hello,
I recently bought CPU in the title. Attached all 4 cores to Einstein@Home and hopes to get good RAC, but I have some doubts about its' current performance.
On my previous box I had to do special actions in BIOS and OS to switch off advanced technologies like SpeedStep, which back then decreased the frequency when processed low-priority applications even at 100%. And I suspect somethign like this is happenning here too. At least, I don't have another explanation why more antique Quad E6600 are way out more performant, like this one for instance
http://einsteinathome.org/host/915960
My host is
http://einsteinathome.org/host/1737308
Also I checked the temprature of CPU today. It's relatively low at 30 C, while tempratures of each core is about 70 C. Is it normal?
Regards,
Andrew
Help with configuring Core 2 Quad E9300
)
It's not your pc most likely regarding the difference in speed you're seeing. The other pc you mentioned uses the Optimized apps and you don't. Simple to fix by learning to install and using them on your pc's that are compatible. Use cpu-z to determine which ones are.
The core temp you mentioned does sound high however, so I would do what it takes to get that under control on which ever pc's you're seeing this on, before using any optimized app on it. Cleaning it out from dust if you haven't already, and checking on a better cooling system might be in order.
RE: The other pc you
)
I used Power Apps before. Didn't want to switch to this one, since was expecting it would be released soon as official version. Also I'm still getting to use to my new box.
But I'm not sure the difference is just Power App vs. Official App. Please, check MIPS and FPS of CPUs in question. There is a huge difference too.
I'm also concerned about core temp. I found out it's high only today. Will check what I can do. I bought CPU in box, which comes with cooler, and the whole system is brand new, so I wonder what might cause such temp.
Thanks for replying anyway. Will check Power Apps next week.
Regards,
Andrew
Your fastest runs are below
)
Your fastest runs are below 32,0000, the performance looks about correct. The Q6600 you linked to is overclocked to 3.6GHz and uses the newer app.
The temp isnt bad for a stock cooler, most user reports of these Q9000 CPUs running at ~60c with stock cooler were from using a too low TJmax value in coretemp, at the time noone knew for sure what was the correct value but Intel later said to use 100 and then reported temps went up to more realistic levels.
Memory settings affects E@H performance, check the specs of your RAM and set the manufacturers recommended voltage and timings in bios, it could give you ~5% extra performance. Combined with the optimized app you should see a very nice improvement.
Team Philippines
A core temperature of about
)
A core temperature of about 70 degree Celsius is normal while running all four cores at 100% cpu-power and 24/7. I use a very big cpu-fan (Noctua NH-U12P) which is able to cool down the very hard working cpu cores in a silent way. The Intel-boxed-fan was too noisy and the core temperatures were the same.
Core temp 0.98.1 is used by me.
The previous used Intel Core2Duo E 6420 ( 2130 Mhz, 2 MB Cache) didn't get hotter than 40 degree Celsius.
Intel Q9300 Quadcore, 2500 Mhz, 4096 MB RAM, GeForce 9800 GT, Vista Ultimate 64-bit, Ubuntu 10.10 64 bit
RE: Memory settings affects
)
As a builder of most of my own pc's I am throwing my 2 cents in here...memory speed can make a difference too. For instance if your pc comes standard with pc-6400(ddr2 800) memory but your motherboard will handle pc-8400 memory(ddr2 1066), put in the higher speed memory. It will NOT make a HUGE difference, but if you are looking for that last little bit, and it sounds like you are, do it it WILL help.
RE: As a builder of most of
)
Actually, this isn't entirely correct. Simply dumping in faster RAM won't boost performance unless the RAM dividers are matched to the FSB. You'll often get better overall performance using a 1:1 divider on the RAM. I overclocked my Q9550 to 3.4GHz - not for the CPU speed boost (although it's a nice bonus), but so that I could run the RAM at 1:1 with the FSB (both of which are set to 400MHz). That being said, I'm using an aftermarket heatsink which does a better job than the stock cooler.
RE: Actually, this isn't
)
Faster RAM helps more than tight timings for core2, but the difference is small.
Remember 1:1 is for the real frequency, since FSB is quad data rate and RAM is dual data rate the optimal divider is 1:2, only then will FSB and RAM both operate at its maximum theoretical bandwidth. Thats why you see some performance gain going from DDR2 to DDR3 on lga775 systems, not enough to warrant the higher price of DDR3 though, the FSB/MCH is very outdated and doesnt scale well with faster RAM. Core i7 has higher bandwidth in single channel than Core2 in dual channel with same speed RAM, that says more about how antiquated the FSB/MCH is than about how modern i7 is (after all, AMD been using integrated memory controller since 2003, Intel is just late to the party)
Team Philippines
Yes that's true, Intel came
)
Yes that's true, Intel came late with memory-controllers and also 64bit but that says nothing about the quality.
RE: Remember 1:1 is for the
)
No... if you want to talk "real frequencies", then a 1600FSB turns out to a "real frequency" of 400MHz. DDR2-800 operates at 800MHz dual-rate, which translates to a "real frequency" of 400MHz. Hence the 1:1 ratio. 1:2 would happen if you used an older CPU that ran at 800FSB
The Q9000 processors operate on a 1333FSB, or 333MHz (real) at stock speeds. Running DDR2-800 means using a 5:4 divider (400:333). DDR2-1066 would mean some really awkward multipler, and would probably give you worse overall performance than running at DDR2-800.
RE: No... if you want to
)
Yes, 400 FSB and 400 RAM is 1:1, thats the real frequency, which was what i said. You can run 1333 FSB Cpus with divider 1:2 if you got DDR3-1333, and with FSB 400 you then need DDR3-1600 RAM. Its a myth that 1:1 is best, but its true that it is often most practical to use 1:1 when overclocking, and the performance loss is small in benchmarks and virtually non-existing in most real world usage scenarios.
This must be where you took a wrong turn. In this scenario you run a 5:6 divider, 333:400 (FSB:RAM, not RAM:FSB). 5:4 would be if you have a FSB 1333 CPU and 533 RAM (PC2-4300)
Team Philippines