Q9450 vs Q6600 information please

J

jws2346

[H]ard|Gawd
Joined
Feb 28, 2007
Messages
1,768
Now I gots a question for those that are folding with a Q9450. I was kinda' tardy to the party and just got 2x Q6600 G0's in the last couple of months. I was kinda' surfin' around our fine forum and noticed in the Intel thread where a poster said the Q9450 at stock (2.66 GHz) would be equal to a Q6600 at 3.2 GHz OC' in the folding at home arena. I guess that extra cache really helps in folding,12 MB's vs 8 MB's :confused::

I thought I was a "fat cat" smokin' away with my 2x Q6600's foldin' 24/7 and I even had thoughts of getting a Q something 45nm, with 12 MB cache :rolleyes:.. The posters over there are discussing a new revision for the Q9450's (I guess something like B3 to G0 for the Kent) and they're also discussing the Nehalem chip, sayin' it might be 50% faster than the 45 nm Q9450's. If all that's truthin' Intel is really on a roll :eek:

How's all this sound to you of "infinite wisdom" in the computer folding business? :confused:

 
jws2346 said:
Now I gots a question for those that are folding with a Q9450. I was kinda' tardy to the party and just got 2x Q6600 G0's in the last couple of months. I was kinda' surfin' around our fine forum and noticed in the Intel thread where a poster said the Q9450 at stock (2.66 GHz) would be equal to a Q6600 at 3.2 GHz OC' in the folding at home arena. I guess that extra cache really helps in folding,12 MB's vs 8 MB's :confused::

I thought I was a "fat cat" smokin' away with my 2x Q6600's foldin' 24/7 and I even had thoughts of getting a Q something 45nm, with 12 MB cache :rolleyes:.. The posters over there are discussing a new revision for the Q9450's (I guess something like B3 to G0 for the Kent) and they're also discussing the Nehalem chip, sayin' it might be 50% faster than the 45 nm Q9450's. If all that's truthin' Intel is really on a roll :eek:

How's all this sound to you of "infinite wisdom" in the computer folding business? :confused:

Click to expand...

The first issue is indeed cache; it does make a really big difference in folding performance.

As for 45 vs. 65 NM performance in theory cycle per cycle everything except heat should be the only issue, however, in my own experiments playing with a Q9300 and a Q6600 showed I had to push the Q6600 to 2.6 to equal the performance of the Q9300 at its stock 2.5. Frankly I can’t find any logical reason for this however I did try two different Q6600s in the same setup and found one to be a tad faster then the other. Go figure.

As for Nehalem, consider what you have read about all new chips that come to market. Tis all BS until you get your hands on one and even then tons depends on it’s over clockability, maturity of the mother board and last but not least even it turns out to be 3 times faster (don’t hold your breath) it’s probable this chip will hit the market at well over a grand especially when you consider the current extreme chips are still well over a grand.

Bang for the buck ATM in my humble opinion is still the Q6600 G0;)


 
Hey thanks BillR for your clear and concise answer. :D

These questions prove I'm getting to be F@H anal. In your opinion would 2.5 GHz (9300) vs 2.6 GHz (Q6600) make any difference in the F@H times? The 9300, I believe, has 6 MB's of cache vs 8 MB's of cache for the Q6600. Would the difference in cache mean the Q6600 would fold any faster? :confused:

It sounds good to me to read where someone with "creds" thinks the Q6600 is still ATM the best bang for the buck cpu. I had another IMO credible computer individual post the same thing a while back, but you know how IT is, one day it's cool and the next day it's horse manure (a fine example of "Politically Correctness") :p

Oh yeah, I know what you mean by BS and having a Nehalem chip in your hand before "crowing" about how fast it is. The only problem is, with Intel being on a "roll" with it's Core 2's, it's hard not to believe the Nehalem is gonna' be one fast "motor scooter" (err... I mean cpu)

 
jws2346 said:
Hey thanks BillR for your clear and concise answer. :D

These questions prove I'm getting to be F@H anal. In your opinion would 2.5 GHz (9300) vs 2.6 GHz (Q6600) make any difference in the F@H times? The 9300, I believe, has 6 MB's of cache vs 8 MB's of cache for the Q6600. Would the difference in cache mean the Q6600 would fold any faster? :confused:

It sounds good to me to read where someone with "creds" thinks the Q6600 is still ATM the best bang for the buck cpu. I had another IMO credible computer individual post the same thing a while back, but you know how IT is, one day it's cool and the next day it's horse manure (a fine example of "Politically Correctness") :p

Oh yeah, I know what you mean by BS and having a Nehalem chip in your hand before "crowing" about how fast it is. The only problem is, with Intel being on a "roll" with it's Core 2's, it's hard not to believe the Nehalem is gonna' be one fast "motor scooter" (err... I mean cpu)

Click to expand...

The only reason I have the Q9300 is due to a one day deal, curiosity about the 45 nm technology and it is in a Shuttle for home theater use and runs nice and cool.

All that aside were I building another Folding Box in the near future the Q6600 would still be my chip of choice.

Aside from that running a Q6600 at 2.6 is just short of blasphemy and treason around these parts. 3.2 to 3.6 seems to be a comfortable operating range for that chip as long as you have good cooling.

Hey, if you fold for the [H] you have to be [H]ard so don’t waste all those free cycles;)


 
jws2346 said:
Now I gots a question for those that are folding with a Q9450. I was kinda' tardy to the party and just got 2x Q6600 G0's in the last couple of months. I was kinda' surfin' around our fine forum and noticed in the Intel thread where a poster said the Q9450 at stock (2.66 GHz) would be equal to a Q6600 at 3.2 GHz OC' in the folding at home arena. I guess that extra cache really helps in folding,12 MB's vs 8 MB's :confused::

How's all this sound to you of "infinite wisdom" in the computer folding business? :confused:

Click to expand...

I'm not so sure about this, but it may matter how the SMP clients are running. I ran a single SMP client on the following system:
[email protected] on an Asus Rampage motherboard, with 4Gig corsair XMS2 Cl4 ram, Vista 64

With only one SMP client running it was cranking through a 2665 at 12 min 27 sec per frame (2191ppd) .... this just seems to be on par with a Q6600 at about 3.2GHz as well. Next time I get a 2665 on my Q6600 I'll switch off one of the clients for a while and see what the production is to have an accurate comparison. Is the Q9450 worth the 65% price premium over a Q6600? ....I think the jury is still out with that one...

0.png
 
I can speak from experience, owning both the Q9450 and Q6600. While it's a bit hard to provide some apples-to-apples comparison due to the higher clock speed of the Q9450 and extreme RAM speed, I can say certaintly that at identical speeds, the new one is 5-10% more efficient, thus much faster at crunching than the Q6600.

However, from the overclocking and value point, nothing beat a Q6600 due to the 9x multiplier vs the lower multiplier of the new parts (7x for Q9300 and 8x for the Q9450). This is why the rest of my farm is composed of Q6600.

I bought the Q9450 just for the bleeding edge factor and since I was initially hoping that with watercooling, I would see a 4 GHz quad. However, it,s not temps but motherboard issues which is the stumbling block. I'm struck at 3.5 GHz :(

 
BillR said:
Aside from that running a Q6600 at 2.6 is just short of blasphemy and treason around these parts. 3.2 to 3.6 seems to be a comfortable operating range for that chip as long as you have good cooling.




Click to expand...

I've got 3 Q6600's running at 3.2 - 3.4Ghz

Decent cooling and good mobo and there is no reason you can't hit it. ;)

 
Xilikon said:
I can speak from experience, owning both the Q9450 and Q6600. While it's a bit hard to provide some apples-to-apples comparison due to the higher clock speed of the Q9450 and extreme RAM speed, I can say certaintly that at identical speeds, the new one is 5-10% more efficient, thus much faster at crunching than the Q6600.

However, from the overclocking and value point, nothing beat a Q6600 due to the 9x multiplier vs the lower multiplier of the new parts (7x for Q9300 and 8x for the Q9450). This is why the rest of my farm is composed of Q6600.

I bought the Q9450 just for the bleeding edge factor and since I was initially hoping that with watercooling, I would see a 4 GHz quad. However, it,s not temps but motherboard issues which is the stumbling block. I'm struck at 3.5 GHz :(

Click to expand...

Yeah I completely agree that the Q9450 isn't the best bang for buck for folding. It's quicker clock for clock but mine isn't much better than Xilikon's, I can run P95 for 4 hours at 3.6 but folding will have random reboots. I'm stuck at 3.56 to be folding stable.

But overall, I'm more than happy with it, my WC'ing is somewhat limited with a dual 120mm rad and an 8800GT in the loop, getting the 45nm chip ensured I'd have a little more headroom for temps.

I'm getting ~1950ppd according to fahmon on the 2665 with 2 vista 32bit clients running right now.


 
Wow :eek: Many Thanks to all you computer knowledgable members out there for being patient and clear responding to my question. I've got to admit people like me are getting a little spoiled on this team, we know all we've got to do is ask a question and it'll be answered almost immediately by some very credible computer people. ;)

OBTW: I was kinda' "bummed out" (not the leaping off of tall buildings kind) when I read the Q9450 was 50% faster than the Q6600 in folding. Sheech, I realize IT is ever changing, but I didn't think it got to the point where a the new quad C2D's (penryn, I read where it wasn't much more than a die shrink) were going to be 50% faster than the previous quad C2D's (Kentfields) in folding :rolleyes:

What really "twisted my shorts" was when I read the Nehalem is 50% faster than the Q9450, which is 50% faster than the Q6600, they shoulda' added the cpu after the nehalem is faster than God. (if Intel had anything to say about it there wouldn't be any question)

Edit: Hey please, don't anyone get the impression I'm complaining because I'm not, I'm just freaking amazed and I have the utmost trust in the opinions of my brothers and sisters on Team 33 when it comes to anything concerning computers (having trust is not my big suit) ;)

 
jws2346 said:
OBTW: I was kinda' "bummed out" (not the leaping off of tall buildings kind) when I read the Q9450 was 50% faster than the Q6600 in folding. Sheech, I realize IT is ever changing, but I didn't think it got to the point where a the new quad C2D's (penryn, I read where it wasn't much more than a die shrink) were going to be 50% faster than the previous quad C2D's (Kentfields) in folding :rolleyes:

What really "twisted my shorts" was when I read the Nehalem is 50% faster than the Q9450, which is 50% faster than the Q6600, they shoulda' added the cpu after the nehalem is faster than God. (if Intel had anything to say about it there wouldn't be any question)

Edit: Hey please, don't anyone get the impression I'm complaining because I'm not, I'm just freaking amazed and I have the utmost trust in the opinions of my brothers and sisters on Team 33 when it comes to anything concerning computers (having trust is not my big suit) ;)

Click to expand...

Wait a minute there. The Q9450 is by no means 50% faster than a Q6600 at the same clock speed. You might get 5-15% of a boost but that would be about it. If anyone has data to the contrary please post it. Again as soon as my Q6600 lands a 2665 WU I'll run another test so we have a hard and direct comparison.

0.png
 
HighYield said:
I'm not so sure about this, but it may matter how the SMP clients are running. I ran a single SMP client on the following system:
[email protected] on an Asus Rampage motherboard, with 4Gig corsair XMS2 Cl4 ram, Vista 64

With only one SMP client running it was cranking through a 2665 at 12 min 27 sec per frame (2191ppd) .... this just seems to be on par with a Q6600 at about 3.2GHz as well. Next time I get a 2665 on my Q6600 I'll switch off one of the clients for a while and see what the production is to have an accurate comparison. Is the Q9450 worth the 65% price premium over a Q6600? ....I think the jury is still out with that one...

0.png
Click to expand...

Ok here is the direct comparison (as direct as I can get it). Both SMP clients (single instance only) were running at 3.2GHz: one on a Q6600 (356x9) and one on a Q9450 (400x8). Both systems had 64 bit vista with 4 gig of ram (the Q9450 was a fresh install and the Q6600 was an old install and had plenty of bloat (sorry that is my main rig and apparently I don't travel light). Both systems had a 1:1 FSB-DRAM setting so the Q9450 will be running with faster ram. The memory timings on the Q6600 were 5-5-5-15 and the Q9450 were 4-4-4-12 .

[email protected] with WU 2665 13 min 26 sec per frame 2058 ppd (fahmon)
[email protected] with WU 2665 12 min 27 sec per frame 2191 ppd (fahmon)

The Q9450 is folding about 7% faster than a Q6600 at the same clock speed. I don't know how to subtract out the difference in faster ram on the Q9450 or the fact that the Q9450 was on an Asus Rampage mother board and the Q6600 is on an older P965 Asus Commando which is starting to show its age a bit.

There might be more of a difference when folding two instances of SMP or with different work units. I'll leave that up to someone else to test.

0.png
 
EvilAlchemist said:
The one thing I do like aout the New 45mm CPU's ... is the lower power draw ...must lower....

Click to expand...

Allow me to throw a cautionary note in here. The 65 nm parts were rated at a max VCORE of 1.35 and right from day one people were running 1.5+ as their new VCORE for over clocking.

That was right about the time Nvidia released the 680i board and right on Nvidia’s site they had a nifty over clocking guide that said point blank to set the VCORE to 1.5 and use good cooling.

Frankly after all but two years into the 65 nm part there have been amazingly few failures and most of the few I’ve read about were the result of, well, stupid people.

Now the 45 nm part is out and is tagged at a max of 1.25 and I’ve already read about more failures in this parts short life then I ever did in the 65 nm part. The small die automatically lowers the point at witch the smoke can be let out of the part and we all know you can’t ever put the smoke back inside.

I might add the new part also requires 14 AMPS at the 8 pin CPU connector and can easily be pushed past 60c with little or no effort.

With future die shrinks and more cores on board the CPU we just might see much more carefully binned parts and that just might cost us the over clocking edge we have all come to love.

Just some random observations and thoughts ;)


 
BillR said:
Allow me to throw a cautionary note in here. The 65 nm parts were rated at a max VCORE of 1.35 and right from day one people were running 1.5+ as their new VCORE for over clocking.

That was right about the time Nvidia released the 680i board and right on Nvidia’s site they had a nifty over clocking guide that said point blank to set the VCORE to 1.5 and use good cooling.

Frankly after all but two years into the 65 nm part there have been amazingly few failures and most of the few I’ve read about were the result of, well, stupid people.

Now the 45 nm part is out and is tagged at a max of 1.25 and I’ve already read about more failures in this parts short life then I ever did in the 65 nm part. The small die automatically lowers the point at witch the smoke can be let out of the part and we all know you can’t ever put the smoke back inside.

I might add the new part also requires 14 AMPS at the 8 pin CPU connector and can easily be pushed past 60c with little or no effort.

With future die shrinks and more cores on board the CPU we just might see much more carefully binned parts and that just might cost us the over clocking edge we have all come to love.

Just some random observations and thoughts ;)


Click to expand...

Excellent observations and thoughts. The 65nm parts definitely appear to be more robust (way more so from a Vcore standpoint) and I too have seen many reports of people killing 45nm parts with too high/prolonged vcore that was mild in comparison on a similar 65nm processor. Very hugh thread over at XS about this issue on the 45nm CPU's and from everything I've read, the "long term" max on the 45nm seems to be a max of 1.4 Vcore.

Caution and over zealous use of Vcore should be avoided IMHO ;) but to clear up actual numbers per the Intel specs, the absolute max Vcore for the 65nm is 1.55 per the processor datasheet (see page 20) and for the 45nm parts max Vcore is stated as 1.45 per The 45nm proc datasheet (pg. 17).

So that seems to follow that people are killing the newer processors with a constant Vcore over 1.4.

Don't mean to correct your VERY important and overall correct observations, but wanted to see the specs for myself and pass them along....

Although if enough people read that and went into a panic underclocking I might catch some of them faster. Hmmmmm :D

Forget what I said... TURN THOSE PUPPIES DOWN QUICK!! :p:D

0.png
 
You must log in or register to reply here.
Back
Top