One other point on the Q6600 vs. Q9550 matter. I wonder whether people's perception of the relative performance is adjusted to reality.
Before I went to undervolted stock frequency, my Q6600 ran for months at 2.79 GHz. This makes it a near frequency match to the new Q9550 I run at stock 2.83 GHZ.
This graph shows good coverage across the infamous variation cycle for the Q6600, and although the taller parts of the peaks are sadly still not covered after more than a month of operation, shows enough of the Q9650 to give the clear impression that the important valley which contains most work is lower by more than the 1.4% frequency difference, and to at least hint that the peaks may be improved by a considerably higher proportion.
To be sure this is not a pure comparison. Aside from the (slight) frequency mismatch, the motherboards, chipsets, and RAM settings differ somewhat. But all are mainstream and I think in the real world of probable user experience it is a rather fair comparison of the clock-for-clock expected performance.
Of course the Q6600 at 2.79 in this comparison was appreciably overclocked and somewhat overvolted, whereas the Q9650 I am running at 2.83 is appreciably undervolted, so running at far lower CPU power to produce somewhat higher science output.
In reply to the question about the 45nm CPUs, my Q9550 is overclocked to 3.4GHz. The RAC has been climbing steadily, and it's just crossing over the 3.6K mark. We'll see where it tops out in the end. :)
Your surmise that more people bought the Q6,600 Core 2 quad for BOINC crunching, and overclocking, than do for the Penryn CPU series is probably correct.
Now the Q9,550 is at a comparable price point to the Q6,600 last year, then this may change.
Wait 12 months and the Core i7 will probably repeat this price point.
I am looking forwards to the 32nM fabrications of Sandy Bridge. Reports are coming out the Core i7 can just top 80 Gigaflops, and the same for Sandy Bridge is 200.
The latter will make some BOINC cruncher, and allow 4 projects concurrently each with an RAC around 6,000 CS.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
Now the Q9,550 is at a comparable price point to the Q6,600 last year, then this may change.
Wait 12 months and the Core i7 will probably repeat this price point.
As for the CPU itself, the 920 model of i7 which my favorite store currently prices at US $295 is essentially there already. Knowing only the CPU price I had originally intended my recent Q9550 build to be a 920 i7. However the dates were tight for running into the holiday season, and I was eventually discouraged by expensive motherboards and memory and by concern that buying motherboards on first release was likely to expose me to some teething problems.
The memory situation is less bad already than I thought, so the missing element for compelling economy i7 BOINC hosts seems to be the motherboard at the moment. I have good hope that will improve in well under twelve months, but no inside information.
As you can see cyclic behavior remains very much a part of the picture. The "Variance = .208" value is a representation of the degree of variation--working from memory I think that is well in the range of historic observations.
Quite typical.
Quote:
I chose not to include all the valley points you had collected, for fear that Mike's computation would be distracted into trying to fit the wiggles and noise among them, rather than fitting the major peak to valley feature. To my eye his fit appears quite good in this case. Your ability to provide points surrounding a peak and going down a decent distance form the peak really helps.
The basic fit seems good.
Quote:
Mike, if you see this please comment if there is something beyond 7g we should be using.
No, I haven't made much progress in the wiggle area. I've been trying to get an analytical function fit with more parameters straight off ..... but I might lunge for a two stage approach : fit the basic curve as we have been, then take the residuals from that and try to get a best fit two component sinusoidal 'beat' alignment.
Quote:
Also, the copy I downloaded today still has an internal headline and some other tagging describing it as for "S5R3" while I think the default constants and such are actually S5R4-specific.
Thanks I'll edit that.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
the missing element for compelling economy i7 BOINC hosts seems to be the motherboard at the moment. I have good hope that will improve in well under twelve months, but no inside information.
Do you think Intel will make a value chipset for LGA1366? Maybe they will position 1366 as a luxury line and make value chipsets only for socket LGA1160. LGA775 could be the last socket where low end and higher end CPUs can use the same mainboards. I dont think Intel was thrilled that most core2quads were placed onto cheap P35/P45 boards instead of X38/X48, and they dont need LGA1366 to compete with AMD, LGA1160 can do that without much trouble. By splitting up into two lineups they dont have to compete with themselfes on the higher end anymore.
Do you think Intel will make a value chipset for LGA1366? Maybe they will position 1366 as a luxury line and make value chipsets only for socket LGA1160. LGA775 could be the last socket where low end and higher end CPUs can use the same mainboards. I dont think Intel was thrilled that most core2quads were placed onto cheap P35/P45 boards instead of X38/X48, and they dont need LGA1366 to compete with AMD, LGA1160 can do that without much trouble. By splitting up into two lineups they dont have to compete with themselfes on the higher end anymore.
I think I agree with you on all points.
With the memory controller vanished into the CPU, I don't know how much room there is for Intel to charge a super-premium price for the chipset itself, but agree that the positioning is likely not to put LGA1366-socket systems really cheap. Quite possibly the LGA1160-socket systems will be the value point for BOINC work, as for much else.
One other point on the Q6600
)
One other point on the Q6600 vs. Q9550 matter. I wonder whether people's perception of the relative performance is adjusted to reality.
Before I went to undervolted stock frequency, my Q6600 ran for months at 2.79 GHz. This makes it a near frequency match to the new Q9550 I run at stock 2.83 GHZ.
This graph shows good coverage across the infamous variation cycle for the Q6600, and although the taller parts of the peaks are sadly still not covered after more than a month of operation, shows enough of the Q9650 to give the clear impression that the important valley which contains most work is lower by more than the 1.4% frequency difference, and to at least hint that the peaks may be improved by a considerably higher proportion.
To be sure this is not a pure comparison. Aside from the (slight) frequency mismatch, the motherboards, chipsets, and RAM settings differ somewhat. But all are mainstream and I think in the real world of probable user experience it is a rather fair comparison of the clock-for-clock expected performance.
Of course the Q6600 at 2.79 in this comparison was appreciably overclocked and somewhat overvolted, whereas the Q9650 I am running at 2.83 is appreciably undervolted, so running at far lower CPU power to produce somewhat higher science output.
In reply to the question
)
In reply to the question about the 45nm CPUs, my Q9550 is overclocked to 3.4GHz. The RAC has been climbing steadily, and it's just crossing over the 3.6K mark. We'll see where it tops out in the end. :)
archae86 Your surmise that
)
archae86
Your surmise that more people bought the Q6,600 Core 2 quad for BOINC crunching, and overclocking, than do for the Penryn CPU series is probably correct.
Now the Q9,550 is at a comparable price point to the Q6,600 last year, then this may change.
Wait 12 months and the Core i7 will probably repeat this price point.
I am looking forwards to the 32nM fabrications of Sandy Bridge. Reports are coming out the Core i7 can just top 80 Gigaflops, and the same for Sandy Bridge is 200.
The latter will make some BOINC cruncher, and allow 4 projects concurrently each with an RAC around 6,000 CS.
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
RE: Now the Q9,550 is at a
)
As for the CPU itself, the 920 model of i7 which my favorite store currently prices at US $295 is essentially there already. Knowing only the CPU price I had originally intended my recent Q9550 build to be a 920 i7. However the dates were tight for running into the holiday season, and I was eventually discouraged by expensive motherboards and memory and by concern that buying motherboards on first release was likely to expose me to some teething problems.
The memory situation is less bad already than I thought, so the missing element for compelling economy i7 BOINC hosts seems to be the motherboard at the moment. I have good hope that will improve in well under twelve months, but no inside information.
RE: As you can see cyclic
)
Quite typical.
The basic fit seems good.
No, I haven't made much progress in the wiggle area. I've been trying to get an analytical function fit with more parameters straight off ..... but I might lunge for a two stage approach : fit the basic curve as we have been, then take the residuals from that and try to get a best fit two component sinusoidal 'beat' alignment.
Thanks I'll edit that.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: the missing element
)
Do you think Intel will make a value chipset for LGA1366? Maybe they will position 1366 as a luxury line and make value chipsets only for socket LGA1160. LGA775 could be the last socket where low end and higher end CPUs can use the same mainboards. I dont think Intel was thrilled that most core2quads were placed onto cheap P35/P45 boards instead of X38/X48, and they dont need LGA1366 to compete with AMD, LGA1160 can do that without much trouble. By splitting up into two lineups they dont have to compete with themselfes on the higher end anymore.
Team Philippines
RE: Do you think Intel will
)
I think I agree with you on all points.
With the memory controller vanished into the CPU, I don't know how much room there is for Intel to charge a super-premium price for the chipset itself, but agree that the positioning is likely not to put LGA1366-socket systems really cheap. Quite possibly the LGA1160-socket systems will be the value point for BOINC work, as for much else.