With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.
Not to be a jerk, but I disagree. Lets say a Haswell era CPU has twice the cores, but accomplishes this in the same time. I am sure the power consumption would be much higher.That's not really the point of the discussion, but yes, if both take the same time then the average power measurement relationship will be mirrored to the total energy consumed.
7-Zip is clearly not the tool to use to measure IPC.
Not to be a jerk, but I disagree. Lets say a Haswell era CPU has twice the cores, but accomplishes this in the same time. I am sure the power consumption would be much higher.
Are we to assume Alder Lake is better in any scenario that you don't bring up?7-Zip decompression hm, why is it slower vs i5 11400F/12mb L3 Cache.
How can be slower, if i5 12400F has higher IPC+bigger L3 Cache 18mb.
How do you explain Rocket Lake beating Alder Lake?AT said the contrary, at the time they used 7 ZIP as representative of Integer perf, FTR Zen 3 improved more INT than FP relatively to Zen 2.
How do you explain Rocket Lake beating Alder Lake?
How do you explain Rocket Lake beating Alder Lake?
I am also in the UK now and these listings took literally two seconds to find, and it doesn't include some other motherboard OEMs who haven't announced B660/H670 series boards (CES being still ongoing).
The IPC from Rocket Lake to Alder Lake is just too much to believe PL1/PL2 have anything to do with it.It's worth reminding people that the PL1 and PL2 settings are more guidelines than actual rules.
Just for curiosity I locked my 12700K in 6+0 @ 4Ghz and ran the default 7z benchmark. Package power was around 60W. This means that depending on the average silicon quality of the 12400 we might see clock regression in 7z bench and also means the 11400 is very likely power limited in this benchmark.7z is very likely not hitting any power limits, not in my testing, it is sipping power with truckload of LLC misses.
Let's say you have CPU that uses 10W per second in that benchmark and finishes in 10 seconds, that would be 100W to finish that bench, if they ran it 3 or more times and got the average of those 3 or more runs to make sure that it's accurate then that's that.It doesn't change the core argument that PBO was used and the numbers are flawed for comparison for that reason.
It doesn't matter what unit of measure the chart for Igor's is in because you can't get that average without maintaining that level consistently or exceeding it at various points. It doesn't matter how you want to arrive at it either because it doesn't change even if you just take the total watts and the amount of time to calculate it at the end or how you do it.
Did you not read ? You can't talk about watts that way, its an instantaneous current flow measurement. What you are describing is the way joules are calculated. If its 10 watts, it could be that way forever, its still a 10 watt draw...Let's say you have CPU that uses 10W per second in that benchmark and finishes in 10 seconds, that would be 100W to finish that bench, if they ran it 3 or more times and got the average of those 3 or more runs to make sure that it's accurate then that's that.
You can run way lower power but the average amount of power to finish the job will still be high.
Not the average amount of power used while running the bench but the sum of power to complete it.
Watt = joule/s. Watt per second is something else, not instantaneous power draw. 10 seconds at 10 W is 100 joule of total energy spent in running the benchmark.Let's say you have CPU that uses 10W per second in that benchmark and finishes in 10 seconds, that would be 100W to finish that bench, if they ran it 3 or more times and got the average of those 3 or more runs to make sure that it's accurate then that's that.
You can run way lower power but the average amount of power to finish the job will still be high.
Not the average amount of power used while running the bench but the sum of power to complete it.
I believe I said that right above your post !Watt = joule/s. Watt per second is something else, not instantaneous power draw. 10 seconds at 10 W is 100 joule of total energy spent in running the benchmark.
Electricity is paid for in units of total energy consumed.
One last point: I also ran the 7z bench with my default 8+0 configuration. Clocks went up around ~4.6Ghz and package power reached my limit of 150W. I don't know what your current configuration is, and whether you're running stock voltage or not, but 7z bench can definitely reach sane power limits on ADL-S.
Let's say you have CPU that uses 10W per second in that benchmark and finishes in 10 seconds, that would be 100W to finish that bench, if they ran it 3 or more times and got the average of those 3 or more runs to make sure that it's accurate then that's that.
You can run way lower power but the average amount of power to finish the job will still be high.
Not the average amount of power used while running the bench but the sum of power to complete it.
Hello There ! Need some help for doing a hard choice !
I wanted to upgrade to an i7-12700 as i'm buying a fanless / semifanless customised machine. (the computer is build by the manufacturer of the case).
only 65W cpu are allowed, but with a max of 120/140W, so when i saw that the i7-12700 was reaching more than 180w, it was a"no more".