News Alder Lake Allegedly Commands 28% Higher Peak Current Than Rocket Lake

Will this require new power supplies or is this increase well within the performance envelope of current PSUs?

If it is a change in standards but one that all supplies already currently meet then it is no big deal.

If it means that every Alder Lake system will require a new reworked PSU then it is a very big deal.
 
Will this require new power supplies or is this increase well within the performance envelope of current PSUs?

If it is a change in standards but one that all supplies already currently meet then it is no big deal.

If it means that every Alder Lake system will require a new reworked PSU then it is a very big deal.
Of course they do. And all existing 500W power supplies (and above) are even capable of suppling continuous current greater than what these peak current requirements are. Also if you look at the table, Intel's 10th gen 165W cpus (these are the Cascade lake cpus like the 18-core 10980XE) have a peak current requirement of 40A. The peak current requirement for 12th gen 125W Alderlake-S (according to this leak) is 39A which is lower than that. So unless a PSU was not capable of running existing HEDT cpus at stock then it should be able to run Alderlake as well.
 
Last edited:
Of course they do. And all existing 500W power supplies (and above) are even capable of suppling continuous current greater than what these peak current requirements are. Also if you look at the table, Intel's 10th gen 165W cpus (these are the Cascade lake cpus like the 18-core 10980XE) have a peak current requirement of 40A. The peak current requirement for 12th gen 125W Alderlake-S (according to this leak) is 39A which is lower than that. So unless a PSU was not capable of running existing HEDT cpus at stock then it should be able to run Alderlake as well.
That's 500w for your entire machine, though, not just to the socket. A decent GPU in the mix and you definitely need a beefy PSU to keep up with peak consumption.
 
  • Like
Reactions: prtskg
Peak current doesn’t mean maximum sustained current – that’s what continuous current is for. Peak current is the peak value it will reach momentarily during a transient that lasts microseconds or milliseconds. What’s important with overcurrent is heat – and there is no much heat produced even if you are pulling a KA if it only lasts a microsecond. Have you ever looked at a time-current characteristic curve of a fuse? And yes, I know, there are secondary long-term modes of failure where peak current is involved but that’s beside the point.

And your argument of what happens when you pair it with a GPU is confusing things as you are now talking about a scenario where sustained power can exceed the power rating of the PSU. Sure, if you have a 125W cpu with a PL2 limit of 228W (19A@12V) and you unlock power limits and run a heavy AVX512 sustained workload and at the same time you also pummel your 320W RTX 3080 (26.67A@12V) with Furmark or some other torture test, you will have a sustained power consumption of over 500W (41.67A@12V) thereby exceeding the maximum sustained power rating of your PSU (and its maximum continuous current of the rail at 12V) . But peak current doesn’t come into consideration whatsoever. You would still be running out of Watts (or continuous amps in the 12V) in such scenario regardless of the momentarily peak current value.
 
Alder Lake does not have avx512.

It does have pcie5 and ddr5 io, both features supporting significantly higher data rates than supported by Intel's prior chips, so would be expected to require higher power than the pcie4 ddr4 chips while running io at full rate.

It also has the Gracemont small cores, so there is the potential to run at lower power when not doing compute intensive or io intensive tasks.

Samsung created some laptops with TGL that can be switched to fanless operation. I suspect that will be even easier with Alder Lake, just by turning off its Golden Cove cores.
 
No surprise there.

Intel is doing now with CPUs, what AMD was doing with GPUs (to keep up with Nvidia). I.e: Large increases in power draw compared to past gen, to get a bit more performance. It was a desperate act for AMD to do in the past to keep up with Nvidia (not the case anymore though), and now Intel is doing the same desperate act to keep with to AMD.

The problem here is, however, that when AMD was trying to stay competitive with Nvidia, AMD was selling their video cards for same or less than Nvidia's (performance being equal between the two). Meanwhile, Intel's processors, being inferior to AMDs, are mostly selling for more than what AMD is asking for. And if you take the total-cost-of-ownership of a CPU into perspective, including the electricity cost, a cheaper to buy Intel CPU (not that many - mostly low end) becomes more expensive because of the need to pay for the extra electricity to power them over their lifetime.

Intel is hurting. It will take them another ~3/4 years to catch up even if they throw ton of money at the problem, because AMD isn't standing still. In the meantime, buy some AMD shares and sell them when it look like Intel is about to come up with something competitive....maybe 2024/2025