Intel Arrow Lake Power Testing Might Not Be Easy: Power Test Bench Build Log
video description
Date: 2024-10-23
Related videos
Comments and reviews: 20
CatalystDestiny
Arrow lake is going to be a challenge to fully benchmark test properly I think, but it's going to be required to test it in quite a lot of different ways. CPU power is one thing with those leaked full power benchmarks causing noticeable performance increase, so good reviewers need to test all different power setups. RAM is going to be the huge one though. I've seen some reviewers say they're choosing only one type of RAM to test and this, IMO, will taint the data because arrowlake supporting CUDIMM's IS one of its features and may make a big impact in performance once fast kits hit. Testing will need even things like both regular DDR5 RAM tests, and CUDIMM RAM tests side by side to see comparisons, differences, and even potential for those considering arrow lake for their next upgrade compared to AMD.
reply
Arrow lake is going to be a challenge to fully benchmark test properly I think, but it's going to be required to test it in quite a lot of different ways. CPU power is one thing with those leaked full power benchmarks causing noticeable performance increase, so good reviewers need to test all different power setups. RAM is going to be the huge one though. I've seen some reviewers say they're choosing only one type of RAM to test and this, IMO, will taint the data because arrowlake supporting CUDIMM's IS one of its features and may make a big impact in performance once fast kits hit. Testing will need even things like both regular DDR5 RAM tests, and CUDIMM RAM tests side by side to see comparisons, differences, and even potential for those considering arrow lake for their next upgrade compared to AMD.
reply
opachki8325
as weird and bad as intel cpus can be I'd like to give them a shoutout for actually keeping their cpu-pch interface large as opposed to amd who have like half the bandwidth. I'm just pissed the x870 boards don't really offer a significant number of m.2 drives.
Why would I need 4 or more Well. Games. I don't wanna have to delete them. And I mean, a 4090 is capable of being limited by a 4.0x8 pcie interface. I've seen comparisons of up to 10% (would be cool if you'd test that) variance between those which imo is a significant bit. Might not be a big deal but still, it's unused potential just because a manufacturer wanted to cheap out.
reply
as weird and bad as intel cpus can be I'd like to give them a shoutout for actually keeping their cpu-pch interface large as opposed to amd who have like half the bandwidth. I'm just pissed the x870 boards don't really offer a significant number of m.2 drives.
Why would I need 4 or more Well. Games. I don't wanna have to delete them. And I mean, a 4090 is capable of being limited by a 4.0x8 pcie interface. I've seen comparisons of up to 10% (would be cool if you'd test that) variance between those which imo is a significant bit. Might not be a big deal but still, it's unused potential just because a manufacturer wanted to cheap out.
reply
gagarin777
The idea with measuring power out of the pinout or onboard elements is not realistic. That is because to measure the DC power draw you need to monitor voltage and current, but from stabbing elements you can only get the voltage. To monitor the current you would need to physically replace the paths on the mobo with your own wiring and pass it through the multimeter/currentmeter or similar device and spill it back to motherboard. And if you decide to read the current out of the power section IC's then that would be the same as using HWinfo, Aida or any monitoring software you wanted to avoid in the first place.
reply
The idea with measuring power out of the pinout or onboard elements is not realistic. That is because to measure the DC power draw you need to monitor voltage and current, but from stabbing elements you can only get the voltage. To monitor the current you would need to physically replace the paths on the mobo with your own wiring and pass it through the multimeter/currentmeter or similar device and spill it back to motherboard. And if you decide to read the current out of the power section IC's then that would be the same as using HWinfo, Aida or any monitoring software you wanted to avoid in the first place.
reply
brianfamilyguy
So possibly dumb question, if someone didn't have the elmor panel could you measure every other part besides the PCIB. Measure the total power using values at supply (minimal losses). Then subtract other parts from the total power to get the power draw from PCIB and other parts This would indicate if you would have to even test for 'missing' power. It is a cool showcase for the elmor lab gear.
reply
So possibly dumb question, if someone didn't have the elmor panel could you measure every other part besides the PCIB. Measure the total power using values at supply (minimal losses). Then subtract other parts from the total power to get the power draw from PCIB and other parts This would indicate if you would have to even test for 'missing' power. It is a cool showcase for the elmor lab gear.
reply
stephanematis
I have PTSD for this, reminders of lab work and our Frankstein hand soldered power monitoring cables, oscilloscopes, custom data parsing, historical logging setups, etc ... essentially, a birds nest of wiring. And the signs DO NOT TOUCH ... which somehow got ignored once overnight.
This is a cleaner setup but I still get flashbacks.
reply
I have PTSD for this, reminders of lab work and our Frankstein hand soldered power monitoring cables, oscilloscopes, custom data parsing, historical logging setups, etc ... essentially, a birds nest of wiring. And the signs DO NOT TOUCH ... which somehow got ignored once overnight.
This is a cleaner setup but I still get flashbacks.
reply
1050-Argon
Very interesting stuff! From what I have heard so far this appears to be different depending on the motherboard vendor and possible even depending on the SKU. Roman (der8auer) has mentioned that this appears to be an ASUS thing espeically on boards that also have VRMs on the south side of the CPU socket. We'll see how it develops.
reply
Very interesting stuff! From what I have heard so far this appears to be different depending on the motherboard vendor and possible even depending on the SKU. Roman (der8auer) has mentioned that this appears to be an ASUS thing espeically on boards that also have VRMs on the south side of the CPU socket. We'll see how it develops.
reply
gamersnexus
Maybe move the GPU out to a slot extender, cut power to the MB, and them power the extender board from a power connection added to the extension board. Example something like PCIe 4.0 x16 Slot Dual Port Backplane Adapter from microsatacables might work. Similar to what you did but off the board.
reply
Maybe move the GPU out to a slot extender, cut power to the MB, and them power the extender board from a power connection added to the extension board. Example something like PCIe 4.0 x16 Slot Dual Port Backplane Adapter from microsatacables might work. Similar to what you did but off the board.
reply
NecroFlex
Do not convert pcie and EPS between eachother
Me, who converted an 8-pin PCIe to fit into an 8-pin EPS socket (and flipped the pins) to get power from 2x 6-pin PCIe connectors cause i needed a cable to use on the tesla cards that need EPS not PCIe cables: Yeah, what he said, don't do it
reply
Do not convert pcie and EPS between eachother
Me, who converted an 8-pin PCIe to fit into an 8-pin EPS socket (and flipped the pins) to get power from 2x 6-pin PCIe connectors cause i needed a cable to use on the tesla cards that need EPS not PCIe cables: Yeah, what he said, don't do it
reply
jsphj
I would take 2 of the ATX 24-pin connectors at this point as long as it’s not lighting the board or cards on fire anymore. Sell them on the choice between dual socket or a big gamer GPU in one of the slots on a single socket board. Bring back the on-card SLI and Crossfire with this also.
reply
I would take 2 of the ATX 24-pin connectors at this point as long as it’s not lighting the board or cards on fire anymore. Sell them on the choice between dual socket or a big gamer GPU in one of the slots on a single socket board. Bring back the on-card SLI and Crossfire with this also.
reply
EdDale44135
It I interesting to see the hoops you have to jump through to get to a point where you can be confident in your results. Particularly when tracking down a strange rumor. Looking forward to hearing what the results are for this test and the CPU performance, once the embargo lifts.
reply
It I interesting to see the hoops you have to jump through to get to a point where you can be confident in your results. Particularly when tracking down a strange rumor. Looking forward to hearing what the results are for this test and the CPU performance, once the embargo lifts.
reply
aVesela3_14cka
in regulary gaming scenario , moving my 990 PRO nvme in to a lower slot and not share PCI- lanes is better than putting it in upperr slot
Thank you for all your work i can see a lvl of enjoj in your eyes while working of this type of staf :D Greetings from Croatia !
reply
in regulary gaming scenario , moving my 990 PRO nvme in to a lower slot and not share PCI- lanes is better than putting it in upperr slot
Thank you for all your work i can see a lvl of enjoj in your eyes while working of this type of staf :D Greetings from Croatia !
reply
1Grainer1
wouldn't riser cable with telemetry be a banger for testing you could perfectly isolate GPU power draw from the rest... i doubt we can isolate ram etc., so there would be a bias, but totally excluding GPU power draw from system power draw would be a massive for testing data
reply
wouldn't riser cable with telemetry be a banger for testing you could perfectly isolate GPU power draw from the rest... i doubt we can isolate ram etc., so there would be a bias, but totally excluding GPU power draw from system power draw would be a massive for testing data
reply
bryantchan6156
Thanks for running through this testing process. Quick question even though you can isolate the pci-e power from 24 pin, but doesn’t the 24 pin also power the ram, nvme, and chipset How can you isolate those components from the 24 pin power
reply
Thanks for running through this testing process. Quick question even though you can isolate the pci-e power from 24 pin, but doesn’t the 24 pin also power the ram, nvme, and chipset How can you isolate those components from the 24 pin power
reply
Shepherd_Solutions
Having to use a screwdriver to fully seat a 12VHPWR cable made it the sketchiest part of that power supply cable spaghetti build. Love the format of showing off crazy and cool solutions to things other people might not consider. Thanks!
reply
Having to use a screwdriver to fully seat a 12VHPWR cable made it the sketchiest part of that power supply cable spaghetti build. Love the format of showing off crazy and cool solutions to things other people might not consider. Thanks!
reply
toddabbott781
But if you take a a 14900K and compare it to this system using the same memory, storage, PSU, graphics... you should get a good comparison just using total system power. If they say it runs 70w less then it should show that 70w difference.
reply
But if you take a a 14900K and compare it to this system using the same memory, storage, PSU, graphics... you should get a good comparison just using total system power. If they say it runs 70w less then it should show that 70w difference.
reply
graxxor
Love to see your BTS videos. Nerds rejoice!! This is the kind of hands-on stuff this particular home made PC builder enjoys to watch. A great stream of consciousness counterpart to your thorough news segments and highly produced documentaries.
reply
Love to see your BTS videos. Nerds rejoice!! This is the kind of hands-on stuff this particular home made PC builder enjoys to watch. A great stream of consciousness counterpart to your thorough news segments and highly produced documentaries.
reply
ModernOddity728
After MLID just leaked how varied performance with Arrow Lake is run-to-run, paired with more complicated power testing required to see the actual power draw, I cannot wait to see the results. Thank you guys so much for what you do!!
reply
After MLID just leaked how varied performance with Arrow Lake is run-to-run, paired with more complicated power testing required to see the actual power draw, I cannot wait to see the results. Thank you guys so much for what you do!!
reply
burakozc3079
If intel did carry 14th gen to tsmc n3 with almost no change in design (except bug corrections) it would be better both performance and efficacy. It is ridiculous they wasted their time to design this ultra crap.
reply
If intel did carry 14th gen to tsmc n3 with almost no change in design (except bug corrections) it would be better both performance and efficacy. It is ridiculous they wasted their time to design this ultra crap.
reply
lyteness859
Would you measure the total system power from after the power supply, because the power supply efficiency would make the numbers different with different rating power supplies if you measure it from the wall.
reply
Would you measure the total system power from after the power supply, because the power supply efficiency would make the numbers different with different rating power supplies if you measure it from the wall.
reply
kaseyboles30
7:10 ish. Camera man missed an opportunity there, should have given a thumbs up in front of the camera just to tie that reference even deeper. Of course I didn't think of it either for almost a full minute.
reply
7:10 ish. Camera man missed an opportunity there, should have given a thumbs up in front of the camera just to tie that reference even deeper. Of course I didn't think of it either for almost a full minute.
reply
Add a review, comment
Other channel videos
