Photo Credit: Gigabyte
Gigabyte has launched their budget LGA-1700 ITX motherboard under the UD line; the H610I DDR4. "Features" include support for up to DDR4-3200 memory, a PCI-e 4.0 expansion slot, and support for a single PCI-e 3.0 M.2 drive. The board features a 4+1+1 VRM design with no cooling. According to Gigabyte's website, the H610I DDR4 supports an i9-12900K; a processor known to draw 250 watts under load...

My take on this:

I normally try to keep my opinion out of this but nothing pisses me off more than motherboards with VRMs that don't even include at least a $1 chunk of metal. The 10 phase VRM Gigabyte Z690I, with a legitimate heatsink, hits 95C in my Meshlicious.
Please don't put an i9-12900K in this board.

The price wasn't available at the time of writing but if this sells for more than $100 then the world is truly lost. Gigabyte should bundle this with a Celeron and AARP membership because this board is probably going to end up in your grandmother's computer for e-mail and Facebook. I'm not even going to link this board. Instead, here is Buildzoid talking about Z690 boards.

 
Last edited:

Gotmachine

Chassis Packer
Oct 7, 2020
14
29
The 10 phase VRM Gigabyte Z690I, with a legitimate heatsink, hits 95C in my Meshlicious.
Please don't put an i9-12900K in this board.
I expect PL1/PL2 to be limited to something like 100/125W in the BIOS. There are tons of low-mid level boards with 4/5/6 CPU phases and zero VRM extra cooling that perform perfectly fine, but they have max power limits set according to what the board can deliver. For example the ASrock B660M-ITX/ac (5 CPU phases) has 150/180 PL1/PL2 limits, a very minimal VRM heatsink that cover only 4 phases, and it perform perfectly well thermally. Arguably, many boards VRM heatsinks are oversized junk that perform terribly, and some boards just have sub-optimal VRM designs that will overheat no matter what.

Technically, you can put any 12th gen SKU on any board, and in most real world use cases even a 12900K will perform nominally, only heavily multi-threaded performance will suffer significantly when power limits are too low.

Something else to consider is that boards (and especially entry-level boards) are designed for top-flow CPU coolers (like the intel stock ones) providing quite a bit of airflow for neighboring components. In that regard, layouts using only an AIO and zero extra ventilation reaching the board surface can be a big issue.
 

Revenant

Christopher Moine - Senior Editor SFF.N
Original poster
Revenant Tech
SFFn Staff
Apr 21, 2017
1,670
2,697
I expect PL1/PL2 to be limited to something like 100/125W in the BIOS. There are tons of low-mid level boards with 4/5/6 CPU phases and zero VRM extra cooling that perform perfectly fine, but they have max power limits set according to what the board can deliver. For example the ASrock B660M-ITX/ac (5 CPU phases) has 150/180 PL1/PL2 limits, a very minimal VRM heatsink that cover only 4 phases, and it perform perfectly well thermally. Arguably, many boards VRM heatsinks are oversized junk that perform terribly, and some boards just have sub-optimal VRM designs that will overheat no matter what.

Technically, you can put any 12th gen SKU on any board, and in most real world use cases even a 12900K will perform nominally, only heavily multi-threaded performance will suffer significantly when power limits are too low.

Something else to consider is that boards (and especially entry-level boards) are designed for top-flow CPU coolers (like the intel stock ones) providing quite a bit of airflow for neighboring components. In that regard, layouts using only an AIO and zero extra ventilation reaching the board surface can be a big issue.

I accept no excuses from the manufacturers. If you can’t support PL1 and PL2 as marketed for the 12900K then it’s not fully compatible. While it may be technically correct that it can run a 12900K it’s practically wrong because you’ll never see the full performance of the chip.

If manufacturers made it plainly clear that you can’t expect full performance than it’s a different story.

For example: Asus lists the 9900K as fully compatible with the Z370-I board. During our internal testing we discovered that once you exceed about 150 watt pull for the chip, the VRM throttles hard. Even placing a high RPM fan on the VRM cooling blocks did nothing to help. Removing the blocks cut the performance down even further to about 100 watts even with the fan. For the full performance of the 9900K you need over 200 watts of support.

While 95 watts is the technically correct spec range it’s also practically wrong as even Intel showed benchmarks at far higher wattage. People will buy a 9900K as an upgrade and then be frustrated when it throttles down to a fraction of the all core 4.7ghz they saw in the reviews. This is marketing BS and manufacturers need to be called out on it or we aren’t doing our jobs as journalists.

…and that’s exactly what’s happening here. Intel showed performance numbers from a fully powered 12900K. They sure didn’t tell reviewers to limit it to 125 watts. So if a motherboard manufacturer wants to say it’s compatible with the 12900K, it better be fully compatible with the 12900K or they will be called out on their BS. This is the reason why MB manufacturers don’t want to sample low end boards to us reviewers; we will hold their feet to the fire for poor designs.
 

Gotmachine

Chassis Packer
Oct 7, 2020
14
29
If you can’t support PL1 and PL2 as marketed for the 12900K then it’s not fully compatible
I agree that it's unfortunate that MB vendors don't communicate clearly what power limits the board power delivery is designed for.
Intel showed performance numbers from a fully powered 12900K. They sure didn’t tell reviewers to limit it to 125 watts.
To me, the core of the issue is that paradigm shift where Intel (and AMD to a lesser extent) are essentially making self-overclocking chips. Pushing 250W in a CPU is nothing new, it's just that it wasn't officially supported like it is now, and nothing has changed in practice : to get that overclocked performance, you need to pay a huge premium on all components (PSU, MB, cooling...), and you can't expect it on entry-level cheap hardware.

The point is, at the exception of a handful of very specific workloads, you're letting very little performance on the table by running a 12900K or any of the other high end SKUs at a 125W power limit. If you account for total platform cost vs performance, it make a lot of sense to use such power limits : lower wattage PSU, cheaper cooling solution, better efficiency... This still is a valid choice and use case.

For the vast majority of usages, single-threaded performance on a handful of threads is still king, and things aren't going to evolve much from now on. Contrary to popular belief, we already have more or less reached the peak of multithreaded software implementations. The 12900K at 241W Cinebench MT scores are just that : scores. This is useless unless you're actually doing 3D renders daily (which doesn't happen in reality, even in a low scale professional environment you would have a dedicated render machine for that).
This is the reason why MB manufacturers don’t want to sample low end boards to us reviewers; we will hold their feet to the fire for poor designs.
It seems boards manufacturers have learned their lesson with 10/11th gen, and are now locking max PL1/PL2 on low end boards in the BIOS so end users don't do silly things (and I wouldn't be surprised if Intel lobyied quite a bit to enforce that). But still, yep, they should definitely list what those power limits are in the spec sheets.