When Intel's Arrow Lake chips were released, the tech world was hopeful that the company would release a generation of processors without the issues of the previous two. So far, that's been true enough, as Arrow Lake doesn't appear to have the same overvoltage issue that led to Raptor Lake and its refresh deteriorating prematurely.

But, that's the only good news out of the release, as the Intel Core Ultra 2 launch came with its own laundry list of issues. Performance was all over the map, with reviewers having wildly different benchmark results, often from the same setup. RAM compatibility appears to be another large issue, with some kits refusing to boot with XMP settings and others refusing to boot at all.

Intel has investigated since the launch and found some things that didn't line up with its internal testing, promised updates to fix the performance issues, and quickly released a succession of updates, both to Windows and to motherboard manufacturers who rushed out new BIOS versions. Except, the performance doesn't appear to have improved, and we've noticed RAM compatibility regression with the kit that we used for one of our Arrow Lake reviews.

Arrow Lake didn't start off on the right foot

Inconsistent benchmarks and Windows weirdness was everywhere

Arrow Lake's launch didn't go smoothly, with performance issues that persisted after trying every troubleshooting step we'd learned in our years of testing. Some of these issues were clearly BIOS bugs with settings not being changed properly, like voltages on DDR5 RAM when XMP was enabled. Others were Windows issues, as changes to the power plan helped, even if they didn't fix everything.

To Intel's credit, they took immediate ownership of the issues, saying that optimization issues were the culprit, and promising some fixes before the end of the year. The company also said that it didn't notice these issues with internal testing and was slightly confused as to why reviews had shown otherwise. Whatever was happening, the message was ,"we hear you, we're looking into it, and we'll endeavor to fix it."

Intel promised it could fix things

Internal testing before launch was inconsistent with review benchmarks

Midway through December, Intel made an announcement about the progress on Arrow Lake's fixes. The company had found five things that were hampering performance, and at least three of them were already out in the wild.

1. A missing Performance & Power Management (PPM) package.
2. Intel Application Performance Optimizer (APO) could not take effect.
3. BSODs when attempting to launch game titles utilizing the Easy Anti-Cheat service.
4. Select performance settings misconfigured on reviewer or early enabling BIOSes.
5. New BIOS performance optimizations.

Two of the fixes were related to missing updates in Windows, which Arrow Lake needs to perform at its best. Those were already inserted into Windows Updates before the announcement, and many (including me) had been using them without knowing. One was a fix for Easy Anti-Cheat, which fixes the issues with Arrow Lake computers being kicked out of multiplayer games.

The last two are BIOS updates, one available now from Z890 motherboard manufacturers and another update to come in January 2025. I've updated to the latest BIOS version with the relevant settings enabled, and I can't tell you it's made any difference to my results. Intel says this particular BIOS could recover 2-14% of performance depending on the task, but 2% is within the margin of error for many tests, so it's possible I haven't hit the ones that made the large improvement.

Even with that said, the testing programs we use here at XDA are chosen to show the tasks for which most of our users would be using these CPUs. A 2% increase is nothing, really, and disappointing when Intel said their internal testing was higher. Also, the RAM settings that worked during our review of the Core Ultra 9 285K are no longer booting, so we're stuck at 5,600MT/s or 6,400MT/s currently.

We'll be rechecking when the last promised BIOS update is ready for testing, but the picture at this point seems dismal.

The fixes so far are inconclusive

RAM compatibility seems to be hit-and-miss, mostly miss

None of the fixes Intel has identified, mentioned, or promised have to do with the largest issue I have with Arrow Lake right now. That's RAM compatibility, which is abysmal. It feels like the AMD Ryzen Zen 1 launch when the majority of DDR4 kits available either wouldn't boot at the speeds they were sold at, or straight up refused to boot at all. At that time, RAM kits using Samsung's B-Die quickly emerged as the savior of the platform, but I've yet to find a similar kit this time around.

I've amassed a large number of DDR5 kits over the years, from 5,600MT/s all the way through to 8,800MT/s, and a wide variety of latencies and capacities. I've tried them all with the Core Ultra 9 285K, with varying success. Maybe I shouldn't say success, but managing to boot into Windows on some kits did feel like a victory after days of trying combinations of settings with each different kit.

There is a small chance it's my specific motherboard configuration, but I'm not the only one I know struggling with RAM compatibility, so it appears to be a platform issue. This can likely be solved by Intel, but the fact they haven't mentioned anything yet suggests that it's a problem that doesn't have an easy or quick fix.

The jury is out, but there's hope

So far, the updates that Intel has rolled out for Arrow Lake for Windows 11 and motherboard BIOSes haven't affected my experience since I reviewed the Intel Core Ultra 9 285K. I still see lower-than-expected performance, to the point where 12th-generation Intel chips are putting up bigger numbers. It's possible that a different LGA1851 motherboard could change things, but based on the in-depth testing of TechPowerUp, I'm not holding my breath.

Their testing also included the 0x114 microcode, which was the final piece of the fixes Intel promised. While it brought some small gains, it didn't improve overall positioning against AMD Ryzen or even Intel Raptor Lake. The microcode does improve things for Windows 11 24H2 users but doesn't change much for 23H2 users, which is interesting, to say the least.

Will Intel be able to identify more fixes for future microcode updates, or was the late switch to TSMC's process and the internal struggles of 18A just too much for Arrow Lake? Only time will tell, but at this point, I'm leaning towards the latter option. Intel will need to swing for the fences with 20A and hope that the next desktop release is a hit, otherwise it's going to be stuck in a rut for years, just like AMD was during the Bulldozer era.