Arrow Lake Retested with Latest 24H2 Updates and 0x114 Microcode

[W1zzard]

Thanks W1zzard!
BIOS 1203 for the ROG MAXIUMS Z890 Hero is out with the 114 ucode
If anyone flashes back and forth to do comparison testing, there is a chance that the ME firmware could impact results as it does not downgrade in most situations when using BIOS flashback etc.

A newer ME firmware is also on the ASUS ROG forums (19.0.0.1854), it is newer than the ME firmware contained in bios 1203 (19.0.0.1827).

2024/12/18
"1.Intel microcode updated to 0x114..
2.M.R.C updated to v1.4.6.64.
3.ME FW updated to v19.0.0.1827.
4.GOP updated to v1057.
5.RST VMD updated to v20.1.0.5850.
6.Improved memory compatibility and stability for frequencies >6000MHz; added CUDIMM Dual PLL mode.
7.Enabled Wi-Fi 7 (320MHz) and Wi-Fi 6E (6GHz) support for multiple countries.

Updating this BIOS will simultaneously update the corresponding Intel ME to version v19.0.0.1827.
The ME version will remain updated even if you roll back to an older BIOS later.
We recommend using EZ Flash to update the BIOS, as it supports ZIP format and auto-updates the ME.
If using USB BIOS FlashBack, ensure the ME version matches the BIOS version. - this should mean ensure the ME firmware is equal to or newer than the target BIOS, in this case 19.0.0.1827 or later.

Starting testing right now, will update this article accordingly

@W1zzard pls - where are the "average across apps" and "average across games" charts I come to TPU for?

Not in this article. Time is limited, and no reason to still be benching on xmas eve, for these meager gains

"All eyes are now on the 0x114 Microcode Update Intel is planning to roll out in January 2025."

How should this be interpreted?

Intel's wording (https://www.techpowerup.com/329990/...formance-targets-jan-2025-for-0x114-microcode) keeps talking about mid-Jan, but the ASUS Hero BIOS just came out with 0x114, so no idea why Intel picked this date .. I'll be testing it right away and update this article

edit: bios update started, now the updater is caught in a reboot loop that power cycles the machine
edit2: long power off didnt help
edit3: cmos clear worked, says "updating led firmware now"
edit4: flashing same bios once more, to be sure the update worked correctly
edit5: worked perfectly this time

 

[westlake]

It should be clear to everyone that Intel (Robert Hallock) is talking nonsense. They are doing nothing but confusing users. And in my opinion it was unnecessary to make so many measurements, one "then" (ARLS release) and one now with the latest published versions would be enough. These two would show how much of what the blues claim is true. Otherwise, essentially nothing. You can investigate how much performance has changed as a result, but what's the point? Ultimately, everyone is interested in the end result.

 

[Max Mojo]

0x114 available already since 2024/12/18 :

 

[westlake]

@W1zzard

You do know that you need to update your ME, right?

 

[Max Mojo]

@W1zzard

You do know that you need to update your ME, right?

It's integrated in the BIOS Update.

 

[westlake]

It's integrated in the BIOS Update.

I think you should take a closer look.

 

[Max Mojo]

I think you should take a closer look.

ME 19.0.0.1854

 

[westlake]

ME 19.0.0.1854

And which version is in the BIOS?

 

[Max Mojo]

Probably you know.

 

[westlake]

Yes, I know, as you should too, since you've linked to a picture that contains the information.

 

[W1zzard]

@W1zzard

You do know that you need to update your ME, right?

If you flash the ZIP (dont unpack it), it updates both. But I'll test with the newer ME too

 

[westlake]

If you flash the ZIP (dont unpack it), it updates both. But I'll test with the newer ME too

I know exactly how the upgrade works on ASUS Z890 motherboards, but the focus is on the newer version. Especially since Intel claims that the 114 microcode "requires" ME 19.0.0.1854.

 

[Tom Yum]

Intel arrogance forever.

It might be as basic as Intel not having the margin to do so, especially for the 245 and 265k. Arrow Lake uses 243mm2 of TSMC expensive N3B process, Raptor Lake uses 257mm2 of the cheaper (to Intel) Intel 7 (10nm) process.

When you also factor in the packaging complexity of Arrow Lake, the rumoured loss of TSMC discounts, and the fact Arrow Lake doesn't really move the performance needle from Raptor Lake (performance per watt improvements aside), from a business perspective it makes sense to keep Arrow Lake volume low (and prices high) and maintain supply of the cheaper Raptor Lake processors.

 

[Hattu]

“We didn't allow Windows Update to install all the updates and started benching instead.”

just 24 h2 none of the patches

My bad. Thanks, somehow I didn't notice that.

 

[W1zzard]

New charts are up with 0x114 and also 0x114 and manual ME update.

Will update the texts

 

[westlake]

Thank you for your work! It can be seen that Intel's claims of improvement are not true.

 

[W1zzard]

All charts and texts have been updated

 

[CyberPomPom]

Moving back to single-threaded cores doesn't hurt multi-thread applications. SMT was introduced to improve performance when CPUs were a single core, and then to improve thread count when core count was low because Intel was stuck on large processes. Now that we have advanced processes and can fit dozens of cores on a single CPU, Intel has shown they can get much better results by carefully optimizing smaller, single-threaded cores instead of wasting space on the components needed for SMT. Just look at the benchmarks. It matches the 9950X in multi-thread Blender rendering and bests it in Cinebench multi-thread, despite having 8 less threads and most of its cores running significantly slower than the full cores on the 9950X.

Every company has bad launches. Just look at the review for the initial Ryzen 7 1800X. https://www.techpowerup.com/review/amd-ryzen-7-1800x/16.html It had many of the same problems Arrow Lake has. Poor game performance with new architecture, apps/OS needed to be optimized for the dual-CCD design, etc. I'm not saying Arrow Lake is great, if you just look at performance it's a sidegrade to Rapter Lake and Zen 5. I'm saying the new architecture is the right direction. I will not be surprised when AMD follows suite with single-threading cores and big/little desktop CPUs in the next generation or two. It has obvious advantages to performance.

I don't know much about the whole SMT against p/e cores in terms of die area cost versus performances so I won't comment on it even though it doesn't look like such a clear for Intel as it requires a lot more cores and better scheduling.

No, what is wrong is your description of the 1800X launch. It didn't have most of the flaws you cite.
Game performance was about twice better than their previous architecture and it had no CCDs to optimise for as they appeared with Zen2.
And power efficiency increase versus previous gen was even better than ARL.
And the brand new SMT was praised with a 30% scaling versus the 15% of HT in Intel parts.
The main issue I remember was a very touchy memory support at a time when Intel was so easy to use.
Almost every reviews of the 1800X were praising its price to performance ratio, even with its obvious new platform flaws.

I don't know how it should compares to ARL that only improves power consumption against previous gen, is not priced competitively and which technical flaws are not so obvious to correct.

 

[LittleBro]

So, for software 24H2 with 0x113 microcode is best, for games 24H2 with 0x114.
There is absolutely NO 6-30% performance improvement on average, so Hallock is shitting bollocks or hallucinating or both at same time.

I wonder what actually changed with microcode updates. Hopefully someone will do a retest also on Linux and Win10.
@W1zzard, are there any noticeable changes to power draw? Voltage/frequency curve? Memory latency?

It's more than clear that Tom's Hardware was right about Arrow Lake suffering from a hardware flaw. Moving memory controller out of compute tile imposes 20-25ns memory penalty and mixing P/E cores (to improve temps) increases L3 cache latency. AMD chiplets handle memory latency better, but not frequencies. Now imagine performance boost that AMD could achieve by moving IMC into tile with cores along with releasing new improved IOD. I'd say Zen 6 does not require any significant changes to core architecture (as introduced with Zen 4 vs. Zen 5) to increase performance, only changes to IMC location and IOD bandwidth are required.

Also, AMD should not drop SMT, as it is delivers much more performance than Intel's HT and produces no power draw increase.

 

[warriorunited]

I can't even download the ME 19.0.0.1854 from Asus forum, it's blocked all the time reporting as a Virus. I think i will wait for updated ME Tool.

 

[W1zzard]

I can't even download the ME 19.0.0.1854 from Asus forum, it's blocked all the time reporting as a Virus. I think i will wait for updated ME Tool.

Attached here, if that helps. The exe is digitally signed by Intel, and I doubt the ME files will flash if they have been tampered with

are there any noticeable changes to power draw? Voltage/frequency curve? Memory latency?

Didn't test, no time and no point given these minimal differences

 

[warriorunited]

@W1zzard Same result, i can't download the file, instantly reported with virus.

Trojan:Script/Wacatac.B!ml

 

[W1zzard]

@W1zzard Same result, i can't download the file, instantly reported with virus.

Trojan:Script/Wacatac.B!ml

Meh .. guess report to your antivirus vendor so they can investigate and fix

 

[friocasa]

So, about the same

"I expected nothing, and I'm still disappointed"

 

[_roman_]

The broken INTEL Promise. No positive, impressive XMAS 2024 surprise from INTEL.

I miss those poeple who wrote pages over pages on comments on ryzen 9000. This intel issue is far worse with barely any attention. Every company should get their equal share of postive and negative comments.

---

Thank you for all the efforts and time spend for those tests.


I do not see a difference. I do not want to copy blender and the other picture.

When I just use my calculator and divide = 1.02977
I do not want to round up - So there is not even a 3 percent deviation on windows. That's not really a big difference. (I see it money wise, not enough, round down always).
Intel should have labelled the processor with a lower number - as it performs much worse.