• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Arrow Lake 2024 (and beyond)

Arrow Lake Refresh but just K and KF chips are going to made. I'll wait around for Nova Lake hopefully that will come closer to AMD offerings. It's going to be funny no matter what Intel does with Arrow Lake Refresh which will make Arrow Lake a dead socket after launch anyway.
I think people got the wrong impression with the arrow lake. Because it looks to me like it was just a more efficient 12th or 13th gen chip than something that was supposed to be the next step up.
 
I think people got the wrong impression with the arrow lake. Because it looks to me like it was just a more efficient 12th or 13th gen chip than something that was supposed to be the next step up.
We’ve known that for a while. Royal Core was supposed to be the next real one, but it got screwed up by various bits of internal politics and Kessler left
 
I think people got the wrong impression with the arrow lake. Because it looks to me like it was just a more efficient 12th or 13th gen chip than something that was supposed to be the next step up.
Its a completely different core architecture. And not only due to tiles/chiplets.
The cache setup is completely different. The layout of the cores into groupings, is different.
I think the only thing somewhat similar, is its another 'Ring' setup, to connect everything. But im sure aspects of that ring are different. Due to the completely different core layouts, tiles, etc.

And without Hyperthreading, they have totally changed the way the cores process threads. My 12 e-cores score as well in Cinibench R23, as the 3900x I used to have (12 cores/24 threads). And R23 usually scores about 40% worse without SMT/Hyperthreading.
 
Its a completely different core architecture. And not only due to tiles/chiplets.
The cache setup is completely different. The layout of the cores into groupings, is different.
I think the only thing somewhat similar, is its another 'Ring' setup, to connect everything. But im sure aspects of that ring are different. Due to the completely different core layouts, tiles, etc.

And without Hyperthreading, they have totally changed the way the cores process threads. My 12 e-cores score as well in Cinibench R23, as the 3900x I used to have (12 cores/24 threads). And R23 usually scores about 40% worse without SMT/Hyperthreading.
Well it was supposed to be a redesign and that was expected. However in the discussion 7 years ago Intel getting feedback from hardware engineers and OS devs we told them they were putting too many extra things inside which a lot of us in OS development: Linux and Windows had to patch a lot of timing and processing scheduling on the OS level because those chips are just built wrong. Same with AMD but they fair better at the end just because when they crank the speed inside to compensate for their additions to their chip, it didn't fall apart like the Intel chips. But that is because intel uses parallel bus technology and AMD uses a type of serial mux which the later is better at overclocking internally. Its the reason why overclocking these chips are harder because they are already overclocked.
 
I downloaded this earlier, but didn't install it.

Asrock inexplicably took it down for at least a couple of hours.

Its back now. file name and size are the same.

.....guess I will use the "new" one!
My Steel Legend got the update on Wednesday. I hadn't noticed. Thanks for the heads up.

Finally a release version. They've been sitting there with a series of betas since December. I'll have to run a few before & after game benches. I'm still on the 0x113 microcode since I generally don't install a beta BIOS unless I have to. More fps isn't a reason unless I really want to play a game that just isn't running well enough. So far that hasn't happened.
 
Im kinda hoping the new CEO makes them stay on Z890 a little longer and force new designs to work with it.
Because these ITX boards are really nice. And I wont need anything better for years ;)
 
Im kinda hoping the new CEO makes them stay on Z890 a little longer and force new designs to work with it.
Because these ITX boards are really nice. And I wont need anything better for years
Well they need to make a better chip to support ram >5200 without having to sit there and configure it for faster ram. Or what I should say, overclock the chip set so it can support the faster ram they advertise.
 
Well they need to make a better chip to support ram >5200 without having to sit there and configure it for faster ram. Or what I should say, overclock the chip set so it can support the faster ram they advertise.
Max stock DDR5 speed for Arrowlake is 6400 ;)

All CUDIMMs should be set for this, to default to 6400. Since they released with Arrowlake or after Arrowlake release.

Hopefully newly released UDIMM sets will include 6400 JDEC profiles. Because CUDIMMs are a ridiculous upcharge, right now.
 
Max stock DDR5 speed for Arrowlake is 6400 ;)

All CUDIMMs should be set for this, to default to 6400. Since they released with Arrowlake or after Arrowlake release.

Hopefully newly released UDIMM sets will include 6400 JDEC profiles. Because CUDIMMs are a ridiculous upcharge, right now.
Intel is a little different story than AMD. Since they divided the memory controller between the processor and chip set, they had to crank the ram speed up to compensate for the delay they caused in the circuit by doing this. If Intel didn't try to copy AMD who fully put the controller on chip, Intel processors would be 30-50% cooler.
 
No performance improvement in FF14 benchmark. But, I dont think that one is a scheduling or utilization issue. I think its all about latency, which Arrowlake sucks at.
 
No performance improvement in FF14 benchmark. But, I dont think that one is a scheduling or utilization issue. I think its all about latency, which Arrowlake sucks at.
If you want some help, Download and run CPU-Z and screen shot the memory and SPD tab and lets see if some alternative timings will fair better.
 
Intel is a little different story than AMD. Since they divided the memory controller between the processor and chip set, they had to crank the ram speed up to compensate for the delay they caused in the circuit by doing this. If Intel didn't try to copy AMD who fully put the controller on chip, Intel processors would be 30-50% cooler.
AMD doesn't put the memory controller on the CPU chiplets, they're on the I/O die. At least on the Ryzen desktop CPUs. A lot of the APUs are monolithic, but not all. For example, Strix Halo is 2 CPU chiplets plus an I/O die with a big fat iGPU in it and a 256-bit memory bus connected to the I/O+graphics die. EPYC and Threadripper are the same as the desktop chips. Intel Xeons are different. They're multiple tiles, but each tile has CPU cores and a memory controller. Like we have some Sapphire Rapids machines at work with 4 tiles and each tile has a 128-bit DDR5 controller. So "dual channel" but actually quad since DDR5 is 2 32-bit channels per 64-bit DIMM. Putting the memory controller on a different die can work quite nicely, Intel just didn't get it quite right on their first real try.
 
If you want some help, Download and run CPU-Z and screen shot the memory and SPD tab and lets see if some alternative timings will fair better.
I will keep that in mind in a week or so. As the timings I used for big improvements on my 7200 set with a 13600K-----barely improve latency on my 265K.

I took a chance on some Kingbank brand RAM.

DDR5 8400 48GB. It was $120 when I bought it a few days ago. Its now $135. UDIMM, not CUDIMM. Hasn't shipped yet.

https://www.amazon.com/gp/product/B0DSZPHLCX?ref=ppx_pt2_dt_b_prod_image&th=1
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
I will keep that in mind in a week or so. As the timings I used for big improvements on my 7200 set with a 13600K-----barely improve latency on my 265K.
The issue seems like with the 265K, its internally overclocked, and need to approach the timing of the ram accordingly instead of setting a timing they want. Ideally you want to time the chip so the ram has a crazy fast latency like 1-1-1-1 or 2-2-2-2 but with the multiplier cranked inside the chip, I don't think its possible.
 
Is there anything that needs to be done in bios or win11 to get a 285k tweaked for gaming at its best?

XMP on but anything else?
 
Is there anything that needs to be done in bios or win11 to get a 285k tweaked for gaming at its best?

XMP on but anything else?
BIOS with 0x114 or newer microcode. The Windows bits should be taken care of by Windows Update. If you're not running an up to date Win11 24H2 set the power profile to "performance" in Windows. Maybe do that anyway. Intel says they fixed it (or rather had MS push out a patch to fix it), but the patch changed the default "balanced" to be the same as "performance", so no point in not setting it to performance.

That's all the basic stuff I can think of. The big one is the BIOS update.
 
Is there anything that needs to be done in bios or win11 to get a 285k tweaked for gaming at its best?

XMP on but anything else?
Increasing Ring multiplier and D2D (Die to Die) multiplier.
NGU can also benefit. Although, it seems overall less beneficial.
 
Is there anything that needs to be done in bios or win11 to get a 285k tweaked for gaming at its best?
if there is bios updates do them. Memory timing is more important than a powerful processor. Windows will go through some more Kernel patches as well as Linux before its over. Even though for them they are about to incorporate a rust based subsystem to replace gnu tools.

Clocking memory is way different than overclocking the processor. Because you time it so the memory flow is efficient no matter what speed the clock wants to be at. Since the platform is already overclocked, you have to use these set of rules instead of just running at 'X' speed with 'Y' timings and focus on the goal of high frame rate. Your memory looks quite impressive you selected, I hope it times well for you. The issue becomes how well it works with the motherboard. Because the pcb traces can effect timing results. That is why you will see oddly shaped trace patterns of the PCB traces from processor to memory. That is physical time compensation. Memory, I would try to find a good vendor for. Amazon would be the last place in the world because you never know if its counterfeit, b stock, returns or rejects from someone cherry picking the best ones. Because I imagine that is going on since people are aiming at a certain speed instead of focusing on the performance. They have to hand select from many to achieve both.
Last time I had to update an old computer, I bought some DDR4 ram from someone who couldn't get it to time to 2400Mhz. I didn't care it set up at 2100 Mhz because I timed my cas to 8 instead of 20 and the difference ended up 170 FPS faster running it 8-8-8-8 1T at 2100 than it did 20-20-28-32 - 2T @2400 which is an odd timing that I found that worked at 2400, but it didn't run well compared what speed and delay it wanted to operate.
 
Does anyone know if the 18A process is on track for release in the second quarter this year? I find it interesting, they are saying TSMC's 2nm is denser while Intel's process is projected to be faster. Wasn't it always the Intel fanboys extolling the density of the Intel Processes?
 
I will keep that in mind in a week or so. As the timings I used for big improvements on my 7200 set with a 13600K-----barely improve latency on my 265K.

I took a chance on some Kingbank brand RAM.

DDR5 8400 48GB. It was $120 when I bought it a few days ago. Its now $135. UDIMM, not CUDIMM. Hasn't shipped yet.

https://www.amazon.com/gp/product/B0DSZPHLCX?ref=ppx_pt2_dt_b_prod_image&th=1
This RAM arrived today and I installed it.

It booted up without issue. Asrock Z890i Nova ITX. 265K.

DDR5 8400 CL38. 48GB. $120. Excellent deal.

With overclocked D2D, Ring, and NGU: my 32GB DDR5 7200 was 82ns latency, in Aida.

This Kingbank 8400 is 73.1ns-----without the D2D, Ring, and NGU tweaks. That's a bigger improvement than I was expecting!
 

Attachments

  • PXL_20250411_181228710~2.jpg
    PXL_20250411_181228710~2.jpg
    135.8 KB · Views: 0
  • PXL_20250411_181234892~2.jpg
    PXL_20250411_181234892~2.jpg
    124.9 KB · Views: 0
  • PXL_20250411_181252846~2.jpg
    PXL_20250411_181252846~2.jpg
    257.9 KB · Views: 0
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
This RAM arrived today and I installed it.

It booted up without issue. Asrock Z890i Nova ITX. 265K.

DDR5 8400 CL38. 48GB. $120. Excellent deal.

With overclocked D2D, Ring, and NGU: my 32GB DDR5 7200 was 82ns latency, in Aida.

This Kingbank 8400 is 73.1ns-----without the D2D, Ring, and NGU tweaks. That's a bigger improvement than I was expecting!
I got this RAM after reading about how well it worked for you. Thanks!
 
Ended up getting a Asus Z890 BTF motherboard for my upcoming build later this year. Was looking at the 285K, but if a better CPU comes out by the end of the year or if a better one is due in a few more months, I'll probably look into getting that. Hopefully the new stuff in 2026 will use the same 1851 sockets but can always get another motherboard if needed. (And if those CPUs are worth the upgrade)
I know the AMD CPUs are better at gaming, but as long as I can play games and do productivity stuff well, I'll be happy.
(Probably will take a good while to get a 5090 anyhow at a decent price since I won't pay anything other than MSRP and that means no new GPU till next year most likely unless i get VERY lucky or if they start to make a ton of them in the coming months) :banghead:
 
Last edited:
Think they'll have the 300 refreshes in that socket, but will go to the next standard if it's a good enough upgrade. (Not too hard to replace the motherboard and CPU if need be)
 
I wonder if the Intel core 385K is going to drop right into these existing motherboards? That's a question that's been on my mind since the release of the 285k LOL
 
Think the 300 series should be able to, but not the 400 series from what I've read and from what Vegeta mentioned. (But that's not set in stone though)
The 285K is lackluster somewhat in some respects but who knows about the future sockets? 300 and 400 series could turn out great or continue the 200 trend.
 
This RAM arrived today and I installed it.

It booted up without issue. Asrock Z890i Nova ITX. 265K.

DDR5 8400 CL38. 48GB. $120. Excellent deal.

With overclocked D2D, Ring, and NGU: my 32GB DDR5 7200 was 82ns latency, in Aida.

This Kingbank 8400 is 73.1ns-----without the D2D, Ring, and NGU tweaks. That's a bigger improvement than I was expecting!
NGU 35 | D2D 35 | Cache ratio 41 (no voltage increases required on my CPU)

Latency dropped to 71.1

And to be clear, changing those things usually only results in a small latency reduction. However, the improvement in some games can be decent. So it seems that pure latency reduction is not the only thing which Arrowlake benefits from.
 
I tested Elden Ring with the new 8400 RAM + the other tweaks (D2D, cache, NGU) standing in my usual spot, staring at a wall.

~176fps. That's up from ~ 148fps with my DDR5 7200 set and no other tweaks. 156 with the tweaks.

and 1% low went up from 138 to 157

My previous 7800X3D optimized with higher clocks and less heat via PBO + CO and that same set of 7200 heavily tweaked: was ~ 224 fps in Elden Ring

All with the same 4070 TI Super

Pretty good gains, IMO.

Addtionally: the two best Pcores on a 265k boost to 5.5. and all E-cores are 4.6. With all Pcores locked to 5.5 and Ecores to 4.9------Elden Ring has minor gains. Staring at the wall, it was about 2fps better.
 
Last edited:
Monster Hunter Wilds isn't a problem for Arrowlake. But my benchmark average framerate went up 7fps. 1440p DLSS Quality. all settings maxed. Would probably see more gains with a bigger GPU. I was surprised there were any gains, as the game is GPU heavy.


FF14 benchmark is virtually unchanged. No benefit.
 
Back
Top