INTEL CPUs *OC Benchmark & Discussion*

[congster2]

Really surprised not many posts here so far. Curious, for those using Alder lake, any issues with virtualisation so far?

 

[DaNuker]

I'm thinking of pairing the 12700 with the H610 since this is the only 65W CPU with E cores and I don't intend to overclock anything or any GPU. However, I don't see this in the price list. Only the F variant is being sold here. Hope they will bring it in.

I am also looking at 12700 as a stopgap measure.

Do note that 12700 is severely multicore limited on 65W PL on a motherboard that doesn't let you unlock it. I noticed this slide on MSI's B660 launch trailer from youtube:

 

[DaNuker]

12400 non-K running at 5.2 Ghz all core.

Looks like the whole bios "bug" where you could tweak non-K CPUs is back again on this Asus DDR5 Apex and Hero motherboards (0811 bios version). 12400 non-K overclocked via BCLK like Intel OC days before Sandy Bridge.

I expect Intel to force Asus to change this quickly though just like they tried to remove AVX-512.

 

[elmariachi]

12400 non-K running at 5.2 Ghz all core.

Looks like the whole bios "bug" where you could tweak non-K CPUs is back again on this Asus DDR5 Apex and Hero motherboards (0811 bios version). 12400 non-K overclocked via BCLK like Intel OC days before Sandy Bridge.

I expect Intel to force Asus to change this quickly though just like they tried to remove AVX-512.

Then again how many people would run a 12400 on an Apex board? The odds are extremely slim.
Even his SP rating at 158 is a bug. Look at the voltages, it shows N/A. This means that the V/F points in the BIOS is not actually loaded properly. Looks like 0811 has not updated the microcode for the non-K variants.

 

[stanlawj]

https://www.tomshardware.com/news/easy-mod-reduces-alder-lake-cpu-temperatures-5-degrees-celsius
The transition from the LGA1200 socket to the LGA1700 socket saw a 42% increase in pin count, resulting in a more extended socket. Consequently, Alder Lake chips are larger and more rectangular than Rocket Lake, which is smaller and has a more squared shape. The problem falls on the Independent Loading Mechanism (ILM). The ILM still latches to Alder Lake processors at the same pressure points as Rocket Lake. The result is that the mechanism pushes down on the processor in the center, causing it to sit lower than on the edges. It's the reason why Alder Lake's integrated heat spreader (IHS) eventually takes a concave shape after several hours of operation

Wow, thanks.... it is true.
Asrock Z690 Extreme CPU holder puts a lot of tension on the CPU IHS. I remember it was like, heck why so hard to lock it. Now with the washers, it is much more reasonable force comparable to my previous Haswell CPUs.
I checked the CPU heat spreader and now it shows a very slight concave (like 0.1mm).
After adding washers, it didn't change the max CPU temps.

Probably will only affect extreme overclocking or watercooling.

 

[elmariachi]

Wow, thanks.... it is true.
Asrock Z690 Extreme CPU holder puts a lot of tension on the IHS.
I checked the CPU heat spreader and now it shows a very slight concave (like 0.1mm).
After adding washers, it didn't change the max CPU temps.

Probably will only affect extreme overclocking.

The CPU holder mechanism is indeed stiff even on my Z690 Apex. People that have added washers to some of these extreme boards have had their memory clocks not running stable or properly after that. I think the pressure exists for a reason or clearance for that matter. The one thing that somewhat worked is loosening the torx screws on the socket. But that's like maybe a 3 degrees difference at max though I don't feel the need to try it with my current temps.

I'm like mid to high 60s on my 5.5Ghz OCTVB profile with all cores around 5.1/5.2Ghz. Disabled E-cores. Will probably enable them later but ring bus at 4.7Ghz is quick with 6000C32 1T. Very noticeable difference in games which is all I do anyways.

 

[stanlawj]

The CPU holder mechanism is indeed stiff even on my Z690 Apex. People that have added washers to some of these extreme boards have had their memory clocks not running stable or properly after that. I think the pressure exists for a reason or clearance for that matter. The one thing that somewhat worked is loosening the torx screws on the socket. But that's like maybe a 3 degrees difference at max though I don't feel the need to try it with my current temps.

I'm like mid to high 60s on my 5.5Ghz OCTVB profile with all cores around 5.1/5.2Ghz. Disabled E-cores. Will probably enable them later but ring bus at 4.7Ghz is quick with 6000C32 1T. Very noticeable difference in games which is all I do anyways.

Nope, the pressure need not be alot. Most likely those who had problems is because their CPU PCB already bent too much, and trying to fix the problem with washer will simply not work because the edge of CPU is lifted off the pins.

As long as the the CPU fully contacts and press on all pins in the socket, it cannot go down any further because the plastic socket walls is preventing the CPU from going deeper into the socket.

The most important is the socket itself and the CPU PCB must not be bent for the washer mod to still work. Therefore it must be done at the earliest moment possible. Before 1st installation is the best.

I believe the IHS continues to bend more and more over weeks and eventually will bend the CPU PCB and socket.

 

[elmariachi]

Nope, the pressure need not be alot. Most likely those who had problems is because their CPU PCB already bent too much, and trying to fix the problem with washer will simply not work because the edge of CPU is lifted off the pins.

As long as the the CPU fully contacts and press on the pins, it cannot go down any further because the plastic socket walls is preventing the CPU from going deeper into the socket.

I'm not saying the pressure needs to be alot or is responsible for the CPU bending. It is definitely stiffer than LGA1151/1200 sockets I've worked with. The main issue is that some of the stock installed board backplates are not levelled correctly or not rigid enough so when the standoffs are screwed in there is unequal pressure on the CPU causing 1 side to lift. This is further worsened if you're using huge aircoolers like the Noctua.

https://www.overclock.net/threads/o...imus-z690-owners-thread.1794802/post-28930039

 

[matique]

Washer mod done on my MSI Unify ITX. Improved temps by about 3C overall, quite happy.

Did another comparison between the Optimus Sig V2 and EKWB Velocity Full Nickel. Even with washer mod, optimus still edges out of the velocity2.

Optimus Sig V2:

Velocity 2:

It's about 3C diff between the two. Not much, but still quantifiable enough.

Also on full load 315w load with optimus block, it does not thermal throttle.

All in all pretty satisfied with the setup. I have also finalised my ram tuning with the unify, ended up with 6666c36 2T. One can only push samsung so far, but I am quite pleased.

 

[DaNuker]

Then again how many people would run a 12400 on an Apex board? The odds are extremely slim.
Even his SP rating at 158 is a bug. Look at the voltages, it shows N/A. This means that the V/F points in the BIOS is not actually loaded properly. Looks like 0811 has not updated the microcode for the non-K variants.

He's following up on 2 possible B550 boards that have an external clock gen that *might* support this. I agree nobody is going to run a Z690 with really expensive DDR5 MB just to do this and it might be too buggy for stable daily use.

But seeing a Celeron G6900 Ghz at 5.3 Ghz is hilarious just like the ye olde Celeron 300A.

 

[GoldenStars]

any cheap b660 motherboard available at the moment? most that I see seems quite expensive

 

[xonix]

Gaming Performance: i3 vs i5, i7, i9 Intel 12th-gen​

 

[matique]

Latest benches:

12900k 5.3p/4.1e/4.2e
3090 FE
6666c36 Gskill Z5 Trident [Samsung IC 5600c36 base]

Timespy

Time Spy Extreme

Should be my final tune for this rig pretty sure it is the fastest 12900k in Singapore by virtue that there isn't much competition lol. Hopefully ddr5 prices come down soon and makes this platform more affordable for others to join in and have fun with.

edit: some R23 benches. I can't run 5.3 on r23 due to thermal issues with my particular bin. Cinebench has an avx load which pushes my 5.3 1.38v tune to above 100c, so here's one at 5.2 all core:

The MSI z690 itx unify has very high auto voltages for per core pbo (can't use static), but my cooling is still enough to handle it at 5.1 MT and 5.4 ST.

 

[TanKianW]

Wah.......RTX3090FE is really no joke.

A card which I really want to get for building a GPU server cluster.

 

[Kairos]

...

Should be my final tune for this rig pretty sure it is the fastest 12900k in Singapore by virtue that there isn't much competition lol. Hopefully ddr5 prices come down soon and makes this platform more affordable for others to join in and have fun with.

...

G.Skill Trident Z5 RGB 2x16GB DDR5 6400 @ S$935 on Newegg SG!

 

[elmariachi]

G.Skill Trident Z5 RGB 2x16GB DDR5 6400 @ S$935 on Newegg SG!

Yeah most of the 6400 kits are around that price or slightly less. From what I understand, these 6400 kits current batches are G.Skill's last Hynix batches and they would be going full on Samsung after this. I've yet to receive mine direct from Taiwan due to CNY holidays.

 

[matique]

G.Skill Trident Z5 RGB 2x16GB DDR5 6400 @ S$935 on Newegg SG!

Yah man even I can't justify. My 5600c36 kit was bought under $500. Even that was hard to justify heng it performed well.

 

[zjllee]

I am coming from a i7-2600k what would be a logical upgrade if I want the enthusiast style overclockability and raw multi processing power for scientific and encoding work? Some light gaming on the side like city skylines and strategy games.

I am asking here as the build thread is dead...very sad.

TIA

 

[elmariachi]

I am coming from a i7-2600k what would be a logical upgrade if I want the enthusiast style overclockability and raw multi processing power for scientific and encoding work? Some light gaming on the side like city skylines and strategy games.

I am asking here as the build thread is dead...very sad.

TIA

I would go for a 12600k + DDR4 board. If you have the budget, go for 12700k + DDR4 board. Feel that 12700k would be better suited for your requirements as well as being abit more future proof with the additional cores.

 

[matique]

I am coming from a i7-2600k what would be a logical upgrade if I want the enthusiast style overclockability and raw multi processing power for scientific and encoding work? Some light gaming on the side like city skylines and strategy games.

I am asking here as the build thread is dead...very sad.

TIA

For high multi thread work I'd still recommend amd though, like the 5950x. Far easier to cool with full at full whack for scientific and encoding work. For reference for me to beat a 5950x I need to hit 5.2 all core and this draws roughly 300w. Also I'm assuming you need a lot of ram for that kind of work, so ddr4 still the way to go. New zen4 CPUs on the horizon though...