Intel 12th-Gen CPUs

[RasmusP]

Is it Alder Lake specific?

It's a new thing since rocket lake.
I had a quick read:
Gear 2 = half bandwidth, same latency

Apparently you can overclock this too though. So you can try to run 3600 mhz in gear 1. But you might need pretty bad timings to get it stable.
At some point using lower timings and using gear 2 will yield better results, depending on the load you have.

So it's mhz vs timings all over again but with another switch you throw into the mix.

I didn't fully read through this topic yet but somehow most reviewers seems to get stuck at 3200 mhz when using gear 1.
But depending on the sticks they can use very low timings.

 

[James Skinner]

Been putting a new i7-12700k through it's paces lately - although I've only realIy tested a few sims in VR I have to say I'm impressed - it's running very cool with low utilisation with 56 cars on lemans in AC in VR, and barely breaking a sweat in other sims too - even ACC which I gather is quite cpu heavy. Need to spend a bit more time with it to get the full story, but so far so good... in the UK they seem to be readily available at MRSP...

EDIT - as it's referenced above, this is running DDR4 AT 3600mhz...

 

[RCHeliguy]

It's a new thing since rocket lake.
I had a quick read:
Gear 2 = half bandwidth, same latency

Apparently you can overclock this too though. So you can try to run 3600 mhz in gear 1. But you might need pretty bad timings to get it stable.
At some point using lower timings and using gear 2 will yield better results, depending on the load you have.

So it's mhz vs timings all over again but with another switch you throw into the mix.

I didn't fully read through this topic yet but somehow most reviewers seems to get stuck at 3200 mhz when using gear 1.
But depending on the sticks they can use very low timings.

I'm hoping that the Gigabyte boards will have preset memory profiles and recognize my memory like it did for the 3200MHz memory in my current system.

I'm not into pushing overclocking.

 

[Durge Driven]

Just looking now 12600k + motherboard + 2 sticks of DDR5 is at minimum $1000. (while reusing current psu / gpu / and everything else.)

Comparing that to

5900x + motherboard + 2 sticks of DDR4 total $900

Using 5900X is not really fair to AMD build

You only need the 5800X and where I live that's another $170 cheaper

https://www.pccasegear.com/category/187_2138/cpus/amd-am4-5000

 

[Deleted member 197115]

I'm hoping that the Gigabyte boards will have preset memory profiles and recognize my memory like it did for the 3200MHz memory in my current system.

I'm not into pushing overclocking.

Gigabyte is like the worst brand for anything under the sun, why not Asus?

 

[RCHeliguy]

Gigabyte is like the worst brand for anything under the sun, why not Asus?

There is no good answer to a question like that.

Why not Chevy? You actually like that Found On Road Dead ?

 

[Deleted member 197115]

There is no good answer to a question like that.

Why not Chevy? You actually like that Found On Road Dead ?

Who drives either brand these days? Are they still making anything?

 

[RasmusP]

Gigabyte is like the worst brand for anything under the sun, why not Asus?

Well over here, the gigabyte boards around 200€ have a closed slotcover, 8x USB, good VRMs, good amount of extension slots and fan headers.

The Asus board with similar features would be the prime A, which costs 300€.
Cheapest one costs 250€ and lacks quite a lot.

The msi boards around 200€ don't even have a slotcover..

Never had an issue with anything from gigabyte. 3 friends have a mobo, 6600k, 8600k, 9600k. And I have a z490 gaming-X for my 10600k.

The gtx 970 windforce 3x of a friend still goes strong and the 2560x1440, 160 Hz IPS Monitor of another friend is superb too.

Meanwhile I know about 4 broken Asus mobos and their IPS monitors have very bad backlight bleed and IPS glow.

Of course Asus build good boards (and other stuff), but they are quite expensive and my subjective experience isn't great.



(my GPU is the best cooling design I've ever heard though. Cool and quiet! And I mean like... I don't hear it against my be quiet silent wings case fans at 900 rpm)

 

[Deleted member 197115]

My personal list of brands I always use in my builds and consider high, headache free quality:
- GPU: EVGA
- MB: ASUS ROG line
- PSU: Corsair, 80+ Platinum or higher
- DRAM: G.Skill
- CPU: Intel
- AIO Cooling: Corsair
- SSD Storage: Samsung

Brands to avoid:
- MSI
- Gigabyte
- AMD

As always, personal experience, YMMV.

 

[Trebormoore84]

Oh god you sound like a PC master race fanatic. Does my head in seriously. What a load of old nonsense.

+1 Gigabyte / Aorus boards are excellent. ASUS are marginally better but the costs simply aren’t worth it. To me it they just looks like pure PC master race marketing bollox. I have one ROG board it boosts very slightly higher than my Aorus board but who cares. 50 quid more expensive not worth it.

I have a EVGA gpu, it brick itself trying to apply a firmware via X1. Absolute disaster IMO and again rely on PC master race nonsense. yes having the 3 year warranty is great but Jesus what a waste of time. Wouldn’t say it’s the best company at all.
Still waiting for the thing to be warrentied after 3 weeks now. Used it for like 20 minutes.

If I were you I’d avoid making lists of brands it won’t do you any good, things change all the time.

 

[RCHeliguy]

Who drives either brand these days? Are they still making anything?

Your Speed Racer avatar needs a small troll its shoulder.

I have an i7-2600K that's been running 24x7 on an ASRock MB for about a decade now. I've never considered ASRock a premium brand.
So YMMV on lots of these parts.

Computers are disposable appliances. I don't care enough to argue about them. Get one and use it!

Rasmus recommended a MB, I consider him knowledgeable on this subject, and I purchased it. Done.

I look at replacing a computer MB as a hassle. All I think about, is that I'll need to find all my key's to prove I have a legal copy of Windows and reregister a bunch software on my dev box and I'll need to do the same and reconfigure my audio cards in SimHub on my gaming computer, etc.. etc..

The peripherals are the fun part.

The irony is that it looks like I could have all my parts here by Sunday, but I may not have time to mess with this until after the Thanksgiving break.

 

[RCHeliguy]

This video shows a lot including differences in how Intel and AMD handles memory speeds.

In some cases the Intel saw a drop with 3600 and a bump up with 4000 and AMD saw a bump up with 3600 and a drop at 4000.

However the changes overall are pretty small 2-3%

This shows an example where 3200 outperformed 4000. In many cases the 4000 increase was very minor or identical to 3200.

 

[Deleted member 197115]

From JayzTwoCents video you posted. Quite a gain at 1080p, but it quickly diminishes as you go up in resolution and shift from CPU bound to GPU bound scenario. BUT, in VR and in most if not all sim titles we are usually CPU bound so we should still see some gain even at higher res. Of course all that Gear 1/Gear 2 thing that Rasmus mentioned may throw it all away.
I have seen some boost on Coffee Lake going from 3200 to 4000mhz.

 

[RobertR1]

XMP timings vary all over the place and from to kit and board to board. Just looking at the first 4 timings and frequency is almost pointless. Unfortunately, ram is difficult to learn and tune properly. It's like the "car setup" of the PC. XMP is just a preset but there's a lot to tune beyond that.

 

[RasmusP]

This video shows a lot including differences in how Intel and AMD handles memory speeds.

In some cases the Intel saw a drop with 3600 and a bump up with 4000 and AMD saw a bump up with 3600 and a drop at 4000.

However the changes overall are pretty small 2-3%

View attachment 517470

This shows an example where 3200 outperformed 4000. In many cases the 4000 increase was very minor or identical to 3200.

View attachment 517471

So about these videos:
The TBG video is with the 9900k, which doesn't have Gear 1 or 2. It has one memory controller "sync pattern" and that's it. So you only have MHz vs Timings, but higher MHz with tighter timings will always give better results.
The 3600X on the other hand has the normal Ryzen "Infinity Fabric", which works basically like Gear 1 and 2. It's comparable at least.
For Ryzen 3000 and 5000, the maximum for 1:1 seems to be around 3600 MHz, some CPUs can run 3833 MHz RAM though.

That means in the Dirt Rally chart from TBG, the Intel actually "runs faster" with each step, while the 3600X drops into 2:1 with the 4000 MHz ram and therefore drops in gaming fps. Almost all applications actually lose performance when not running 1:1 on AMDs.

With Rocket Lake, running Gear 1 is mostly the way to go, similar to Ryzen. But with DDR5 and the massive MHz boost, things change for Alder Lake.
You can gain performance with DDR5 for applications that can make use of the bandwidth but when you have latency critical applications like most games at high fps, DDR4 with way lower timings is better.

So it's very important to put Gear 1, Gear 2 and the Inifinity Fabric ratio next to any charts from now on!
Jay2Cents video sadly doesn't say which CPU they used. But since the videos are from before Rocket lake, I guess it's an Intel 9th or 10th gen, so it doesn't matter.

I have seen some boost on Coffee Lake going from 3200 to 4000mhz.

Yep, like I wrote above, that's normal and to be expected, since Coffee Lake only had "Gear 1".

Of course all that Gear 1/Gear 2 thing that Rasmus mentioned may throw it all away.

Hehe yeah it does.. Things were getting too easy with overclocking Intel CPUs. They had to come up with something fancy to play around with

 

[RCHeliguy]

Hehe yeah it does.. Things were getting too easy with overclocking Intel CPUs. They had to come up with something fancy to play around with

I'm expecting the MB to have a preset for my DDR4 memory. It will recognize it as 4000MHz and try to assign something it considers appropriate. It doesn't seem like the difference will be that large in either direction, so I'm not sure I should care.

Everything arrives tomorrow except the CPU which arrives Thursday. I'm a bit surprised at how quickly it's getting here.

I'll make a backup of my boot drive on the gaming system and keep my i9-9900K on ice. If there are any hiccups, I'll put it back until after Thanksgiving break is over.

 

[RasmusP]

I'm expecting the MB to have a preset for my DDR4 memory. It will recognize it as 4000MHz and try to assign something it considers appropriate. It doesn't seem like the difference will be that large in either direction, so I'm not sure I should care.

Everything arrives tomorrow except the CPU which arrives Friday. I'm a bit surprised at how quickly it's getting here.

I'll make a backup of my boot drive on the gaming system and keep my i9-9900K on ice. If there are any hiccups, I'll put it back until after Thanksgiving break is over.

Sounds good!
I'm not sure the xmp profiles include gear 1/gear 2 so I would check that. Gear 1 seems to be way better for cpu bound games.
4000 MHz might not be stable though.
Most high MHz sticks come with some lower xmp profiles too, which also have lower timings.
So maybe you could run gear 1 with 3600 MHz and great timings.

Since you can't use the extremely high MHz of ddr5 but nobody is doing Alder lake DDR4 gear 1 vs 2 comparisons, you can look at rocket lake gear 1 vs 2 comparisons between DDR4 sticks:

DR = dual ranked. But I'm out of the loop regarding that.. It has something to do with how the modules are soldered together (not of there are modules on both sides! That's only an indicator of probability of DR!).

 

[RCHeliguy]

At some point I'm just going to be happy with the overall higher performance of the i9-12900K over my i9-9900K and call it a day.

This shows an overall improvement of just 23% ... Hmmm...

This shows a bunch of faster core speeds offset by a lower memory latency.

UserBenchmark: Intel Core i9-12900K vs i9-9900K

 

[RasmusP]

At some point I'm just going to be happy with the overall higher performance of the i9-12900K over my i9-9900K and call it a day.

This shows an overall improvement of just 23% ... Hmmm...

This shows a bunch of faster core speeds offset by a lower memory latency.

View attachment 517763

UserBenchmark: Intel Core i9-12900K vs i9-9900K

I'm not telling you to manually test and tighten the timings. I'm just saying it might be worth it for VR to test Gear 1 with 3200-3600 MHz xmp profiles (depending on what profiles your sticks come with) vs the gear 2 + 4000 MHz profile.

It's only changing 2 lines in the bios.

Userbenchnark: it's sketchy.. Almost every known tech YouTube did a video or post about them. It's one other, sometimes nice indicator but by no means anything really trustworthy.

 

[RobertR1]

Gear 2 and ddr4 is a bad idea but each his own.