Discussion Alder Lake - Builders Thread

Page 10 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

AdamK47

Lifer
Oct 9, 1999
15,206
2,838
126
This thread is for those of us that own or are looking to get an Alder Lake CPU.

Bought mine on release day. 12900K. Memory selection was limited to only Crucial DDR5 4800 at the Micro Center I went to. Bought two 2 x 8GB kits. Motherboard is a Gigabyte Z690 Aorus Master. I did buy a 360mm Lian Li Galahad 360mm AIO. Ended up returning it since the Lian Li Socket 1700 backplate that I bought separately worked well with my NZXT Kraken X73.

46RRrSN.jpg
 
Last edited:

Justinbaileyman

Golden Member
Aug 17, 2013
1,980
249
106
Well, yes, gigabyte screwed me (on an Intel motherboard circa 2007) on an RMA, plus its $100 more than the one I mentioned. Also, this would be for DC work, and I could care less about type-c ports. I did spend the extra for a PCIE4 NVME drive, but I will be using the on-board video. The point of this box is to evaluate performance of the CPU, and how well it works vs 5900x in DC jobs. As far as newegg screwing me, I have never had a problem with an open box, but there is always a first. If they do, its the last I will ever buy from them. (like gigabyte and creative labs cheating on a rebate, so 2 companies that will never get a dime more from me) I have spent over $100,000 at newegg, and now Amazon is telling me I qualify for business pricing, due to how much I spend there.

Anyway, I am looking for other good mobos, but I went with Z690 for the best chipset and upgrade-ability to Raptorlake.

Edit: I saw the DDR5 speed mentioned. I have some DDR4 4000, I might take that out from one of my 5950x's and try that, its only running at 3200 now to match the other memory in the box.

Edit 2. Most of my 5950x's are in sub $200 motherboards, but X570. So this would be an equal test as well. And my 3900x and 5900x.
I don't hate Newegg or anything I just refuse to buy open box from them any more after those particular items I purchased that they purposely screwed me on. I try to only purchase on there competitor sites unless Newegg has an unbeatable price on something.

Anyways back on topic here.. My vote is for the Asus Prime Z690-P Wifi 6e.. Its cheap and pretty and has all the bells and whistles at that range!!

Edit: Plus it is a DDR4 Board like you are wanting, Mark.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
I don't hate Newegg or anything I just refuse to buy open box from them any more after those particular items I purchased that they purposely screwed me on. I try to only purchase on there competitor sites unless Newegg has an unbeatable price on something.

Anyways back on topic here.. My vote is for the Asus Prime Z690-P Wifi 6e.. Its cheap and pretty and has all the bells and whistles at that range!!

Edit: Plus it is a DDR4 Board like you are wanting, Mark.
I looked at that. Its a clone of the one I mentioned (the M) but no NVME heatsink, but I will be using a better one that I have on the SN770 WD NVME drive. The VRM heatsinks look the same, and the rest of the board. But $50 less.
 
  • Like
Reactions: Drazick

Justinbaileyman

Golden Member
Aug 17, 2013
1,980
249
106
Max without memory overclock is DDR5-4800. Why fill all four slots? Going with 128GB?
I wanted to go 4 x 16gb DDR5 because right now on my 3950x X570 rig I am using 4 x 8gb 3200 and I am close to maxing it out every day with all my tabs open and all my programs running in the back ground. I obviously need more and faster Ram. I use Handbrake and WinFF alot.
Should I be shooting for 128gb Instead??
 

Tech Junky

Diamond Member
Jan 27, 2022
3,407
1,142
106
Question on socket 1200/1700 mounting compatability with older sockets. I have a prolima Megahalems cpu cooler that fits Intel Socket 775, 1156, 1366 and 2011.

Any chance that will work here ? the motherboard says it has holes for socket 1200 and 1700.
Probably not.

Better off getting something with the right bracket or a bracket upgrade kit for some of the coolers out there for prior gen chips.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
Probably not.

Better off getting something with the right bracket or a bracket upgrade kit for some of the coolers out there for prior gen chips.
First, I went with the Prime-P 690 since it had 14 stage vrm instead of the M one had 10. I got the WD black SN770 PCIE 4.0 250 gig NVME.

As far as the heatsink, I have 2 of these sitting here doing nothing, thats a lot of $$$ that I really did not want to waste. I mean $180 wasted. Look at this compatability !
 
  • Like
Reactions: Drazick

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
So total of $577, I have a 12700k, which has a heatsink (crummy factory one) and a good motherboard with 14 phase VRM and a PCIE 4.0 NVME drive that has the heatsink on the motherboard. I have case PSU and memory laying around. This is for 20 threads. Not bad. We will see how it performs when I get it.

Stupid question. By using the IGP, will any CPU intensive tasks suffer due to more heat ?

I only have one complaint about this build. It appears that the most compatible heatsink for Intel is not compatible with this. AMD heatsink's seem to work from AM2 all the way to AM4, most without even a kit. So I may have to spend another $50-100 more to get a good heatsink, just due to their incompatibility.

Remember, I had nothing but Intel from 2006 to 2017, and these are from that era. I go back to Intel, and get screwed.
 
  • Like
Reactions: Drazick

Tech Junky

Diamond Member
Jan 27, 2022
3,407
1,142
106
Megahalems Rev.C | Prolimatech
I'm not seeing any adapter for that.

I went cheap with a $45 cooler and graphite pad and keep within 5F on idle temps and under load hit ~130F / 54C. I'm also not using a GPU in my build as it's mainly a server and the only time I need a display is when it's not working right. Usually to roll back a kernel upgrade that's not 100% working upon reboot.

I was thinking about reusing a Fuma cooler but, the co. wanted $8 for the adapter unlike some other options that just required an e-mail to request one w/ PoP of the original. I changed my remodel to a rebuild from the ground up and sold the prior system. The only hesitation I would have on an adapter kit / old cooler being used is the surface covering the larger LGA1700 footprint and where the heat pipes land on the die.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
I'm not seeing any adapter for that.

I went cheap with a $45 cooler and graphite pad and keep within 5F on idle temps and under load hit ~130F / 54C. I'm also not using a GPU in my build as it's mainly a server and the only time I need a display is when it's not working right. Usually to roll back a kernel upgrade that's not 100% working upon reboot.

I was thinking about reusing a Fuma cooler but, the co. wanted $8 for the adapter unlike some other options that just required an e-mail to request one w/ PoP of the original. I changed my remodel to a rebuild from the ground up and sold the prior system. The only hesitation I would have on an adapter kit / old cooler being used is the surface covering the larger LGA1700 footprint and where the heat pipes land on the die.
That looks like exactly what I will do, after it comes, and mine don't work. It appears that this company is out of business, no recent updates or products.

Stay tuned....
 
  • Like
Reactions: Drazick
Jul 27, 2020
16,158
10,236
106
I wanted to go 4 x 16gb DDR5 because right now on my 3950x X570 rig I am using 4 x 8gb 3200 and I am close to maxing it out every day with all my tabs open and all my programs running in the back ground. I obviously need more and faster Ram. I use Handbrake and WinFF alot.
Should I be shooting for 128gb Instead??
Browser tabs? I hate them coz they keep eating up my 32GB on my Thinkpad. Better start with 2x32GB and see how that goes.

 
Last edited:

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
I would not bother touching GPU clocks at all, they are irrelevant for CPU intensive tasks, a few watts at idle.
The thing that is most important to get right in ADL build is memory. Stock 3200 with whatever timings is holding CPU back, need to go to 3600C16 with Gear1 mode region to unlock perf. It's like Ryzen1, where B-Die mem ~3400 worked wonders for performance.
 
  • Like
Reactions: igor_kavinski

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
Isn't the iGPU going to be effectively power gated when using a dGPU, except maybe when doing extremely niche GPGPU calculations?
I will comment since I am the one who started asking, but in my case, the IGPU will only be used to drive a monitor, nothing else. No 2D or 3D anything, just like a terminal. Might see a few web pages, thats as graphic as I am going to use it for. But the CPU's big and small will be at 100% load for use 24/7.
 
  • Like
Reactions: Drazick

DrMrLordX

Lifer
Apr 27, 2000
21,620
10,829
136
I will comment since I am the one who started asking, but in my case, the IGPU will only be used to drive a monitor, nothing else. No 2D or 3D anything, just like a terminal. Might see a few web pages, thats as graphic as I am going to use it for. But the CPU's big and small will be at 100% load for use 24/7.

Oh okay. There might be a tiny amount of overhead from driver calls, but basic desktop rendering won't use much power.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
Do you use a watt-meter? How are you going to determine power efficiency between Alder Lake and Zen 3 for your use case?
Yes, I have 3 kill-a-watt meters. I will compare the WCG/Rosetta production against the price of the CPU's and used wattage(at the wall, all have comparable gold rated PSU's, almost al EVGA,). I can't find it now, but Stefan in the DC forum did some comparing. EPYC won for wattage/points, but 5950x won when you consider the hardware costs. I will be comparing the 12700k against the 5900x and the 5950x. Core count and price compare to the 5900x. So I am curious how this will work out in that equation. To give you an idea, a 400 watt 3080TI in a 5950x is taking 570 watts at the wall, with all 32 threads fully loaded, and the GPU at 98% load.

Note to others... My use case is not exactly typical, but representative of highly threaded, high use cases, possibly like rendering or engineering apps. My comparisons are for my use case, but it will apply to some extent to any high load use case. For one, gaming, which Alder lake excels at, is NOT a high use case. single threaded performance is key, even though some games are fairly highly threaded, not to the extent I use. Power consumption during gaming is much lower than peak.
 
Last edited:
  • Like
Reactions: Drazick
Jul 27, 2020
16,158
10,236
106
To give you an idea, a 400 watt 3080TI in a 5950x is taking 570 watts at the wall, with all 32 threads fully loaded, and the GPU at 98% load.
Should be very interesting. How long does it take to get results? I hope you will get results for the standard DDR4-3200 speed as well as lowest CAS latency and highest memory speed, like so:

12700K - Stock DDR4-3200
12700K - Lowest possible CAS latency (any speed, even if it is lower than DDR4-3200)
12700K - Highest overclocked speed (no matter how high the CAS latency has to be)

and all three data points for the 5900X and 5950X also.

Is this possible?
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
Should be very interesting. How long does it take to get results? I hope you will get results for the standard DDR4-3200 speed as well as lowest CAS latency and highest memory speed, like so:

12700K - Stock DDR4-3200
12700K - Lowest possible CAS latency (any speed, even if it is lower than DDR4-3200)
12700K - Highest overclocked speed (no matter how high the CAS latency has to be)

and all three data points for the 5900X and 5950X also.

Is this possible?
Not really. All the 3200 I have is c14, and many of the boxes have 3600 cl 16 but virtually none run at that speed, like 3200 c14 speed. And since work units are not exactly all the same this is a test that will be more like "are they very different ?" or "margin of error?", so like 5% maybe even 7% is margin of error. I have a 14 core Xeon that run at 2.5 ghz, that gets beat 2 to one by a 12 core 3900x running at 3.6 to 4 ghz or so. Thats very different.
 
  • Like
Reactions: Drazick
Jul 27, 2020
16,158
10,236
106
And since work units are not exactly all the same this is a test that will be more like "are they very different ?" or "margin of error?", so like 5% maybe even 7% is margin of error.
Since all the threads are fully loaded, is it possible to quantify the work done in terms of work units accomplished per hour or something like that? Is it possible for a single unit to take more than an hour?