With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.
I think it is a number of reasons, the biggest being bad press.Def looks like Alder Lake is a bust for DIY. Rocket Lake might be outselling it. I think it's the board prices and perhaps the perception you need DDR5.
Based on what? Tea Leaves? It hasn't even been on sale for 10 days yet.
Forget clock speed. What I want is a 12900k forced to 142 watt max under any load. THEN benchmark it vs a 5950x stock at the same 142 watt, using a very large and intensive benchmark suite. And log running benchmarks, like 20 minutes to an hour (to see heat and hitting the wall on temps, etc) Also, both should have the best memory speed and timings available for their platform (within reason. For example, my 5950x's run 4000@3800/1900 speed. cl16) Not sure the most reasonable fast DDR5 speed is today, but the memory should not be the most expensive, but definitely upper tier. Based on these results, I could better judge Alder lake. Even the above is only 95 watt. for 6/4 cores I would expect efficient to be about 65 watt, maybe a little more.What are the actual sales numbers for ADL to date? I'm hoping they are low as that will keep Intel on it's toes. ADL ticked all the right boxes for me and I'm in the process of building a system around the 12700K I got at Microcenter for $400 last Sunday. Mobo should be here in a few days. I'm still rocking the 4770K so it's gonna be a big upgrade for me.
Zen 3 was tempting but for my uses I find the iGPU strong enough and more stable. Rocket Lake didn't do it for me. With ADL I see big cores that are more performant than Zen 3. Most of my compute intensive apps don't use more than 8 cores effectively anyway. I'll probably lock it down to 150W and be done with it.
Do we have an apples-to-apples comparison of Zen 3 to Alder Lake? By that I mean 8 core vs. 8 core at the same frequency doing the same application? I'd love to see that on CB or something at 4GHz. In addition it would be interesting to see them clocked so that the performance is the same and then see the power. But keep the lower performing part at 4GHz and clock the higher IPC one down to match it performance-wise. This way neither get out of the linear part of the shmoo plot.
Take a second look at the 12700 and 12600 skews. Those are much more sane and arguably a superior product.Heat and power usage are really the only things I have against Alder lake.
Hm slap, no E Cores on i5 12600.
Dell leaks specifications of Intel 12th Gen Core non-K Series - VideoCardz.com
The full lineup of Intel 12th Gen Core series leaks Earlier this month Intel launched six Alder Lake-S, desktop processors. However, the plans for this architecture go a lot further. Intel is yet to announce its 65W models called non-K Core series as well as 35W T-Series for low-power/embedded...videocardz.com
Forget clock speed. What I want is a 12900k forced to 142 watt max under any load. THEN benchmark it vs a 5950x stock at the same 142 watt, using a very large and intensive benchmark suite. And log running benchmarks, like 20 minutes to an hour (to see heat and hitting the wall on temps, etc) Also, both should have the best memory speed and timings available for their platform (within reason. For example, my 5950x's run 4000@3800/1900 speed. cl16) Not sure the most reasonable fast DDR5 speed is today, but the memory should not be the most expensive, but definitely upper tier. Based on these results, I could better judge Alder lake. Even the above is only 95 watt. for 6/4 cores I would expect efficient to be about 65 watt, maybe a little more.
Heat and power usage are really the only things I have against Alder lake.
This is the way!
I saw some speculation that decision to drop AVX512 came so late, that each ADL chip has P cores with burned in V/F curves ready for AVX512 stability. People are doing easy >50mV undervolts while keeping perfect stability everywhere.
Kinda shows how much Intel cares about desktop market if they don't mind their CPUs burning 50W extra.
Oh and obviuosly undervolts double dip into power, as now E cores are not being overfed with volts either, so win-win for team Intel.
Oh, I have already commented that they use only a little too much power, and as seen can be undervolted. 95 watt in the example is still a little high for what it is, but its in the ballpark.Take a second look at the 12700 and 12600 skews. Those are much more sane and arguably a superior product.
The top end skew feels like it is overclocked to compete with AMD in the press.
When the cheaper Intel mainboards eventually show up, the software issues get patched, the 5600x, 5800x, and 5900x are going to be facing some real competition.
When AMD drops the price on the 5600x it will kick off the great CPU war of 2022.
Sales charts are nothing?
Its FACTS. Early facts and a small amount of data, but its not nothing. Just early results. And they are not good.For what, one store, one day, in the first week of sales?
Yes, that's nothing.
It's barely out the door, it's MUCH too early to jump to absurd negative conclusions, based on extremely sketchy info.
Its FACTS. Early facts and a small amount of data, but its not nothing. Just early results. And they are not good.
This could change in a month or two, but you are discounting valid information.
I see this as a good thing actually.I think early sales figures could be fairly deceptive, and looking at Micro Center on Thursday (built a 10400 for a client who just needed a workstation for an auto dealership, at $179, it was by far the best price/perf combined with a $119 Asus Prime B560), things are not looking healthy for CPU sales at the moment in general.
They had a fair amount of 12th gen in stock, and mountains of Zen3 accumulating. Motherboards on AM4 were almost comically overstacked. I asked about what was selling and the manager said there's been a fair amount of excitement with Alder but that overall things have been pretty slow outside of the campers for periodic GPU batches. It feels like the pandemic stimulus checks contributed to a solid buying wave in 2020 to early 2021, but it's been declining steadily over time now.
The answer isn't hard to find. The only GPU for less than $1000 in stock was a $199 2GB Quadro P620, basically a card just to give simple video output. With no GPUs in sight, what you're left with are the extreme edge cases of buyers with a couple grand+ for a total new build with scalper level GPUs, and the handful of prosumer guys needing newer faster stuff for their work. The average consumer is boxed out of the market entirely.
It ends up pretty dire looking forward. I've seen all the reviews, looked at all the great stuff. The W11 nonsense, overpriced boards, first wave DDR5 teething, all this shall pass. $200 12400 will be a return to outstanding value in that segment. But in the end, for what? For who? It's mostly tier for tier a bit better than Zen3, but that's been out for ages, and anyone who had money that needed a current gen platform probably already bought for the most part with the stimulus waves. A slightly better product to pair with no GPU at all (at any reasonable price at least) makes no sense to most people.
I don't really see much light in this tunnel my friends, and if all this product starts backing up in distribution and retail, it doesn't really make Raptor, Zen3D, and Zen4 look all that hopeful in turn. Increasingly great products, for a vanishingly small market, shrinking segment.
Part of the reason for slow uptake is probably also the continued expense and scarcity of high end dgpus, added to the already high platform costs for motherboards and ram.
And as others said, it is late to the party. I am glad to see Intel competitive or even in the lead in some cases, but it is not a compelling upgrade to Zen 3.
For what, one store, one day, in the first week of sales?
Yes, that's nothing.
Forget clock speed. What I want is a 12900k forced to 142 watt max under any load. THEN benchmark it vs a 5950x stock at the same 142 watt, using a very large and intensive benchmark suite. And log running benchmarks, like 20 minutes to an hour (to see heat and hitting the wall on temps, etc) Also, both should have the best memory speed and timings available for their platform (within reason. For example, my 5950x's run 4000@3800/1900 speed. cl16) Not sure the most reasonable fast DDR5 speed is today, but the memory should not be the most expensive, but definitely upper tier. Based on these results, I could better judge Alder lake. Even the above is only 95 watt. for 6/4 cores I would expect efficient to be about 65 watt, maybe a little more.
Heat and power usage are really the only things I have against Alder lake.
I agree on all points. EXCEPT... DC so far has worked best for EPYC is come cases (more cores, but still WAY more power than gracemont) or 5950x. In primegrid, I can't believe what a monster the 5950x was. In Rosetta, the EPYC seems to be best, especially because it needs lots of memory and a lot of cores. Gracemont has no place in DC that I can see, but if I could try them, I could see.In a well multi-threaded application more cores running at lower clocks wins over less cores with higher clocks. I don't think anyone would debate that. When the architectures and process are comparable and the application is highly and efficiently threaded more cores at lower clocks win. No doubt about it. For how you are running your rigs the 5950X is the best option. If we were to enter "I wish land" then and Alder Lake with 40 Gracemont cores I think would be quite efficient for distributed computing.
On the other end of the spectrum is someone like me who doesn't need massive core count. I think 6 or 8 really fast cores. Power isn't so much of an issue as I need them to not stutter during heavy compute during playback of multitrack audio with lots of plug-ins or during transitions while previewing during video editing. Encoding isn't a priority as I can do that in the background while I'm getting other work done.
The great thing in my mind about Intel being back in the game is the following.
1. Competition, which will hopefully start a more aggressive leap-frogging of CPU's and process development.
2. Competition, which will drive down prices.
3. Competition, which will drive Intel and AMD to develop a variety of truly different designs (like 5950X vs 12900K) that have different strengths and weakness, which will suit various user needs. It's not a one size fits all game when you bring pricing into it. We want the best bang for our buck for our individual workflow.
Power users can't take the heat of the 12900k for sustained use.
Yes, but who wants the extra cost and possible problems ? And heating your workspace. And for a big company, the power use ? almost twice what a 5950x would draw ? Their power bill would make the CIO's scream...Watercooling seems to tame it. AiO or custom. Still wondering if 10ESF has hotspotting issues though.
Yes, but who wants the extra cost and possible problems ? And heating your workspace. And for a big company, the power use ? almost twice what a 5950x would draw ? Their power bill would make the CIO's scream...
Talk to someone with a small data center. If I replaced all my high core count systems with 12900k, my power bill would double or more. More power also mean more heat, and more AC. I already have the AC running with all the windows open at 55F outside temps. And I only have 1000 cores. A big data center with thousands of cores ??? Or a drafting company like my son works for ? It would kill them.
Prosumer desktop. Small but not completely insignificant market that SKUs like the 5800-5950 and even Threadripper thrive in. To people who need these high core count high performance units to do their jobs, faster performance can be a very direct way to increase productivity and income. Drilling in a bit, someone who needs very bursty type work, say a Photoshop master or production editor using Premiere, they could very reasonably use something like the 12900K to get more done, on demand, as the clients wait. On the flip side, for uses like continually 24/7 loading up encodes, big data, going to a 5950X may make more sense, or leaving consumer sockets entirely for what lies above them.
The 12900k works perfect for @AdamK47 , yes ? He has the budget, doesn't care about power use, as he has ONE, I think he said he was it under water....Agree with Mark, 12900K is a compelling CPU for a *very* specific type of person, but its niche is pretty slim, IMHO. AIOs of the 360mm variety and up (custom loop) are quite a consideration in added cost and annoyance over products that can perform quite well under solid air cooling. As we've seen, running a 12900K in the top power and performance mode is unfeasible under even a monster DH15 rig. Backing it off a bit makes it work, but shaves what, 10% or so off sustained loads? At which point it defeats the advantages over 5950 for bursty/peaky stuff like gaming and Photoshop.
The i5s and i7s otoh are real standouts, wibbly bits of growing pains notwithstanding.
It's all a bit uncomfortable with the overall diy market at present unfortunately. If Intel can deliver on ARC in a big way, it could save not only Intel DIY, but really help AMD as well. A hypothetical monster Zen4 would also come out to "ehhh" if the GPU situation continues to be disastrously bad.
The 12900k works perfect for @AdamK47 , yes ? He has the budget, doesn't care about power use, as he has ONE, I think he said he was it under water....
Edit: its a 360 AIO. Same thing the original threadrippers worked well with, before the NH-U14s tr4-sp3
Aside from him, there are few it works for. The Alder lake builders thread only has him doing the 12900k. There are a couple others doing 12600k and 12700k. That about it, 3 people. 2 considering.
Backing it off a bit makes it work, but shaves what, 10% or so off sustained loads? At which point it defeats the advantages over 5950 for bursty/peaky stuff like gaming and Photoshop.
The i5s and i7s otoh are real standouts, wibbly bits of growing pains notwithstanding.