• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Intel's Next-Gen Desktop Platform Intros Socket LGA1851, "Meteor Lake-S" to Feature 6P+16E Core Counts

The core i7 13700k got 8P-core and 8 e-core. It's a 16 cores/24 threads CPU, not 24 cores.
Obviously I meant 13900K considering I am talking about a 24 core CPU with 16 E cores that does exist. Just a typo.
 
If they stick to 6P they will get completely and utterly trashed by Zen4 3D in gaming. Already it looked dodgy for them given that we'll get that first, never mind all the upgradeability to probably up to Zen6 3D. Even in MT it's not so clear all the e-cores make sense, I'm looking at AV1-SVT stuff and the core-count scaling just isn't there, so that can only get worse as AMD keeps making stronger cores for themselves compared to Intel's focus purely on rendering. Meteor Lake is not looking very promising at all, but I understand it's a necessary step, they had nowhere else to go so now they have to build back up. Unfortunately for them it also means it's not worth buying.

As for AM5 upgrading, yeah most people won't do it but what people are missing is how cheap it will be to upgrade. You can essentially sell the old CPU for 70%ish of the purchase cost then upgrade to the new one (which is likely to be similarly priced); the bigger the market the easier this process is. So in effect you'll get a 50% performance upgrade (and whatever efficiency gains) for a 25-30% cost + the hassle of re-selling. To me that looks primo & a no-brainer.

With every passing day AM5 looks more and more worth it. :rockout:
 
I'm gonna give it to you straight: Yes, and happily so, at least they're not making promises they did not intend to keep. AMD's treatment of TRX40 buyers in fact is far more scandalous than any of these examples brought forward, and I dare say even considering the updated VRD specs between HSW/BDW and SKL+KBL/CFL's two generations, because while the fix was simple - at least in LGA 1151 v2's case by doing a jump from two points in the LGA with a piece of tape - Intel had actually done some hardware level changes. In socket AM4's case, the problem was only AGESA - from the original Excavator chip Athlon X4 970 chip to the 5950X - the electrical specification of the socket was kept intact.

Actually, write this down and charge me on it; if Zen 5 or Zen 6 end up heavily outperforming Meteor/Arrow Lake, X670 motherboards won't be updated citing <insert excuse here>, just as I was for opting for a high-end X370 over a B450 board at the time I first built my rig. If you heard even half of the utter nonsense I was spouted at, such as "Crosshair VI Hero and Extreme have such poor VRM designs that a $60 B450M board will beat them!" or "smh just don't be poor and upgrade your motherboard" (and hell, I did, because otherwise I would be forced to delay my upgrade to Zen 3 by over a year), you'd understand why I feel this way.

Unlike GPU, I do not upgrade my CPU every generation, I used my Core i7-990X from 2011 to 2017, my i7-4770K from 2013 to 2019 - and I only upgraded from Zen 2 to Zen 3 because my brother was on Sandy Bridge(!) and no one deserves to be on Sandy Bridge to this day. He is happily enjoying my old 3900XT now.
I get you, I do, I just don't see why that should lead to Intel, who have a much worse track record and where you can be absolutely shure one socket will only support two CPU "generations", the quotation marks indicating that what Intel perceives as two different generations is not what you and I would perceive as such.

I had bad experiences with AMD, too, I bought one of the first AM2 Mainboards in hope of upgrading to Phenom later. Only it was an Abit KN9 Ultra and Abit went benacrupt a short time after Phenom release and only some of their last AM2 Boards with nForce 520 Chipset got Phenom support. Had I bought a Gigabyte or Asus with nForce 570 Ultra/SLI or one of the first P965 Boards, I would have been able to upgrade from my Athlon64 X2 3800+ to Phenom II X4 945 or Core 2 Quad Q9650, instead I was stuck with X2 5400+ BE@3,1GHz until I upgraded to my current Haswell board.
There I was sure I would upgrade my i5-4670K to Broadwell i7, only it wasn't to be since Broadwell needed Z97 and wasn't all that good. At least we got the 4790k, which I only bought two years ago.

Yes, the 400-series only happened because the 300-series boards were not designed all to well. Yes, it would have been better if 400-series boards could have used PCIe4.0. Yes, it wasn't nice that AMD didn't want to enable Zen3 on 300 series. But it happened, allthough late. But even that an upgrade to Zen2 was possible on nearly all boards early on is much more than ever to be exspected from Intel. Yes, TR40 and TRX40 was horrible for enthusiast buyers, but no one would have asked that of Intel. To be fair, HEDT seems to be dead for the time being, so there seems to simply have been no business case for non-pro Threadripper 5000 and we won't get a new HEDT plattform from AMD.

When I buy an AM5 mainboard, I can be very sure it will at least support three cpu generations. If i had bought a Z690 board early on it would have been ok to upgrade to Raptor lake, but I couldn't decide last year. Now I know Z790 won't get anything better, apart from overpriced 13900KS. I would be nuts to chose Z790 over AM5 if upgradeability is important to me.
 
Probably they have quite a bit more IPC, they overall they get faster performance than 8C + 16C Raptor Lake.
And let's be honest, Intel probably already has Meteor Lake running in the labs, they know exactly the performance and how it stacks up compared to current products.
 
I get you, I do, I just don't see why that should lead to Intel, who have a much worse track record and where you can be absolutely shure one socket will only support two CPU "generations", the quotation marks indicating that what Intel perceives as two different generations is not what you and I would perceive as such.

I had bad experiences with AMD, too, I bought one of the first AM2 Mainboards in hope of upgrading to Phenom later. Only it was an Abit KN9 Ultra and Abit went benacrupt a short time after Phenom release and only some of their last AM2 Boards with nForce 520 Chipset got Phenom support. Had I bought a Gigabyte or Asus with nForce 570 Ultra/SLI or one of the first P965 Boards, I would have been able to upgrade from my Athlon64 X2 3800+ to Phenom II X4 945 or Core 2 Quad Q9650, instead I was stuck with X2 5400+ BE@3,1GHz until I upgraded to my current Haswell board.
There I was sure I would upgrade my i5-4670K to Broadwell i7, only it wasn't to be since Broadwell needed Z97 and wasn't all that good. At least we got the 4790k, which I only bought two years ago.

Yes, the 400-series only happened because the 300-series boards were not designed all to well. Yes, it would have been better if 400-series boards could have used PCIe4.0. Yes, it wasn't nice that AMD didn't want to enable Zen3 on 300 series. But it happened, allthough late. But even that an upgrade to Zen2 was possible on nearly all boards early on is much more than ever to be exspected from Intel. Yes, TR40 and TRX40 was horrible for enthusiast buyers, but no one would have asked that of Intel. To be fair, HEDT seems to be dead for the time being, so there seems to simply have been no business case for non-pro Threadripper 5000 and we won't get a new HEDT plattform from AMD.

When I buy an AM5 mainboard, I can be very sure it will at least support three cpu generations. If i had bought a Z690 board early on it would have been ok to upgrade to Raptor lake, but I couldn't decide last year. Now I know Z790 won't get anything better, apart from overpriced 13900KS. I would be nuts to chose Z790 over AM5 if upgradeability is important to me.

I mean don't take it too strictly, after the parts got older I too bought a few to try out - I've had a X5680, a 3770K, the 5775C (and even a lower cost Z97 board for it), also a X99 build which I still own in the interim, plus all generations of AMD HBM GPUs from Fury X to VII, including the Vega Frontier - I like tech and I've happily tried things out once the dust settled, but it's not bad experiences, bugs aside the experience has not been bad.

It's the rotten ethics, people just need to realize AMD is a company and that they are all too happy to bend you over if it suits them. They're not our friends in a fight against the big evil leather-clad green monster or the deeply religious blue monster. The time has come and gone to acknowledge that AMD is not the good guy in this story, nor are they the small company and the underdog anymore.

Also to beg to differ on the 300 series motherboards being inferior. AGESA is EXTREMELY buggy and it has problems that affect major functionality such as PBO currently outstanding today. If you update the X370 boards they will work perfectly fine, I know a guy using a 5800X3D on his C6H with absolutely zero bugs (exclusive to that motherboard anyway, general AGESA issues still apply such as EDC bug).

There was a notorious stinker, the MSI Titanium but other than that they're no worse than the B450 and X470, even most lower cost B550s. AsRock just copy and pasted their design, it's possible to upgrade a X370 Taichi to X470 by simply replacing its BIOS chip. They're the exact same motherboard.
 
Intel Socket LGA1700/1800, Or How To Bring Some False Hope To Everyone
 
This is why I don’t get the whole Alder Lake / Rocket Lake DDR4 and DDR5 argument. If you buy a DDR4 version today, you will still have to upgrade motherboard, RAM and CPU if you want to upgrade to Meteor Lake in 2024.
If you're buying a CPU today you shouldnt be upgrading in 2024. You shouldnt be upgrading until 2028 at the absolute earliest.
 
As long as all the i5 SKUs get 6 P-cores, I am fine with that. IPC of those cores is all that matters.
 
As long as all the i5 SKUs get 6 P-cores, I am fine with that. IPC of those cores is all that matters.
A naughty thought: what if they don't? How can Intel properly differentiate betwen the i5, i7 and i9 series if all have 6 P-cores?
 
How can Intel properly differentiate betwen the i5, i7 and i9 series if all have 6 P-cores?
E-cores, clocks and cache. Big differences.

But cutting P-cores in favor of E-cores for lower-end SKUs is 100% idiotic (on desktop). If they did a 4P+xE i5, I would not buy that at any price.

Including more than 8 P-cores is pointless, because games do not need more than 8 cores and E-cores are more efficient for productivity. But going down to 4 P-cores would be a brain dead move with gaming-focused i5 SKUs, because no number of E-cores can make up the difference, as games do not scale like that.
 
Starting with Meteor Lake but even more so with Arrow Lake Intel is finally taking power usage seriously. Apart from fresh architecture, not tweaks, use of FPGA, chiplet design, etc I expect power to finally start improving. To those outraged at E cores, imagine the 13900K as a 24 P core cpu, you be looking at 1kW processor with insane cooling requirements or vastly lower clocks to maintain system stability. AMD sees the writing on the wall, with hybrid approach come Zen 5. Why use a V8 when you are stuck in traffic or just cruising at 60km/h? A small 4 cyliner is a much better choice. You want the V8 for actual performance. Lightly loaded P cores would use way more power than E cores. No doubt crestmont cores will be a lot better than gracemont and hopefully they show real improvements in power usage too, given there will be a lot of them, up to 40 in Arrow lake.
 
The P-Cores on the Meteor Lake CPUs are based on the brand-new Redwood Cove architecture while the E-Cores will utilize the Crestmont design. Both of these are new and improved architecture as reported by Coelacanth-Dream a few days back.

intel was check that 6-core is enough.... i guess 7ghz boost..
 
If the i9 is 6 P core, what is the i5, 4 P core? :/
Oh damn... you're right. We figured out the real agenda here!

Intel wants to return to the quad core era! Its clearly the way for them to make money on their endless monolithic rebadged Core product. It used to work, right? Why not again.

It would be funny if it wasn't true :D This really isn't progress... its nogress. Completely pointless exercises to feed marketing for the next wave of same.
 
Makes a ton of sense. I've been saying this even before Alder Lake launched.

Multicore performance for the last several generations has typically been limited by the package power, so the more E-cores they cram in, the better they will be for 100% multithreaded workloads. Games and most other productivity applications are just fine as long as there are a few threads running on performance cores.
 
What argument? It gives consumers a choice to use the newest CPUs, even if they're on a tight budget. This allows Intel to make more money. How does this not make sense?

You would rather be forced to buy a dead platform (AM4) because you can't afford DDR5 and a $300 motherboard (AM5)? That strategy doesn't see to be working out too well for AMD.

It has historically worked awesome for AMD.

Perhaps the slow sales is more a problem of competition than price, where a 7700X is hard to purchase, because, well, AM4 supports the 5800X3D. ;)

Come 2023 the cache versions of AM5 CPUs will arrive, DDR5 is down in price, and hopefuly motherboards are 150USD.
 
The P-Cores on the Meteor Lake CPUs are based on the brand-new Redwood Cove architecture while the E-Cores will utilize the Crestmont design. Both of these are new and improved architecture as reported by Coelacanth-Dream a few days back.

intel was check that 6-core is enough.... i guess 7ghz boost..
Actually, what I suspect is that the new P-cores are larger and dropping to 6 is a way to improve yields and reduce die sizes. The E-cores barely take up any space by comparison, so it’s no big deal to add a bunch of those. It might be that Intel had hoped for a smaller node to be ready for this lineup, and it’s not. We had 10 core 10-series and went backward with the new core and 11-series, maybe for this very reason. It’s just a guess, but that could be why some folks were thinking this was going to be an 8C at the top end and since might be getting scaled back.
 
You would rather be forced to buy a dead platform (AM4) because you can't afford DDR5 and a $300 motherboard (AM5)? That strategy doesn't see to be working out too well for AMD.
It's working out great for AMD. They're selling more of just the 5800X3D than all of Intel's CPUs in total, and the CPUs in 2nd and 3rd place? Also AM4.
 
With 6 P cores instead of 8, Intel will increase the max turbo to 6 GHz. Will be their selling point.

I'm one of those guys who often buy used hardware and stays a couple generations behind what is "new".
What for someone is a dead platform, for someone else can be a valid upgrade, that will last him several years.
The AM4 platform will stay popular for many other years to come. Only the people that have the money to buy the Ryzen 9 5950X can afford the AM5 jump so early. The ones with the Zen 2 and Zen 3 cpus, or even with a Ryzen 7 5800, can simply buy a used more powerful Ryzen and spend the other money in a gpu, if needed.
AMD, Intel and NVIDIA can do all the live shows they want and hype their products in every manner, but this doesn't mean that the people following these events, videos and articles will run to buy their stuff. Half of the people will keep their hardware and wait for better times. Lot of others will buy used parts. And lot of the people with the money will wait for better deals.
 
Last edited:
It's working out great for AMD. They're selling more of just the 5800X3D than all of Intel's CPUs in total, and the CPUs in 2nd and 3rd place? Also AM4.
Your numbers are certainly in line with those that mindfactory.de (MF) makes available. MF has always been a bastion of AMD sales so the figures are not that surprising. For them to sell more Raptor Lake CPUs per week (week 43 2022*) than Zen 4 is surprising. The current Zen 3 fire sale is working out great for AM4 motherboard owners but for AMD itself, look at its latest financial figures. Income from laptop and desktop processors for Q3 2022 has decreased by $1 billion in the space of a year. More than that, a profit of around $500 million for this segment a year ago has turned into a loss for Q3 2022 of $28 million.

(* 7600X 70, 7700X 110, 7900X 60, 7950X 60 - total 300. 13600K/KF 310, 13700K/KF 320, 13900K/KF 60 - total 690.)
 
Last edited:
I will just get 13900ks and skip meteor lake, i don't use the igpu anyway, second year of the socket always get you the best to keep for 2 years even for enthusiasts
 
Dropping cores means less performance, as per 10900K to 11900K transition.

The same old stupid Intel that never learns :D :roll:

View attachment 268013
Intel Core i9-11900K processor review - Performance - Content Creation Blender 2.83 (guru3d.com)
At the flagship end, yes. 6P+16E is worse than 8P + 16E.

Realistically, most people aren't using coolers capable of 350W and what will actually happen is that the reduction in P cores will free up a bunch of power budget for the 16 E-cores and multithreaded performance ought to be better than the current 8P + 16E interations.

With infinite power and cooling, this will likely be a downgrade or a sidegrade. For real-world, normal people with off-the-shelf AIOs and tower coolers, this is going to get more performance out of their limited cooling and motherboard VRMs.
 
This is why I don’t get the whole Alder Lake / Rocket Lake DDR4 and DDR5 argument. If you buy a DDR4 version today, you will still have to upgrade motherboard, RAM and CPU if you want to upgrade to Meteor Lake in 2024.

It would've made since if both men controllers were on the die, like the last transition. New mb/CPU(or just a new CPU if it was backwards compatible with existing socket) at release avoiding the early adopter tax and the rush on the new mem. Then at a later time upgrade ram/mb when cheaper.
 
Back
Top