Friday, June 24th 2022
Intel "Raptor Lake" Core i9 Sample Powers Up, 8P+16E Configuration Confirmed
An engineering sample of a 13th Intel Core "Raptor Lake" Core i9 processor hit the web, courtesy of wxnod on Twitter, which confirms its 8P+16E core-configuration in a CPU-Z screenshot. Based on the same LGA1700 package as "Alder Lake," and backwards compatible with Intel 600-series chipset motherboards, besides new 700-series ones, "Raptor Lake" combines eight "Raptor Cove" performance cores (P-cores), with sixteen "Gracemont" efficiency cores (E-cores).
"Raptor Cove" features a generational IPC increase over the "Golden Cove" P-cores powering "Alder Lake," while the "Gracemont" E-cores, although identical to those on "Alder Lake," are expected to benefit from the doubling in L2 cache per cluster, from 2 MB to 4 MB. The ISA as detected by CPU-Z appears to be identical to that of "Alder Lake." The processor is a monolithic silicon chip built on the Intel 7 (10 nm Enhanced SuperFin) silicon fabrication process.
Sources:
wxnod (Twitter), VideoCardz
"Raptor Cove" features a generational IPC increase over the "Golden Cove" P-cores powering "Alder Lake," while the "Gracemont" E-cores, although identical to those on "Alder Lake," are expected to benefit from the doubling in L2 cache per cluster, from 2 MB to 4 MB. The ISA as detected by CPU-Z appears to be identical to that of "Alder Lake." The processor is a monolithic silicon chip built on the Intel 7 (10 nm Enhanced SuperFin) silicon fabrication process.
104 Comments on Intel "Raptor Lake" Core i9 Sample Powers Up, 8P+16E Configuration Confirmed
Cinebench R23 is one that Intel wins for reasons unknown. My guess is that the room scene is heavily reliant on AVX2 which Alder lake does twice as fast as Zen3. The team here working on Cinema4D threw their jobs at the one and only 12900K rig I bought to test on and said it was slower by about 15% than their workstations (5950X).
Blender is another one that's hard to say which CPU is better for because it can favour Zen3 or Alder Lake depending on the scenario. Blender simulations aren't heavily-threaded and favour IPC on a primary thread. Blender rendering (no 3rd party renderers) favours the 5950X because it'll fully-load all cores quite easily. Blender really isn't a good candidate for CPU rendering though as it's one of the better products for GPU rendering and unless you have a potato GPU, that is the preferred option for anyone modelling and rendering in Blender.
For pretty much everything else I've ever heard of, pure CPU rendering will still always favour the 5950X in real-world testing because it simply has a core count advantage, and each of its cores is close enough in IPC to Alder Lake to make it a brute-force win on core count alone. The other reason is that unless you have a water-cooled workstation, most professional systems will throttle PL2 on Alder Lake and not achieve the performance of a brief benchmark on a larger-scale render job. Yes, if you have an enthusiast PC that's capable of maxing out PL2 and keeping your 12900K within thermal limits with un uninterrupted 250W+ then it's a much closer race. In the real world, workstations typically use air cooling and a 12900K isn't going to have unlimited power or low die temperatures under extended full load.
Meanwhile, a 5950X is basically fine with an air cooler and has a max real-world power draw of 142W unless you manually overclock it. The 5950X I build here have enough cooling to comfortably cool 175W so that's what I set PBO to and call it a day.
Core i9 12900K processor review - Performance - Content Creation Blender (guru3d.com)
Core i9 12900K processor review - Performance - Content Creation POV-Ray 3.7 (guru3d.com)
Core i9 12900K processor review - Performance - SpecWorkstation 3 (guru3d.com)
Core i9 12900K processor review - Performance - Compression software 7-zip (guru3d.com)
Core i9 12900K processor review - Performance - Video Encoding (guru3d.com)
Core i9 12900K processor review - Performance - Content Creation FryRender (guru3d.com)
Core i9 12900K processor review - Performance - Corona Ray Tracing (guru3d.com)
It's kind of irrelevant though as a 5900X is vastly cheaper than a well-cooled, DDR5 12900K and if you need to do this professionally you are almost certainly going to dump larger jobs out to either a cloud rendering platform like Amazon where you can borrow 5000 CPUs for an animation briefly, or push your render out to a farm of renderboxen (like we do, and our vis. department only qualifies as a relatively small studio).
It made me caught between a rock and hard place when I burnt money on 12th Gen thinking I would be happy with only 8 P cores and maybe the e-cores give them a chance, but nope they were crap and buggy. In reality I wanted more strong cores. And I wanted faster IPC and clocks. Zen 3 only 11% behind in IP, but it is so hard to clock it anymore than 4.5 to 4.7GHz all core. Where as Golden Cove even on air you can get to 4.9 or 5GHz all 8 P cores without trouble even with an average to below average bin but beyond that you need great AIO or custom loop or more.
Basically have to wait for Zen 4 as Intel does not have even HEDT anymore as an option with Golden Cove cores and even if they were supposed to by now it has kept getting delayed and delayed yet again if it ever shows up.Problem with Threadripper is not only price, you have a minimum of 24 P cores contributing to that price. And also do not get IPC of Golden Cove or even Zen 3 yet as Zen 3 Threadripper is not even on DIY market yet.
Many people want 10-12 P cores and that is perfect. But 8 is the good minimum for a high end system these days and enthusiasts like to spec and overprovision a few over it and do not like more e-cores. Intel is losing lots of enthusiast customers to AMD than if they offered a 10-12 P core Golden Cove or Raptor Cove chip.
People like HenrySomeone providing nothing constructive as to debunking what Wolverine said: :laugh:
Me: /facepalm
As for those whining about heat, my 12900K has been running surprisingly cool, even when tweaked and overclocked a little. Not that i need to, i run it stock and actually tune the voltage down a little bit without losing performance...which feels really weird, but that let me then tune the turbo bins to get higher average performance for the same heat output and power usage lol my 9900K ran way hotter sat at 5GHz...when under full load anyway, which isn't an event that happens often even when i was running a game, though some did chew that CPU close to 90%! Though i must admit I've had my 12900K since late December last year, hearing anything about 13th gen makes me go grrrrrrr!Oh god, I'm glad i missed the crap he wrote before i posted my last reply or I'd have to flame! I was on Windows 11, maybe Windows 10 was slightly faster at the time due to the scheduler being not so optimized but...did i really notice? No. My GPU was no longer as gimped by my CPU! Only thing I've complained about is the price of DDR5 :laugh: though it *is* better in many ways...the speeeeeed (bandwidth)! Previously only achievable with HEDT and 4 memory channels, you can hit over 100GB/s on a mainstream!
TL;DR The architecture is only slightly worse than Zen3, but out of the box, it is pushed too far.
I do not have a problem per see with e-cores themselves on many SKUs as you can disable them. What I do have a massive problem with is the hard lock at 8 P cores. And Intel is just continuing to add more and more e-cores and staying at 8 P cores which is just wrong and not a good choice to do. Its not even a hybrid arch anymore like Alder Lake. At least Alder Lake had an equal amount of e and P cores or always more p cores. You could just shut off the e-cores and have a great 8 core chip, but no more than 8 strong cores. Now with future gens all Intel has made Big.Little more like Big.LittleXXXXX with the X being e-cores increasing each gen per the roadmaps.
And there is not even HEDT to turn to anymore. Well there was supposed to have been and even been here by now rebranded from HEDT to Mainstream Xeon Workstation or something like that (but who cares about the name), but it keeps getting delayed and delayed yet again and again if it ever sees the light of day.
Intel actually had better options 5-8 years ago back in the day than they do now despite their being no competition from AMD. On mainstream, you were stuck on 4 cores 8 thread chips. But they had a great HEDT platform where you could get 6 core chips (starting with Bloomfield through Ivy Bridge-E in 2010 to 2013) than 8 cores and even 10 cores starting with Haswell-E through Broadwell-E form late 2014 through 2016. Then they had Skylake-X 5 years ago as I type this though that was a step in wrong direction as latency got bad when they went to mesh form the ringbus and since have had no HEDT besides the meaningless Cascade Lake-X refresh in late 2019. Back then 4 core 8 thread chips were the same as 8 core 16 thread chips are today. And 16 cores today is like 8 cores back then. And yet Intel had excellent HEDT options with 8-10 cores that were a bit expensive, but still within reach to high end enthusiasts who have more money to spend than others, but do not want not beyond break the bank with Server high end workstation class parts that could get up to $10K or more just for a desktop.
Now despite their being competition from AMD, Intel is worse in innovation hard locking us to what was the 4 core 8 thread equivalent 5-8 years ago now with 8 cores and 16 threads, but with no HEDT option to be found to go for more on current archiecture. The e-cores are not wanted by everyone so they do not count. At the very least they need choice on mainstream platform Why not both a 10 P core Alder Lake chip in addition to their 8+8 chips as different use cases demand both especially since there is no HEDT option for more than 8 cores. Oh and before anyone like I have seen before in another thread ripping on me for wanting ore P-cores and my suggestion of that for Intel claiming it is bad to have your own products compete against each other like a 10+0 vs 8+8, well tell that to AMD. Company's own products compete against each other all the time and it serves different markets and benefits a company in the long run. I mean AMD technically has the 5800X3D competing against the 5900X and 5950X. One has only 8 cores but lots more L3 cache in 5800X3D. The other 2 have more cores but regular less L3 cache. Depends on use case. The e-cores are more trouble than they are worth for lots of people today and really in near future, and there are many who want more than 8, so a 10 P core chip would be beyond welcome.Yeah it would not be a problem if Intel had chips with more than 8 P cores. AMD is not hard locking us into chips with 8 at most P cores. They plan to use e-cores on lower end and mobile chips, but not flagship Ryzen Zen 5 parts.
Of course things can change and its pretty scary what someone else mentioned that companies could try just throwing more e-cores down our throat while decreasing P cores even more and market high core counts of weak cores and save money. I do not think we are anywhere near that point, but still scary thought.CrAsHnBuRnXp was actually agreeing with me. Read their posts earlier in this thread.I am not a fanboy of either company. In fact I actually slightly prefer Intel when they have comparable IPC and clocks and big boy core counts. If AMD was maxing out at 8 P cores and just adding more e-cores to CPUs, I would be just as upset with them.
I am going all AMD cause Intel has no choice of more than 8 P cores. And while Intel Golden Cove and then Raptor Cove cores will have mild/modest better IPC than Zen 3/ Probably Zen 4 too, the IPC deficit is not enough to make me switch to Intel and be stuck at 8 P cores and hybrid arch if I want more. Its no where near the IPC gap of even the Conroe vs K8 days let alone the 50% or more IPC deficit AMD had with Bulldozer and pilediver compared to Intel counterparts of the day.
We'll get there someday, but today this doesn't bring much to the table.