From the same source, the core configurations for the first 3 ADL SKUs are as follows:
Bwahahahaha. I like how you have to pray that internet rumors are true because you don’t know anything at all and hence cannot even tell when leaks are nonsense.
Or you can like, use that EE degree you claim to have (LOL) and think through leaks and decide whether they are likely to be true or not. Good luck with that.
Keep digging
Says the guy prematurely gloating because of an internet leak. Oh boy. It’s like gloating about going to the Fyre Festival.
Well it shouldn't be long now. I'm more than happy to wait while you dig yourself deeper. That 3.9/3.7 turbo/base number sounds right on the money. I can't wait to revisit this topic on release
If rumours/leaks are to be correct, this could be the best release ever made by Intel for years.
These spec('12900k', 5.3/3.9) are the same as VideoCardz's quote to the rumor/source which was from a chinese forum.
And the thread in that froum was deleted due to unknown reason......(sry but I'm more interested in the rumor rather than product itself lol)
Well it shouldn't be long now. I'm more than happy to wait while you dig yourself deeper. That 3.9/3.7 turbo/base number sounds right on the money. I can't wait to revisit this topic on release
The AC numbers aren't the base, it's presumably the allowed all core turbo.
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,785
- 136
If we believe Raichu's number and few posts ago analyzed Gracemont to be 10% faster than Skylake, Sunny Cove is just 7% faster than the "small core". Intel was calling Lakefield Big/Bigger core, and this would indeed be the case for Alderlake.
Actually if Intel does indeed achieve another 30% gain over Tremont it'll end up better than Skylake.
Tremont = Ivy Bridge, Haswell in some aspects
Haswell = 10% better than IVB
Skylake = 10-15% better than Haswell
Still that's quite a stretch to get 11.5K points.
Rocketlake = 6,000
Tremont 4C 10W = 925
Rocketlake x 1.2 = 7500, which means Gracemont has to get in the range of 4000 points. That's 4x the performance of 4C 10W Tremont with 2x the cores. That means a hypothetical 4C Gracemont will be over 2000 since it doesn't scale linearly. This means 4C Gracemont will perform like 1065G7 peak and 1165G7 in average scores even though it lacks Hyperthreading.
If you add 30% because of HT, 8C Gracemont will get 5,300 points, meaning after you normalize for clocks we're getting into better-than-Rocketlake-perf/clock territory, nevermind Skylake!
2200 points for a 4C Gracemont means either Tremont 10W only runs MT at 2GHz(it does not. The 6W runs slightly over 2GHz) or the perf/clock gain over Tremont is in the 40% range not 30% and in the range of Sunny Cove not Skylake. You can see why some of us think the scores seem too high.
@eek2121 What the hell does it mean by "much better AVX2 than Skylake"? Skylake has dual 256-bit FMA units. You think Gracemont has more? Cause I don't think so. And the load/store units aren't going to be twice larger either.
Last edited:
7600K gets 1609, but I believe that's at 4.2 for all core. 6600K gets 1446. 3000 seems doable if it is Skylake-like IPC but for 4000 it would for sure have to be better.
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,785
- 136
The red flag to me is that Tremont 10W gets 925 score with MT clocks running well over 2GHz, while this rumor is showing 80% gains outside of architectural gains of 30%! If the all core Turbo is 3.7GHz, then either Tremont 10W is just at 2GHz which is wrong, or Gracemont is 40% or even 50% faster per clock. That would be fantastic, but I'm very, very skeptical.
As for Golden Cove, we have that Intel document saying "up to" 20%. If the clocks are at 5.3GHz then we're not getting something performing better than 20% faster per core.
Also, this assumes zero overhead in MT running the cores simultaneously.
Even 3,000 is pushing it.
925 x 1.2(clocks) x 1.3(architecture) x 1.9(cores) = 2750
The J5040 already runs close to 3GHz. The 10W Tremont performs 25% better than that, meaning it's at least 2.8GHz or so. 4,000 requires Gracemont to be 72% faster than Tremont at the same clock!
Let's work the other way.
8x Golden Cove = 7500
8x Gracemont = 4000
8x Golden Cove sans HT = 5800
Let's assume a conservative 4.3GHz MT clock for Golden. And 3.7GHz for Gracemont.
3.7/4.3 x 5800 = 5000, or a mere 25% faster than Gracemont per clock.
As for Golden Cove, we have that Intel document saying "up to" 20%. If the clocks are at 5.3GHz then we're not getting something performing better than 20% faster per core.
Also, this assumes zero overhead in MT running the cores simultaneously.
Even 3,000 is pushing it.
925 x 1.2(clocks) x 1.3(architecture) x 1.9(cores) = 2750
The J5040 already runs close to 3GHz. The 10W Tremont performs 25% better than that, meaning it's at least 2.8GHz or so. 4,000 requires Gracemont to be 72% faster than Tremont at the same clock!
Let's work the other way.
8x Golden Cove = 7500
8x Gracemont = 4000
8x Golden Cove sans HT = 5800
Let's assume a conservative 4.3GHz MT clock for Golden. And 3.7GHz for Gracemont.
3.7/4.3 x 5800 = 5000, or a mere 25% faster than Gracemont per clock.
Last edited:
moinmoin
Diamond Member
- Jun 1, 2017
- 4,952
- 7,663
- 136
Thanks a lot for that reference, reading on it is a riot. Finally a perfect analogy I can use for all the lackluster planning at my workplace.
It’s all AVX. Gracemont’s AVX2 implementation is better than skylake. Intel’s own documentation confirms this. Any AVX driven workload will be faster than Skylake.
Golden Cove will probably be an AVX monster as well.
I expect Intel will eventually (in the coming years) bring the atom cores to full performance/feature parity with the big cores and retire the old design completely.
At 3.7GHz a RKL is at about 4800pts, and that s with something like 32% SMT gain, without it and in 8T/8C mode it would be at 3700pts at said 3.7GHz...
Intel Core i9-11900K & i5-11600K „Rocket Lake-S“ im Test: Benchmarks in Anwendungen
Core i9-11900K und i5-11600K im Test: Benchmarks in Anwendungen / Multi-Core-Szenarien / Single-Core-Szenarien
www.computerbase.de
Intel Core i9-11900K & i5-11600K „Rocket Lake-S“ im Test
Intels finale 14-nm-CPUs für Desktop-PCs alias Rocket Lake-S setzen im Test auf die Brechstange, um auch gegen den Vorgänger zu bestehen.
www.computerbase.de
So 12900K gets a 5.0/3.7 GHz allcore apparently.
162x1.275x8+5000= 8262
115x0.95+8+3700= 3233
Doesn't change much, combined it's almost at 11500 but there might be a small overhead, this 11300 score makes sense. Gracemont is close to Sunny Cove IPC according to the chinese leaker, although it's not clear to me if it refers to Cinebench only or in general. If Cinebench R20 gets a good boost from AVX instructions it can look quite different in non AVX tasks compared to Tremont.
Yeah, thanks, typo. You get what I mean. These clocks seem accurate. Not going to wade into the score debate at this point.
I'll wait to see the actual scores when a finalized product makes its way to reviewers, same as I've done in the past, same as I did with AMD. The internet is rife with pro Intel messaging not unlike the massive hypetrains AMD of yesteryear and even now suffers. Perhaps it's an age thing and I've grown tired of all the BS marketing and so-called leaks.
From a far saner viewpoint, it's expected Intel's product will be better than their competitors. That is what competition is. Even if it's got a 5% lead over Zen 3, it's still competition. Competition spurs better products. The consumer will always win in this matter, even if Intel's product will come out anywhere from 12 months to 15 months later than Zen 3.
We already knew Zen 4 is coming out in 2H22, likely 4th quarter. We've known this for the better part of a year, but going by some posts I've seen elsewhere you'd think this was brand new top secret info.
From a far saner viewpoint, it's expected Intel's product will be better than their competitors. That is what competition is. Even if it's got a 5% lead over Zen 3, it's still competition. Competition spurs better products. The consumer will always win in this matter, even if Intel's product will come out anywhere from 12 months to 15 months later than Zen 3.
We already knew Zen 4 is coming out in 2H22, likely 4th quarter. We've known this for the better part of a year, but going by some posts I've seen elsewhere you'd think this was brand new top secret info.
Gideon
Golden Member
- Nov 27, 2007
- 1,641
- 3,678
- 136
Why I give some credibility to the rumors is that AFAIK both Raichu and Bullsh1t_Buster have been very accurate with their Zen 3 and RDNA2 rumors and they stand by them (both perf and clock speeds)
It also makes some sense, considering even Tiger Lake was architecturaly (high-level, not talking about sythesis) pretty much done by the time Intel's process woes really kicked in and Alder Lake is their first manifest of a reaction to Zen, Intel had quite a time to cook up.
Still, if the rumors are accurate, this is the biggest achievement Intel has done since Conroe and an undersell of similar proportions to Zen 1:
Intel promising 20% more performance (Golden Cove vs Willow Cove) but rumors point to delivering closer to 35% in MT workloads (big-core to big core including the 300 Mhz uplift from 4.7Ghz to 5Ghz) and all in all near doubling of their R20 score when including small cores. And in no special AVX-512 workload, but a legacy AVX2 workload that has represented quite accurate average IPC changes in the past for both AMD and Intel.
Delivering a 60% IPC uplift over Skyake (again, when true) is a bigger achivement than many here give credit to. Yes, Skylake is 6 years old now and it's "nothing special just around way too long" as some said. Yes, it's like comparing Skylake to Nehalem but:
- In 6 years Skylake got only about 43% better IPC than Nehalem (~20% from Sandy Bridge and 5.8 + 11.2 + 3.3 + 2.7 % from the rest)
- That's with 4 full node shrinks (45nm -> 32nm -> 22nm -> 14nm) at a near 8x density increase compared to about 2.8x for Alder Lake (1.5 node shrinks, thereabout).
- Also while Dennard's scaling was running into wall, these nodes still had incredible power-efficency gains allowing not only a ton of extra transistors to be added but also to be turned on at the same time. 10nm doesn't give half that.
- The die sizes of Nehalem 4/8 CPUs (Bloomfield and Lynnfield) were 263 mm2 and 296 mm2 without the IGPU. Bloomfield had about 731M transistors. a 4/8 Skylake is 122 mm² and about 1,750M transistors, so about 8x more per mm2
This is just to give you the scale of what these rumors entail.
coercitiv
Diamond Member
- Jan 24, 2014
- 6,202
- 11,909
- 136
So IF the rumor is true THEN people should have given them more credit? As in before the proof?
coercitiv
Diamond Member
- Jan 24, 2014
- 6,202
- 11,909
- 136
Look who's talking.
Read Gideon's post again, maybe even this for good measure: he made a nice analysis of what this rumor entails in terms of PPC in CB 20. If that rough CB20 score formula is true then Intel have a strong product in their hands, however one cannot keep distance from the leak and wait to see further confirmation while also claiming that many here aren't giving Intel enough credit. Either @Gideon believes the leak and claims other people should follow, or he is still cautiously optimistic and therefore should not ask others to give Intel credit just yet.
Gideon
Golden Member
- Nov 27, 2007
- 1,641
- 3,678
- 136
I don't believe anything yet, i'm just extrapolating from the rumor.
I actually meant just the opposite with that comment. Somebody said that being a lot faster than skylake is trivial because it's that old. I wanted to specify that based on Intel's past performance that is not the case.
They haven't really ever achieved anything like that in the last 20 years, bar Core 2. And 40% from Nehalem to Skylake was with a buch of full node shrinks and 8x density increase.
If there's any truth to it its loads better than what intel has achieved in a decade
My Life Is Dire has also been correct on a few, emphasis on 'few,' occasions but it doesn't mean anyone would give it credence now. I'm not familiar with either of the two people you've listed. Not trying to be rude to you here, Gideon. Best to tame expectations than rile ourselves up and feel like a weather-beaten saggy balloon in four or five months when Alderlake launches and possibly turns out to be a flop for one or more reasons.
Now, on the other hand, if Alderlake final turns out to be better and is this decade's Conroe, then I'd hope to high heaven or whatever deity or supernatural force y'all believe in if any that AMD can answer Intel in the coming year or two. It is infinitely better when these two companies go at it like an old married couple slapping each other in the head when we, the consumer, win with better hardware each year.
Not true. They did design a processor that doubled as a space heater and another processor that smashed its figurative head into a wall to communicate with its twin brother that was also hot and slow.
Last edited:
TRENDING THREADS
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 10K
-
Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)- Started by Vattila
- Replies: 13K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 7K
-
Facebook
Twitter