They're literally releasing 4 new architectures within 1 year. How the hell does that even work? It's confusing as hell to customers. It's a nightmare for vendors. Release one generation, and then 3 months later, release another? Yikes.Reply
Intel 3 is literally an iteration on Intel 4. It’s like being shocked that Intel went from 14+++ to 10SF for H series mobile. Tiger Lake turned out fine, and as long as there aren’t any issues with Intel 3 or Sierra Forest it’ll turn out fine. Reply
Well... - 10SF only worked because it was preceded by a year of failed Canon Lake. - H Series Mobile wasn't ALSO a completely new design ("Granite Rapids is a tile-based architecture, with separate compute and I/O tiles – an evolution from Sapphire Rapids, which even in its tiled form is essentially a complete SoC in each tile.)
Essentially you're asking us to believe that Intel, which has CONSISTENTLY over-estimated how fast it can iterate - in process, in architecture, and in packaging - will now somehow manage to get all three right. "Sure, right now I can only juggle two balls, and sure I frequently drop one, but hell, let me try three chainsaws, I just *know* I can do it."Reply
10SF was preceded by 10nm formerly 10nm+ on a year of Ice Lake. Intel 3 will be preceded by Meteor Lake on Intel 4. Intel 3 isn’t a new node, it’s an iteration. Intel 4 is about as delayed as 10nm, and they made a fairly rapid advancement from the working and unexceptional 10+ to the working and good 10SF. I am inclined to believe they can make a similar advancement in a similar time frame from Intel 4 to Intel 3. Leaks indicate that Intel 4 can clock above 4GHz and is a real node, but that Intel may omit a desktop Meteor Lake launch because it would need to be at 5GHz to outperform Raptor Lake.
As for packaging, Intel has done Lakefield before which is more advanced than Granite Rapids. Intel should also have Meteor Lake out in 2023 which has the same level of disaggregation as Granite Rapids.
The reason why I am inclined to believe in Sierra Forrest is precisely because it will be reliant upon iterations rather than bringing something entirely new to the table. The only new thing should be the micro architecture. Intel knows how to do IO and IO tiles. They know how to do a 4x10, 5x8, or 6x6 mesh (node count depends on whether or not they will include memory controllers on the CPU tile or IO tile). They know how to do EMIB connections. And, they should have the experience of putting all 3 together this year with Meteor Lake. Reply
I agree with most of your points, @Otritus. However, I'd take issue with the notion that Meteor Lake is comparable to GNR or even SPR. The main difference is that MTL has all its CPU cores on the same tile. Its tile boundaries all align with what were traditionally all separate chips, if you go back far enough (North Bridge, CPU, GPU, South Bridge), which should significantly reduce the bandwidth requirements and energy-efficiency penalties. For that reason, I regard it as rather conservative.
What's interesting and more ambitious about Sapphire Rapids (SPR) is that Intel spanned their mesh interconnect topology over EMIB, almost like it's not even there. Excellent analysis, here:
> I am inclined to believe in Sierra Forrest is precisely because it will be reliant upon iterations
I've been suspecting the 144-core variant is actually monolithic (aside from the I/O tile, perhaps). If the quad-core tiles are still the same size as a P-core, and we agree it's plausible that you could make a 36-core monolithic tile on Intel 4, then that seems like a good place to start.
The cool thing about showing that wafer is we can estimate the chip sizes. Assuming it's 300 mm in diameter, I get a height of about 20.7 mm and a width of about 20.3 mm. So, that's 420 mm^2 for what I'm guessing are 60 cores (15 dark blobs, which I take to be quad-core clusters). Compare that to the largest SPR die at 770 mm^2 for 34 cores*.
Doing some rough extrapolation, that would make a 144-core die over 1000 mm^2, even before adding any spare clusters for yield. So, I guess the 144 core Sierra Forest is probably comprised of 2 dies of about 80 cores each.
Sierra Forrest used e-cores not p-cores. 144 cores is 36 clusters. 1 Gracemont cluster is moderately larger than a Golden/Raptor Cove core. The estimates by size vary, but that would be roughly 40-50 cores in size. When you factor in the significantly increased density of Intel 3, that much is definitely possible as a monolithic design.
Another way of looking at it is that the Golden Cove cores took up roughly 2/3 of a Saphire Rapids tile. That’s roughly 17mm2 per core. 48 cores at 17mm2 per core would be 816mm2. Intel 4 roughly doubles density, which would be around 408mm2. Factor in additional space for cache not scaling as well and a larger footprint for the micro architectural enhancements, and you can easily fit 144 cores in less than 600mm2. Locuza on YT had a great breakdown of the Alder Lake die, and Golden Cove server in total takes up around 12-14mm2 rather than the rough 17mm2 I calculated. Monolithic 144C becomes even more reasonable with these numbers.Reply
> Sierra Forrest ... When you factor in the significantly increased density of Intel 3, > that much is definitely possible as a monolithic design.
Yes, this thought experiment is where I started. I worked it out, based on Gracemont figures, although Sierra Forest actually uses Crestmont cores. So, size assumptions based on Gracemont could be well off the mark.
Then, I saw the picture (see article) that purported to be a wafer of SRF dies and that's how I arrived at the estimate of roughly 1000 mm^2 for a monolithic 144 core die. The only way my math is wrong is if that's not 300 mm in diameter, there aren't 60 E-cores on those dies, or that's not even SRF. You see the pic? The caption says it's "a finished Sierra Forest compute tile wafer". Click on it to view full size and count the dies & core clusters for yourself.
We could validate this, if we had a wafer shot of Meteor Lake. Then we should be able to work out the size of a Crestmont cluster on Intel 4.Reply
Q1 2023, Saphire Rapids a delayed 2022 product. Q4 2023, Emerald Rapids a 2023 product on time. Q4? 2024, Granite Rapids a 2024 product.
That’s 3 generations in roughly 3 years. Nothing confusing about that.
H1 2024, Sierra Forest a product that’s meant for a different kind of server than traditional servers due to using e-cores instead of p-cores. Meant to exist alongside Emerald and Granite Rapids and uses the same platform as Granite Rapids. Nothing confusing about figuring out whether your workload is better suited to more e-cores or fewer p-cores and choosing your processor based on that. Reply
We all know SPR was meant to launch well back in 2022. That explains why EMR is following so closely behind. If SPR had launched remotely on schedule, the timing of EMR wouldn't seem so weird.
I have faith that EMR won't slip, since it's basically like Raptor Lake was to Alder Lake. The most puzzling thing is why Intel didn't outright cancel it. But, if they feel it improves their competitiveness meaningfully, then they might want it to hedge against any slips in Granite Rapids.Reply
Assuming these cores are the Gracemont successor, they should have similar IPC to Sunny Cove. Zen 4c should have IPC between Zen 3 and Zen 4. Zen 4c should have similar clock speeds and SMT to e-cores. I’m going to put my money on Bergamo being better although it may use more power. Reply
> Emerald Rapids will be built on the Intel 7 process. This means that the bulk of any > performance/efficiency gains will have to come from architectural improvements.
It's probably the same Intel 7+ process that Raptor Lake uses. Do we know how much of Raptor Lake's improvements were simply due to the node refinement?Reply
I have data that contributes to the inquiry "what is the 'performance gain' from architectural improvements" and I will answer that questions comparing Alder and Raptor desktop 900KS, 900K, 900KF for the G fall out notation and 600K sent to down bin.
This is more a yield than performance analysis however K in relation KS on KS frequency bump for Alder + 300 MHz and for Raptor + 200 MHz in relation to spec is telling of the process for performance improvement.
900K = an index of 1 and 900 KS, 900KF and 600K are all compared to 900K at index of 1 (compared against itself) and that index (other SKU comparison against K) is on channel supply data as a proxy for yield.
Data sample is weekly and begins the week KS is introduced for Alder and Raptor respectively within the overall envelope of same base process subject SKU comparison looking for a sign of process improvement.
The index shows the difference in supply volume subject ease of manufacturability.
13900K = 1 12900K = 1 13900KS = 0.101 shows 197% improvement over Alder KS to achieve + 200 MHz 12900 KS = 0.034 13900KF = 0.069 12900KF = 0.065 approximately the same 13600K = 0.117 12600K = 0.109 shows 7.7% more fall out
We can also look at the SKU split for the sample beginning the week KS is introduced through last week;
For power efficiency we can look at the SKU power split and in this assessment over the full run from day 1 of the introductions;
Alder shows dynamic power within bottom of mid range
8P+8E, 150 to 241W = 1.07% 8+8, 125 to 241W = 38.26% 8+4, 125 to 190W = 24.63% 6+4, 125 to 150W = 4.83% 8+8, 65 to 202W = 4.45% 8+4, 65 to 180W = 4.10% 6C, 65 to 117W = 17.19% 4C, 58 to 89W = 0.78% 2C, 46 to 55W = 0.66% 8/6/4/2 at base 35W = 4.03%
Raptor shows dynamic frequency
8P+16E, 150 to 253W = 7.1% and configuration on 8+16 8+8, 125 to 253W = 69.77% shows top bin frequency improvement 6+8, 125 to 181W = 7.46% new configuration 8+8, 65 to 219W = 8.39% shows an improvement over Alder 8+8 6+8, 65 to 154W = 3.65% new configuration 6+4, 65 to 148W = 0.84% down from Alder but meager volume 4C 58 to 89W = 2.48% and bottom bin quad improvement 8/6/4 at base 35W = 0.31%
You decide on power efficiency there’s sufficient data here to extend the analysis.
Between the two generation samples Alder ramp volume is 88.6% and Raptor run down volume is 11.3% so one can say Raptor for a lower volume improves on the top SKU frequency or Intel mine operations are working overtime in the sort room.
Hey, I have a question that seems right up your alley. What do you make of the i5-12600 (non-K) pricing? It launched at $223 and now sells for $250 (Walmart) to $258 (Newegg), or more. It's the largest and fastest incarnation of the P-core only die (6+0). The i5-12600K is based on the big 8+8 die and frequently sells for less than its non-K namesake. Does this smell right, to you?Reply
mode_13, new on ebay around $180 to $200 any 12600_ pick your flavor. On 12600 hexa only, 65 to 117W v 125 to 150W for K_ so how about a lower priced board for 12600. For $250 to $258 at Newegg and Walmart? I just checked Newegg now $207 so maybe Walmart at $250 still has not updated their page with a competitive price. The high volume price is $111 ($223/2) however on a volume Alder i9/i7 purchase so many 12600 were likely thrown into the sales package for nothing everything below 600K is generally thrown in as sales close. The check is Alder full line Average Weighed Price $1K at $417 / 2 for a high volume procurement = $208 so at this full line product SKU buy in (i9 = 42.6%, i7 = 28.8%, i5 = 24.1%, i3 = 3.8%, Pentium = 0.27%, Celeron = 0.44%) procurement would say no to 12600 as underwater at $233 end sale and there would be a negotiation around the SKU procurement price which is likely n/c. For a tray this isn't the case it's the traditional buy 10 and get 1 free < 10%. I seriously doubt Walmart buys by the tray. I consider Neweeg and Wallmart price similar any OEM these days on competitive desktop market. Specific OEMs its different encompassing mobile and Xeon SKUs negotiated within and around the desktop buy in. mbReply
Thanks for the reply. The $207 price on Newegg is a 3rd party seller that I don't 100% trust, and not sure if Intel will offer warranty claims from. If I'm going to take those risks, I'd rather save more money and go the ebay route. Speaking of the ebay route, I have a search query that excludes any ES (Engineering Sample) chips, which I don't trust.
The reason I like the i5-12600 is that it typically has the fastest single-thread performance among the 65 W-rated models, probably due to the smaller ring bus. It's also the second cheapest model that supports ECC memory, which obviously limits me to more expensive W680 boards.Reply
ebay purchase right save the money and there's the ebay guarantee ECC 12600 good single core thanks for the note. I too would stay away from engineering samples unless you know for example the mainboard design manufacturer compatibility lab tech who was originally given the field sample by the Intel FAE. mbReply
wow you actually believe his BS ??? with out knowing where he gets this from, its all fake and BS, and not to be believed. for all you know, it pulled it out of thin air, and made it all up.
with links to sources that arent HIS page, its just BSReply
I know his posts bother you, but I don't see much harm in them. You're free to ignore them, if you don't find them useful.
IMO, unless you can clearly demonstrate posts are spreading misinformation or violate other rules, then you should try to relax and accept that he's as free to post as you are.Reply
" unless you can clearly demonstrate posts are spreading misinformation or violate other rules " my point is... no one can prove his posts either way. he never posts links to where he gets is info from, there fore, there his no way to confirm or deny any of it. he could very well be making it all up out of thin air, there fore, it could very well be misinformation and total bs he is posting. ive seen this same type of posts on other forums, and its usually fake and misinformation as the OP also cant site sources to where they got it, so its flagged as their opinion onlyReply
Besides, he stopped replying "MB.", as I think you'd requested. That was a legit complaint and he complied. Let's try to find compromise where we can, and move on.Reply
No, I meant that thing he used to do where he'd write one-line replies to people that were just like "mb.". You asked him not to do that, and he stopped.
Anyway, whatever you have against him, please leave me out of it.Reply
Who r u quasar? I am the data scientist and the work is mine as the primary source assessing as you are aware ebay and other supply data on my tenure inside various design manufacturers and design producers basically I am a modern Storeboard. Did you ever have issue with Storeboard, Price Watch or the MDR x86 annual. Marketing and introducing PC AT board products for Orchid; Tiny Turbo, PCNet, IBM Turbo PGA 1st 3D graphics card, Excell memory cards against AST and Quadram and Intel PCEO. Next for Arche tens of PCs across 286, 386, 486 including the first PC MP based on C&T MPAX chip set. Mike goes into the processor business at Cyrix 486SLC, 487SLC, Cyrix 486 DLC, Cyrix 486S. Then to Arm 7, TDMI, associated strong ARM. Then NexGen 586, AMD K100-133, K5, codecs for C-Cube, Alpha 21164 work for Samsung and then the IDT Winchip roll out and by that time Intel associate network and I were getting fed up with each other. In 1998 enlisted by the FTC. In 2008 contracted by the USDOJ retained by Congress. In 2009 reenlisted by the FTC it's been a tough road but also a very interesting and intriguing one. Compliment or confrontation? So you work for Intel? Or are a competing analyst? mbReply
i am some one who would like to SEE your sources for myself, what is wrong with that myke brazzzone ?? is it insulting that i ask for sources ? that i dont believe your posts as some of it seems like it is made up, and false ? but as i have said before, your reluctance and flat our refusal to post direct links your sources show you dont have any. you can post names and such all you want, that is meaning less. i can say my sources are X site, or Y business too, but that would prove nothing as well.
maybe you just dont like being called a fraud. who posts made up info. so, ill put it this way, IF post direct links to where you get your info from for your posts. if you do this, i will apologize for calling you what you seem to be, which is a fraud and a fake.
i guess asking for sources, is just too much to ask from some one. Reply
:IF post direct links to where you get your info from for your posts. if you do this, i will apologize for calling you what you seem to be, which is a fraud and a fake."
see, was that so hard ? but, what about the other sources you use ? those are just the financials, what about the posts claiming the other info you post on here ? like how many products amd intel or nvidia have produced ? or things about the wafers, some of your posts have quite the bit of info in them, i wont deny that, but it would be nice to see the sources for those as well.
for now, i apologize to you. but i would like to see other sources if asked, deal ??Reply
Quasar you are more than welcome to add, to compliment, to offer point and even counter point to my observations is a constructive approach. But to detract without support you know much of my data comes from ebay that is not ebay terapeak on my own filters, same as Mercury by the way, so why do you detract {?] which is not constructive and had nothing to do with management or engineering best practice. I am trained in many management disciplines including constructive confrontation and constructive collaboration and invite you to display eitherconstructively. mbReply
yea ok sure, i agree Myke brazzzonee, BUT i cant do any of what you suggest, unless i have the SAME info that you your self get, so i could say something completely different then what you have, and that in the end would prove squat. correct ? hence the asking for direct sources.
again, you still refuse to post direct sources, is there a specific reason ? is it top secret info that no one else can have access to ? either way, i cant prove what you post is wrong or right, or made up out of thin air.
as i said, asking for sources, which i have seen MANY times on other forms being asked by one person to the other, is just too much to ask from you. and like on those forums when one person is asked for a source, and doesnt provide it, they are also called out on it, like i am doing here, but what ever. keep posting info that no one can conform, or counter, as it is unclear where you really come up with it, say it is from site X and site Y, doesnt really count, IMO, as i mentioned, other people try it, and it never resolves anything.Reply
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
47 Comments
Back to Article
lemurbutton - Wednesday, March 29, 2023 - link
They're literally releasing 4 new architectures within 1 year. How the hell does that even work? It's confusing as hell to customers. It's a nightmare for vendors. Release one generation, and then 3 months later, release another? Yikes. Replylemurbutton - Wednesday, March 29, 2023 - link
Also, they're skipping Intel 4 and going straight from Intel 7 to Intel 3. That's a huge jump. Expect delays as usual. ReplyOtritus - Wednesday, March 29, 2023 - link
Intel 3 is literally an iteration on Intel 4. It’s like being shocked that Intel went from 14+++ to 10SF for H series mobile. Tiger Lake turned out fine, and as long as there aren’t any issues with Intel 3 or Sierra Forest it’ll turn out fine. Replyname99 - Wednesday, March 29, 2023 - link
Well...- 10SF only worked because it was preceded by a year of failed Canon Lake.
- H Series Mobile wasn't ALSO a completely new design ("Granite Rapids is a tile-based architecture, with separate compute and I/O tiles – an evolution from Sapphire Rapids, which even in its tiled form is essentially a complete SoC in each tile.)
Essentially you're asking us to believe that Intel, which has CONSISTENTLY over-estimated how fast it can iterate - in process, in architecture, and in packaging - will now somehow manage to get all three right. "Sure, right now I can only juggle two balls, and sure I frequently drop one, but hell, let me try three chainsaws, I just *know* I can do it." Reply
Otritus - Wednesday, March 29, 2023 - link
10SF was preceded by 10nm formerly 10nm+ on a year of Ice Lake. Intel 3 will be preceded by Meteor Lake on Intel 4. Intel 3 isn’t a new node, it’s an iteration. Intel 4 is about as delayed as 10nm, and they made a fairly rapid advancement from the working and unexceptional 10+ to the working and good 10SF. I am inclined to believe they can make a similar advancement in a similar time frame from Intel 4 to Intel 3. Leaks indicate that Intel 4 can clock above 4GHz and is a real node, but that Intel may omit a desktop Meteor Lake launch because it would need to be at 5GHz to outperform Raptor Lake.As for packaging, Intel has done Lakefield before which is more advanced than Granite Rapids. Intel should also have Meteor Lake out in 2023 which has the same level of disaggregation as Granite Rapids.
The reason why I am inclined to believe in Sierra Forrest is precisely because it will be reliant upon iterations rather than bringing something entirely new to the table. The only new thing should be the micro architecture. Intel knows how to do IO and IO tiles. They know how to do a 4x10, 5x8, or 6x6 mesh (node count depends on whether or not they will include memory controllers on the CPU tile or IO tile). They know how to do EMIB connections. And, they should have the experience of putting all 3 together this year with Meteor Lake. Reply
mode_13h - Thursday, March 30, 2023 - link
I agree with most of your points, @Otritus. However, I'd take issue with the notion that Meteor Lake is comparable to GNR or even SPR. The main difference is that MTL has all its CPU cores on the same tile. Its tile boundaries all align with what were traditionally all separate chips, if you go back far enough (North Bridge, CPU, GPU, South Bridge), which should significantly reduce the bandwidth requirements and energy-efficiency penalties. For that reason, I regard it as rather conservative.What's interesting and more ambitious about Sapphire Rapids (SPR) is that Intel spanned their mesh interconnect topology over EMIB, almost like it's not even there. Excellent analysis, here:
https://chipsandcheese.com/2023/03/12/a-peek-at-sa... Reply
mode_13h - Thursday, March 30, 2023 - link
> I am inclined to believe in Sierra Forrest is precisely because it will be reliant upon iterationsI've been suspecting the 144-core variant is actually monolithic (aside from the I/O tile, perhaps). If the quad-core tiles are still the same size as a P-core, and we agree it's plausible that you could make a 36-core monolithic tile on Intel 4, then that seems like a good place to start.
The cool thing about showing that wafer is we can estimate the chip sizes. Assuming it's 300 mm in diameter, I get a height of about 20.7 mm and a width of about 20.3 mm. So, that's 420 mm^2 for what I'm guessing are 60 cores (15 dark blobs, which I take to be quad-core clusters). Compare that to the largest SPR die at 770 mm^2 for 34 cores*.
Doing some rough extrapolation, that would make a 144-core die over 1000 mm^2, even before adding any spare clusters for yield. So, I guess the 144 core Sierra Forest is probably comprised of 2 dies of about 80 cores each.
* https://www.angstronomics.com/p/monolithic-sapphir... Reply
Otritus - Thursday, March 30, 2023 - link
Sierra Forrest used e-cores not p-cores. 144 cores is 36 clusters. 1 Gracemont cluster is moderately larger than a Golden/Raptor Cove core. The estimates by size vary, but that would be roughly 40-50 cores in size. When you factor in the significantly increased density of Intel 3, that much is definitely possible as a monolithic design.Another way of looking at it is that the Golden Cove cores took up roughly 2/3 of a Saphire Rapids tile. That’s roughly 17mm2 per core. 48 cores at 17mm2 per core would be 816mm2. Intel 4 roughly doubles density, which would be around 408mm2. Factor in additional space for cache not scaling as well and a larger footprint for the micro architectural enhancements, and you can easily fit 144 cores in less than 600mm2. Locuza on YT had a great breakdown of the Alder Lake die, and Golden Cove server in total takes up around 12-14mm2 rather than the rough 17mm2 I calculated. Monolithic 144C becomes even more reasonable with these numbers. Reply
mode_13h - Friday, March 31, 2023 - link
> Sierra Forrest ... When you factor in the significantly increased density of Intel 3,> that much is definitely possible as a monolithic design.
Yes, this thought experiment is where I started. I worked it out, based on Gracemont figures, although Sierra Forest actually uses Crestmont cores. So, size assumptions based on Gracemont could be well off the mark.
Then, I saw the picture (see article) that purported to be a wafer of SRF dies and that's how I arrived at the estimate of roughly 1000 mm^2 for a monolithic 144 core die. The only way my math is wrong is if that's not 300 mm in diameter, there aren't 60 E-cores on those dies, or that's not even SRF. You see the pic? The caption says it's "a finished Sierra Forest compute tile wafer". Click on it to view full size and count the dies & core clusters for yourself.
We could validate this, if we had a wafer shot of Meteor Lake. Then we should be able to work out the size of a Crestmont cluster on Intel 4. Reply
IanCutress - Wednesday, March 29, 2023 - link
Intel 4 and Intel 20A are going to be client focused Replymode_13h - Thursday, March 30, 2023 - link
Greetings, Dr. Potato!: ) Reply
Otritus - Wednesday, March 29, 2023 - link
Q1 2023, Saphire Rapids a delayed 2022 product.Q4 2023, Emerald Rapids a 2023 product on time.
Q4? 2024, Granite Rapids a 2024 product.
That’s 3 generations in roughly 3 years. Nothing confusing about that.
H1 2024, Sierra Forest a product that’s meant for a different kind of server than traditional servers due to using e-cores instead of p-cores. Meant to exist alongside Emerald and Granite Rapids and uses the same platform as Granite Rapids. Nothing confusing about figuring out whether your workload is better suited to more e-cores or fewer p-cores and choosing your processor based on that. Reply
mode_13h - Thursday, March 30, 2023 - link
We all know SPR was meant to launch well back in 2022. That explains why EMR is following so closely behind. If SPR had launched remotely on schedule, the timing of EMR wouldn't seem so weird.I have faith that EMR won't slip, since it's basically like Raptor Lake was to Alder Lake. The most puzzling thing is why Intel didn't outright cancel it. But, if they feel it improves their competitiveness meaningfully, then they might want it to hedge against any slips in Granite Rapids. Reply
dontlistentome - Wednesday, March 29, 2023 - link
144e cores.Larrabee, come back, all is forgiven. Reply
schujj07 - Wednesday, March 29, 2023 - link
That is the competitor to Bergamo. Only question is what will be better 144 e-cores or 128 Zen 4c cores. ReplyPixyMisa - Wednesday, March 29, 2023 - link
We'll have to wait and see, but an e-core is half a p-core, and I'd be very surprised if a Zen 4c is less than 75% of a Zen 4. ReplyOtritus - Wednesday, March 29, 2023 - link
Assuming these cores are the Gracemont successor, they should have similar IPC to Sunny Cove. Zen 4c should have IPC between Zen 3 and Zen 4. Zen 4c should have similar clock speeds and SMT to e-cores. I’m going to put my money on Bergamo being better although it may use more power. Replysonofgodfrey - Wednesday, March 29, 2023 - link
It will probably depend on the size/speed of cache & memory. Either one is bound to be memory starved. Replytipoo - Wednesday, March 29, 2023 - link
And bring Project Offset back with you damn it Replymode_13h - Thursday, March 30, 2023 - link
> Larrabee, come back, all is forgiven.Sierra Forest won't have AVX-512 (or AMX), though. Nor hyperthreading, I'm pretty sure. So, it's really not a Xeon Phi successor. Reply
DannyH246 - Wednesday, March 29, 2023 - link
Yawn...another roadmap, another marketing article from www.IntelTech.com.This time things won't be delayed. We promise. Reply
sonofgodfrey - Wednesday, March 29, 2023 - link
Some typos: Granite Forest ? (AMD is petrified of that core :) ) and Clearwater Falls ? ReplyRyan Smith - Wednesday, March 29, 2023 - link
Thanks. It's a lot of codenames to keep straight, and a lot of these are similar to local (Oregon) locales... ReplyJKflipflop98 - Tuesday, April 4, 2023 - link
That's because they are Oregon locales. We name them after natural features near where they were designed. There's lots of "bridges" in Israel. Replymode_13h - Thursday, March 30, 2023 - link
> Emerald Rapids will be built on the Intel 7 process. This means that the bulk of any> performance/efficiency gains will have to come from architectural improvements.
It's probably the same Intel 7+ process that Raptor Lake uses. Do we know how much of Raptor Lake's improvements were simply due to the node refinement? Reply
Bruzzone - Thursday, March 30, 2023 - link
I have data that contributes to the inquiry "what is the 'performance gain' from architectural improvements" and I will answer that questions comparing Alder and Raptor desktop 900KS, 900K, 900KF for the G fall out notation and 600K sent to down bin.This is more a yield than performance analysis however K in relation KS on KS frequency bump for Alder + 300 MHz and for Raptor + 200 MHz in relation to spec is telling of the process for performance improvement.
900K = an index of 1 and 900 KS, 900KF and 600K are all compared to 900K at index of 1 (compared against itself) and that index (other SKU comparison against K) is on channel supply data as a proxy for yield.
Data sample is weekly and begins the week KS is introduced for Alder and Raptor respectively within the overall envelope of same base process subject SKU comparison looking for a sign of process improvement.
The index shows the difference in supply volume subject ease of manufacturability.
13900K = 1
12900K = 1
13900KS = 0.101 shows 197% improvement over Alder KS to achieve + 200 MHz
12900 KS = 0.034
13900KF = 0.069
12900KF = 0.065 approximately the same
13600K = 0.117
12600K = 0.109 shows 7.7% more fall out
We can also look at the SKU split for the sample beginning the week KS is introduced through last week;
13900KS = 12.89%
13900K = 76%
13900KF = 3.21%
13600K = 7.89%
12900KS = 2.45%
12900K = 83.36%
12900KF = 5.63%
12600K = 8.65%
For power efficiency we can look at the SKU power split and in this assessment over the full run from day 1 of the introductions;
Alder shows dynamic power within bottom of mid range
8P+8E, 150 to 241W = 1.07%
8+8, 125 to 241W = 38.26%
8+4, 125 to 190W = 24.63%
6+4, 125 to 150W = 4.83%
8+8, 65 to 202W = 4.45%
8+4, 65 to 180W = 4.10%
6C, 65 to 117W = 17.19%
4C, 58 to 89W = 0.78%
2C, 46 to 55W = 0.66%
8/6/4/2 at base 35W = 4.03%
Raptor shows dynamic frequency
8P+16E, 150 to 253W = 7.1% and configuration on 8+16
8+8, 125 to 253W = 69.77% shows top bin frequency improvement
6+8, 125 to 181W = 7.46% new configuration
8+8, 65 to 219W = 8.39% shows an improvement over Alder 8+8
6+8, 65 to 154W = 3.65% new configuration
6+4, 65 to 148W = 0.84% down from Alder but meager volume
4C 58 to 89W = 2.48% and bottom bin quad improvement
8/6/4 at base 35W = 0.31%
You decide on power efficiency there’s sufficient data here to extend the analysis.
Between the two generation samples Alder ramp volume is 88.6% and Raptor run down volume is 11.3% so one can say Raptor for a lower volume improves on the top SKU frequency or Intel mine operations are working overtime in the sort room.
Mike Bruzzone, Camp Marketing Reply
Bruzzone - Friday, March 31, 2023 - link
For comparison, Xeon Ice; by cores, power distribution, base and max frequency.Core Grade Split
40 = 4.12%
38 = 3.10%
36 = 7.4%
32 – 22.94%
28 = 11.87%
26 = 1.74%
24 = 8.1%
20 = 3.05%
18 = 3.08%
16 = 11.94%
12 = 10.61%
10 = 9,57%
8 = 11.5%
Power Distribution
300W = 4.5%
270W = 7.63%
265W = 1.19%
250W = 4.85%
240W = 0.99%
235W = 1.79%
230W = 4.03%
225W = 1.19%
220W = 1.19%
205W = 26.11%
195W = 0.98%
185W = 9.27%
165W = 3.97%
150W = 4.93%
140W = 4.74%
135W = 5.18%
120W = 8.82%
105W = 9.41%
Base Frequency GHz
3.6 = 1.09%
3.5 = 0.42%
3.4 = 0.35%
3.2 = 4.63%
3.1 = 2.53%
3.0 = 5.41%
2.9 = 4.39%
2.8 = 14.88%
2.7 = 0.23%
2.6 = 6.56%
2.5 = 0.44%
2.4 = 11.13%
2.3 = 13.44%
2.2 = 9.08%
2.1 = 11.12%
1.0 = 14.31%
Max Boost GHz
4.0 = 1.86%
3.9 = 0.45%
3.7 = 1.37%
3.6 = 21.84%
3.5 = 24.47%
3.4 = 31.65%
3.3 = 9.11%
3.1 = 9.26%
Mike Bruzzone, Camp Marketing Reply
mode_13h - Friday, March 31, 2023 - link
Thanks for the info, Mike!Hey, I have a question that seems right up your alley. What do you make of the i5-12600 (non-K) pricing? It launched at $223 and now sells for $250 (Walmart) to $258 (Newegg), or more. It's the largest and fastest incarnation of the P-core only die (6+0). The i5-12600K is based on the big 8+8 die and frequently sells for less than its non-K namesake. Does this smell right, to you? Reply
Bruzzone - Friday, March 31, 2023 - link
mode_13, new on ebay around $180 to $200 any 12600_ pick your flavor. On 12600 hexa only, 65 to 117W v 125 to 150W for K_ so how about a lower priced board for 12600. For $250 to $258 at Newegg and Walmart? I just checked Newegg now $207 so maybe Walmart at $250 still has not updated their page with a competitive price. The high volume price is $111 ($223/2) however on a volume Alder i9/i7 purchase so many 12600 were likely thrown into the sales package for nothing everything below 600K is generally thrown in as sales close. The check is Alder full line Average Weighed Price $1K at $417 / 2 for a high volume procurement = $208 so at this full line product SKU buy in (i9 = 42.6%, i7 = 28.8%, i5 = 24.1%, i3 = 3.8%, Pentium = 0.27%, Celeron = 0.44%) procurement would say no to 12600 as underwater at $233 end sale and there would be a negotiation around the SKU procurement price which is likely n/c. For a tray this isn't the case it's the traditional buy 10 and get 1 free < 10%. I seriously doubt Walmart buys by the tray. I consider Neweeg and Wallmart price similar any OEM these days on competitive desktop market. Specific OEMs its different encompassing mobile and Xeon SKUs negotiated within and around the desktop buy in. mb Replymode_13h - Saturday, April 1, 2023 - link
Thanks for the reply. The $207 price on Newegg is a 3rd party seller that I don't 100% trust, and not sure if Intel will offer warranty claims from. If I'm going to take those risks, I'd rather save more money and go the ebay route. Speaking of the ebay route, I have a search query that excludes any ES (Engineering Sample) chips, which I don't trust.The reason I like the i5-12600 is that it typically has the fastest single-thread performance among the 65 W-rated models, probably due to the smaller ring bus. It's also the second cheapest model that supports ECC memory, which obviously limits me to more expensive W680 boards. Reply
Bruzzone - Saturday, April 1, 2023 - link
ebay purchase right save the money and there's the ebay guarantee ECC 12600 good single core thanks for the note. I too would stay away from engineering samples unless you know for example the mainboard design manufacturer compatibility lab tech who was originally given the field sample by the Intel FAE. mb ReplyQasar - Friday, March 31, 2023 - link
" Thanks for the info, Mike! "wow you actually believe his BS ??? with out knowing where he gets this from, its all fake and BS, and not to be believed. for all you know, it pulled it out of thin air, and made it all up.
with links to sources that arent HIS page, its just BS Reply
mode_13h - Saturday, April 1, 2023 - link
I know his posts bother you, but I don't see much harm in them. You're free to ignore them, if you don't find them useful.IMO, unless you can clearly demonstrate posts are spreading misinformation or violate other rules, then you should try to relax and accept that he's as free to post as you are. Reply
Qasar - Saturday, April 1, 2023 - link
" unless you can clearly demonstrate posts are spreading misinformation or violate other rules "my point is... no one can prove his posts either way. he never posts links to where he gets is info from, there fore, there his no way to confirm or deny any of it. he could very well be making it all up out of thin air, there fore, it could very well be misinformation and total bs he is posting. ive seen this same type of posts on other forums, and its usually fake and misinformation as the OP also cant site sources to where they got it, so its flagged as their opinion only Reply
mode_13h - Saturday, April 1, 2023 - link
Besides, he stopped replying "MB.", as I think you'd requested. That was a legit complaint and he complied. Let's try to find compromise where we can, and move on. ReplyQasar - Saturday, April 1, 2023 - link
i will.. once he posts sources for his posts. and no.. he still replies with mb at the end. Replymode_13h - Sunday, April 2, 2023 - link
Who cares about that?No, I meant that thing he used to do where he'd write one-line replies to people that were just like "mb.". You asked him not to do that, and he stopped.
Anyway, whatever you have against him, please leave me out of it. Reply
Bruzzone - Saturday, April 1, 2023 - link
Who r u quasar? I am the data scientist and the work is mine as the primary source assessing as you are aware ebay and other supply data on my tenure inside various design manufacturers and design producers basically I am a modern Storeboard. Did you ever have issue with Storeboard, Price Watch or the MDR x86 annual. Marketing and introducing PC AT board products for Orchid; Tiny Turbo, PCNet, IBM Turbo PGA 1st 3D graphics card, Excell memory cards against AST and Quadram and Intel PCEO. Next for Arche tens of PCs across 286, 386, 486 including the first PC MP based on C&T MPAX chip set. Mike goes into the processor business at Cyrix 486SLC, 487SLC, Cyrix 486 DLC, Cyrix 486S. Then to Arm 7, TDMI, associated strong ARM. Then NexGen 586, AMD K100-133, K5, codecs for C-Cube, Alpha 21164 work for Samsung and then the IDT Winchip roll out and by that time Intel associate network and I were getting fed up with each other. In 1998 enlisted by the FTC. In 2008 contracted by the USDOJ retained by Congress. In 2009 reenlisted by the FTC it's been a tough road but also a very interesting and intriguing one. Compliment or confrontation? So you work for Intel? Or are a competing analyst? mb ReplyQasar - Saturday, April 1, 2023 - link
blah blah blah blah blah. more jibber jabberi am some one who would like to SEE your sources for myself, what is wrong with that myke brazzzone ?? is it insulting that i ask for sources ? that i dont believe your posts as some of it seems like it is made up, and false ? but as i have said before, your reluctance and flat our refusal to post direct links your sources show you dont have any. you can post names and such all you want, that is meaning less. i can say my sources are X site, or Y business too, but that would prove nothing as well.
maybe you just dont like being called a fraud. who posts made up info. so, ill put it this way, IF post direct links to where you get your info from for your posts. if you do this, i will apologize for calling you what you seem to be, which is a fraud and a fake.
i guess asking for sources, is just too much to ask from some one. Reply
Bruzzone - Sunday, April 2, 2023 - link
:IF post direct links to where you get your info from for your posts. if you do this, i will apologize for calling you what you seem to be, which is a fraud and a fake."https://www.ebay.com/
https://ir.amd.com/financial-information/historica...
https://www.intc.com/financial-info/financial-resu...
https://investor.nvidia.com/financial-info/financi...
mb Reply
Qasar - Sunday, April 2, 2023 - link
see, was that so hard ?but, what about the other sources you use ? those are just the financials, what about the posts claiming the other info you post on here ? like how many products amd intel or nvidia have produced ? or things about the wafers, some of your posts have quite the bit of info in them, i wont deny that, but it would be nice to see the sources for those as well.
for now, i apologize to you. but i would like to see other sources if asked, deal ?? Reply
Bruzzone - Saturday, April 1, 2023 - link
Quasar you are more than welcome to add, to compliment, to offer point and even counter point to my observations is a constructive approach. But to detract without support you know much of my data comes from ebay that is not ebay terapeak on my own filters, same as Mercury by the way, so why do you detract {?] which is not constructive and had nothing to do with management or engineering best practice. I am trained in many management disciplines including constructive confrontation and constructive collaboration and invite you to display eitherconstructively. mb ReplyQasar - Sunday, April 2, 2023 - link
yea ok sure, i agree Myke brazzzonee, BUT i cant do any of what you suggest, unless i have the SAME info that you your self get, so i could say something completely different then what you have, and that in the end would prove squat. correct ? hence the asking for direct sources.again, you still refuse to post direct sources, is there a specific reason ? is it top secret info that no one else can have access to ? either way, i cant prove what you post is wrong or right, or made up out of thin air.
as i said, asking for sources, which i have seen MANY times on other forms being asked by one person to the other, is just too much to ask from you. and like on those forums when one person is asked for a source, and doesnt provide it, they are also called out on it, like i am doing here, but what ever. keep posting info that no one can conform, or counter, as it is unclear where you really come up with it, say it is from site X and site Y, doesnt really count, IMO, as i mentioned, other people try it, and it never resolves anything. Reply
Oxford Guy - Sunday, April 2, 2023 - link
I have a superior source so do not question me:https://www.youtube.com/watch?v=Vqc8b9nKgoo Reply
mode_13h - Sunday, April 2, 2023 - link
Rickroll! ReplyJayNor - Thursday, March 30, 2023 - link
I saw coverage at STH that said HBM will be covered in a separate presentation. I'd expect them to move to HBM3. Replynealburns5 - Thursday, March 30, 2023 - link
I believe Genoa has been misspelled as Geona a couple of times. Probably a good word to spell correctly for keyword searching. Reply