@@ServeTheHomeVideo Indeed, I still using dual X5675 and E5-2690v3 for my school remote rendering purposes, and boy even tech newbie can quickly understand each number in the SKU's mean.
@@ServeTheHomeVideo Was a tiny tiny bit confusing at first because I was expecting naming similar to the consumer line (ie 7XXX is 7th gen) but MUCH MUCH simpler than the new Xeon naming scheme. O.o
I have a build with a 2665 I built for gaming after mt dad brought an old server home, and it's much easier to decipher the X79 and X99 chipsets than just about anything newer lol. The only real rule you need to remember is v1 and v2 are X79 and v3 and v4 is X99.
Looks like Intel's marketing department is taking binning to a whole new level. I guess 10 nanometer allows them to squeeze more letters out of their product name and bring more value to their shareholders. Or they really really want to see Lisa Su succeed.
AMD: everything gets all the features, the only difference between SKUs is core counts and power limit Intel: Every single customer gets their own SKU!
It's funny because there has been a ton of research that shows this kind of ultra segmented and confusing product offering actually drives away potential clients. For the last half a decade the process for buying Intel servers/cpus for your business requires first commissioning a formal study to figure out which of their hundreds of CPU models is right for your use case. What AMD has done with just offering all the features, none of the segmentation bullcrap, and the product stack being simplified with bigger number equals more cores/tdp is smart marketing and business because it makes any client buy with confidence.
@@RazorSkinned86 Ultra segmentation is great for profit margins when you have effectively no competition. Epyc brought competition back, but there must be some powerful people at Intel that are still in denial about it.
Remember when we got an X for high performance, an E for standard, and an L for low wattage? Sure it was hard to know that an L5640 was 6 cores and an X5667 was only 4 Also, who can forget the x5697 a "2 core" 4.4Ghz processor when most of these other processors didnt even reach 3Ghz, it apparently could be unlocked to 6 cores, but could use upwards of 300w
It's pretty obvious that Intel's SKU naming is confusing on purpose. It makes them able to play fast and loose with pricing because it's all but guaranteed that the client will not be able to do their own comparisons and price matching.
Or those clients could just go choose an EPYC CPU just by looking at their wallet since AMD’s whole range has ALL of the features, only changing in core counts and TDP.
At this point they should just accept defeat and use 128bit UUIDs as names instead. At least it should be consistently bad unlike re-using the same nomenclature for completely different kinds of CPUs
@@ServeTheHomeVideo Whats the point of buying ANY of this crap, when AMD offers 64 cores and a MUCH HIGHER clock speed? Also what do you think about Tachyum Prodigy T16128-AIX CPU,is it all bull or do you think its true? It would be a game changer and solve SO MANY PROBLEMS/128cores,high clock speeds and basically as much RAM.as you need. These are legitimate questions and 40 cores is a joke/that looked to be the most cores you can get from these CPUs,which is kind of sad and like I said the clock speeds are really really low/slow. Its been a year, is cooper lake and ice lake, still the newest thing offered? Sorry I know more about 1700/Alder Lake/SOON Raptor Lake and 30 series and 40 series graphics cards. Also what do you do power supply wise/SOME of these systems COULD require 2000 watt power supplies, also 240 Volts is needed for that/120 Volts won't cut it. My current desktop setups have two ASUS TUF 3090 Ti OC Edition 24GB cards, that I've optimized with code/mainly C++, I also have a i9-12900KSs and 128GBs of 5000GHz DDR5 RAM, I plan to get i9-13900kss and 4090tis, when they come out as well, but I was trying to explore some other options/I emulate OSs/run multiple Virtual Machines at once, to test my code and another reason I would like some better specs is I would like to stream multiple instances of games, to multiple devices, locally and slow clock speeds and low core counts, is a problem for that,it sounds like intel is turning into a joke. Anyway any help, would greatly be appreciated.
Btw this crazy confusing and complex segmentation isn't by choice on the part of Intel engineers. This entire product stack from Intel is the result of every single CPU sku offered having a massive amount of so-called "dark silicon" (non-functioning, partially functioning) on the CPU dies. This isn't Intel doing "artificial segmentation" or "intentional" but the result of the silicon binning process then Intels marketing people having to figure out how to sell CPUs where none of the chips come out even 50% functional. What we are seeing with the confusing and ultra segmented product offering from Intel is due to just how bad things are with manufacturing multi core server class CPUs with a monolithic design. All these different CPUs with different ram speeds, memory channels, PCIE lanes, and random features enabled is literally due to the shameful secret at Intel that with large monolithic CPUs what we are seeing is a whole wide offering of defective barely/partially functional CPU models. AMD on the other hand doesn't have this issue because they are able to ensure all features and IO function on every CPU packaged with the only thing determined by binning being the amount of cpu cores and cache. This is enabled by AMD using an MCM process of smaller (and thus low catastrophic defect rate) chiplets with only CPU cores and L1/L2 cache on the leading edge node with features/memory/pcie/IO on separate chiplet using an extremely mature lowest defect node.
Eh... I mean sure that is part of it, but some things here should clearly have been designed to be toggle-able in the BIOS, while others could be segmented seamlessly (E.g. only "Platinum" Xeons do x and y) Plus, some kind of braindead segmentation was always the standard with Intel.
Xeon product SKUs are easy to understand, its just a 47 dimensional matrix. The confusion is due to the additional SKU letter designators being on the second and third axes of time and are imperceivable in our universe without multilateral time travel.
This is two year old, but you have not changed your quality of video production. It was on point two years ago. 😊 🤜🏼🤛🏼😎🍀🍀🍀☮️☮️☮️ Great delivery above all… that’s what has always made it watchable. 😂😂😂 Fooook how do you turn off Ai text correct… I re-read my comment and it was mangled. 2024
This is why I bought AMD Epyc. It's easy to go to the site, see what's available, and buy what I need within my price range and single or multi CPU configuration.
[11:09] - "I don't know what S stands for..." [11:23] - Shows graph with text explanation showing That the two "S" SKUs "support Intel Speed Select Technology - Performance profile 2.0 (Intel SST-PP)"
Besides being the coolest and best family of processors Intel has ever made in my opinion, all I can think of is: "Every Xeon Scalable processor is personalied"
I just want this: First letter of product line, core count, stepping / revision number, suffix. So a core i9 12900K would become a i160K or an i8-80K depending on your taste of core naming. A 5950X would be a R516X.
When you need an almost half-hourlong video by a 3rd party to explain your naming scheme--and a decent number of folks still can't grok it--you have completely, utterly failed at naming.
Half of those markings are useless, mixed with the useful ones, making the entire thing ridiculous. The most irritating thing is, P was once used for "FPGA", now they don't just changes the definition but also still provides FPGA integrated parts for OEM. If you're doing use case specialization for naming, do it without messing up all the useful ones K. BTW AMD actually did something similar, they put those useless letters on the second digit(like 7F72 7R32 7H12), but not nearly as crazy.
Whoa what a mind reader! Was just visiting Wikipedia to lookup the naming conventions or whatever on Dell Precision Workstation. Now if you could only tell me if U.3 NVMe PCIe Gen4 SSDs (Kioxia CM6 V-Series) are backward compatible with the Dell R750 U.2 slots I could retire happy!
This is Cascade Lake + r grade SKU designation break out seen in WW channel inventory; 1.5 TiB base memory = 93.05%, 2.0 M = 2.51%, 4.5 L = 2.06%, Y speed select = 0.44%, N NFV specialized = 0.05%, V virtual machine density value optimized = 0.11%, T 10 year life thermal package = 0.08%, S search optimized = 0.02%, U 1P = 0.08% and W workstation = 1.67%. Specific Skylake only breakout; Per core performance = 28.48%, 2x DRAM = 5.29%, Performance per Watt = 63.07%, Omni path attempt = 0.70%, 10 year life = 1.33%, Workstation 31/21xx_ = 1.33%. Not enough Ice lake data yet to tell. All data is WW open market channel inventory, trades and sales. Channel VARS/consultants/enterprise who has the applications ability to address specialized Xeon options? Probably waiting for business of compute applications specialization to get out into the contract consulting community associated secondary product sales. Channel does not see Platinum 92xx or Cooper lake volumes; at least not yet? There is one 5320H that shows up from tome to time and it seems snapped up pretty quick like removed from public availability so unlike confidential Christmas presents. intel produced approximately 200 M Cascade Lake + r on top of 200 M Skylake. Skylake and Cascade combined by core grade; 4C = 7.92% 6C = 4.64% 8C = 23.22% 10C = 9.65% 12C = 15.04% 14C = 4.75% 16C = 9.95% 18C = 6.9% 20C = 5.18% 22C = 1.54% 24C = 4.19% 26C = 1.88% 28C = 5.14% LCC = 45.43% or 187,716,644 units of production over four years HCC = 36.64% or 146,573,165 units XCC = 17.93% or 77,710,191 units Dividing by 18 processors per rack on power envelope 2 system control and 16 data processing; 10,095,369 fully configured racks of LCC 8,142,954 racks of HCC 3,983,900 racks of XCC All up 22,222,222 racks of servers which is the largest TAM installed market in computing history currently being targeted for GPU compute cluster refurbishment (to salvage) optimization including by FPGA and innovation replacements. Intel knows where its installed and dGPU lags? Nvidia chomping at the bit. AMD will sacrifice consumer for commercial margins to get out from in between two goliaths in consumer GPU? FPGA like Nvidia knows where it is among hyperscale / cloud; what about enterprise premise? ARM wants a cut. Innovations what a stake. On the mystical magical Intel and certain commercial analysts so said 20 M servers per year that is totally a misrepresentation concealing DCG Xeon laundering theft; 100 M units running Xeon annual / XCC 17.93% = 17,927,547 in this example 20/24/28 core processors for system configuration. 20 M servers, noooo, how about 22 M fully configured racks of servers that do not include v4/v3 enterprise legacy and v4/v3/v2 MP in this example are still sought after on the cheap for hyperscale. mb
Time for playing games is over. Intel better get their act together. AMD SKUs require less head scratching while delivering overall better performance.
It's irrelevant. There isn't a single role in SMB all the way to up Enterprise that justifies buying intel over AMD outside of adding servers to a preexisting intel VMware cluster. The architecture is archaic and loses to Milan in every way from performance, security, cost, and now even compatibility is a non issue.
I still remember when Xeons still using 4 digit SKU and a "Vx" for generation, such a simpler time back then
Even the Xeon E5 was quite a bit easier. Xeon E5-2670 V4 is easier than Xeon Platinum 8380HL
@@ServeTheHomeVideo Indeed, I still using dual X5675 and E5-2690v3 for my school remote rendering purposes, and boy even tech newbie can quickly understand each number in the SKU's mean.
@@ServeTheHomeVideo Was a tiny tiny bit confusing at first because I was expecting naming similar to the consumer line (ie 7XXX is 7th gen) but MUCH MUCH simpler than the new Xeon naming scheme. O.o
I have a build with a 2665 I built for gaming after mt dad brought an old server home, and it's much easier to decipher the X79 and X99 chipsets than just about anything newer lol. The only real rule you need to remember is v1 and v2 are X79 and v3 and v4 is X99.
@@AlpineBishop exactly! But server chipset is already kinda messy, like 5100 series or C620
Looks like Intel's marketing department is taking binning to a whole new level. I guess 10 nanometer allows them to squeeze more letters out of their product name and bring more value to their shareholders.
Or they really really want to see Lisa Su succeed.
AMD: everything gets all the features, the only difference between SKUs is core counts and power limit
Intel: Every single customer gets their own SKU!
It's funny because there has been a ton of research that shows this kind of ultra segmented and confusing product offering actually drives away potential clients. For the last half a decade the process for buying Intel servers/cpus for your business requires first commissioning a formal study to figure out which of their hundreds of CPU models is right for your use case. What AMD has done with just offering all the features, none of the segmentation bullcrap, and the product stack being simplified with bigger number equals more cores/tdp is smart marketing and business because it makes any client buy with confidence.
This
AMD EPYC also shits on these jokes of processors
@@Hogdriva Remembers when opteron was kind of the joke and intel was laughing
@@RazorSkinned86 Ultra segmentation is great for profit margins when you have effectively no competition. Epyc brought competition back, but there must be some powerful people at Intel that are still in denial about it.
When I dont know if youre talking about a cpu and monitor, maybe its time for a re-brand intel
Remember when we got an X for high performance, an E for standard, and an L for low wattage?
Sure it was hard to know that an L5640 was 6 cores and an X5667 was only 4
Also, who can forget the x5697 a "2 core" 4.4Ghz processor when most of these other processors didnt even reach 3Ghz, it apparently could be unlocked to 6 cores, but could use upwards of 300w
Intel - Making Microsoft licensing feel self-explanatory since 1968.
It's pretty obvious that Intel's SKU naming is confusing on purpose. It makes them able to play fast and loose with pricing because it's all but guaranteed that the client will not be able to do their own comparisons and price matching.
Or those clients could just go choose an EPYC CPU just by looking at their wallet since AMD’s whole range has ALL of the features, only changing in core counts and TDP.
At this point they should just accept defeat and use 128bit UUIDs as names instead. At least it should be consistently bad unlike re-using the same nomenclature for completely different kinds of CPUs
Ha!
@@ServeTheHomeVideo Whats the point of buying ANY of this crap, when AMD offers 64 cores and a MUCH HIGHER clock speed?
Also what do you think about Tachyum Prodigy T16128-AIX CPU,is it all bull or do you think its true?
It would be a game changer and solve SO MANY PROBLEMS/128cores,high clock speeds and basically as much RAM.as you need.
These are legitimate questions and 40 cores is a joke/that looked to be the most cores you can get from these CPUs,which is kind of sad and like I said the clock speeds are really really low/slow.
Its been a year, is cooper lake and ice lake, still the newest thing offered?
Sorry I know more about 1700/Alder Lake/SOON Raptor Lake and 30 series and 40 series graphics cards.
Also what do you do power supply wise/SOME of these systems COULD require 2000 watt power supplies, also 240 Volts is needed for that/120 Volts won't cut it.
My current desktop setups have two ASUS TUF 3090 Ti OC Edition 24GB cards, that I've optimized with code/mainly C++, I also have a i9-12900KSs and 128GBs of 5000GHz DDR5 RAM, I plan to get i9-13900kss and 4090tis, when they come out as well, but I was trying to explore some other options/I emulate OSs/run multiple Virtual Machines at once, to test my code and another reason I would like some better specs is I would like to stream multiple instances of games, to multiple devices, locally and slow clock speeds and low core counts, is a problem for that,it sounds like intel is turning into a joke.
Anyway any help, would greatly be appreciated.
That moment your naming convention is so convoluted that it needs a 20m+ video to explain it.
I could listen to this *Patrick* dude all day long.
If you think the naming is complicate, wait till you learn about their adaptive boost frequencies
(Techtechpotato video intensifies)
Yes. Like 90%+ of server buyers do not understand that at all.
Intel's nomenclature is more sku'd up than ever.
Agreed.
THANK YOU! WE NEED THIS SO MUCH!
Btw this crazy confusing and complex segmentation isn't by choice on the part of Intel engineers. This entire product stack from Intel is the result of every single CPU sku offered having a massive amount of so-called "dark silicon" (non-functioning, partially functioning) on the CPU dies. This isn't Intel doing "artificial segmentation" or "intentional" but the result of the silicon binning process then Intels marketing people having to figure out how to sell CPUs where none of the chips come out even 50% functional. What we are seeing with the confusing and ultra segmented product offering from Intel is due to just how bad things are with manufacturing multi core server class CPUs with a monolithic design. All these different CPUs with different ram speeds, memory channels, PCIE lanes, and random features enabled is literally due to the shameful secret at Intel that with large monolithic CPUs what we are seeing is a whole wide offering of defective barely/partially functional CPU models.
AMD on the other hand doesn't have this issue because they are able to ensure all features and IO function on every CPU packaged with the only thing determined by binning being the amount of cpu cores and cache. This is enabled by AMD using an MCM process of smaller (and thus low catastrophic defect rate) chiplets with only CPU cores and L1/L2 cache on the leading edge node with features/memory/pcie/IO on separate chiplet using an extremely mature lowest defect node.
Excellent point
Eh... I mean sure that is part of it, but some things here should clearly have been designed to be toggle-able in the BIOS, while others could be segmented seamlessly (E.g. only "Platinum" Xeons do x and y)
Plus, some kind of braindead segmentation was always the standard with Intel.
Omg the scarecrow bit, I lol'd
Feel free to use that one!
Xeon product SKUs are easy to understand, its just a 47 dimensional matrix. The confusion is due to the additional SKU letter designators being on the second and third axes of time and are imperceivable in our universe without multilateral time travel.
you should of made a 'sound board' with 'scalable' ,'gold', 'platinum' ,'8','L','N', etc... to play out whenever you named any of them :D
I probably would have gone crazy doing that.
Thank you for making this video and so many like it. This gets all the algo ups
Not a very popular one :-(
This is two year old, but you have not changed your quality of video production.
It was on point two years ago. 😊
🤜🏼🤛🏼😎🍀🍀🍀☮️☮️☮️
Great delivery above all… that’s what has always made it watchable. 😂😂😂
Fooook how do you turn off Ai text correct… I re-read my comment and it was mangled. 2024
Wish they kept the V# series going. Was already in place forever and had no need to change it.
This is why I bought AMD Epyc. It's easy to go to the site, see what's available, and buy what I need within my price range and single or multi CPU configuration.
amazing video, I was trying to figure this all out and this made it easy. Buying an epyc instead... rofl
AMD: I don't need so many SKUs.
Intel: All Xeon processor is personalized.
Just once, I would love to hear Patrick say "THIS" in the Doug DeMuro style.
I think Doug DeMuro's AND's are much crazier
[11:09] - "I don't know what S stands for..."
[11:23] - Shows graph with text explanation showing That the two "S" SKUs "support Intel Speed Select Technology - Performance profile 2.0 (Intel SST-PP)"
Pyrite == "Fools Gold", bet that flies well with the Intel marketing people. 😂
I had also suggested "Jade" at some point. Yes though, I think over a few beers they did not love my suggestions.
Besides being the coolest and best family of processors Intel has ever made in my opinion, all I can think of is:
"Every Xeon Scalable processor is personalied"
I hope that's a joke
Intel the master of confusing SKU naming
What I realized from this vid is that I'm probably better of buying an Epyc... Lol
I just want this: First letter of product line, core count, stepping / revision number, suffix. So a core i9 12900K would become a i160K or an i8-80K depending on your taste of core naming. A 5950X would be a R516X.
When you need an almost half-hourlong video by a 3rd party to explain your naming scheme--and a decent number of folks still can't grok it--you have completely, utterly failed at naming.
Half of those markings are useless, mixed with the useful ones, making the entire thing ridiculous. The most irritating thing is, P was once used for "FPGA", now they don't just changes the definition but also still provides FPGA integrated parts for OEM. If you're doing use case specialization for naming, do it without messing up all the useful ones K.
BTW AMD actually did something similar, they put those useless letters on the second digit(like 7F72 7R32 7H12), but not nearly as crazy.
Should I be worried that I exactly knew what the SKU's in the introduction mean?
Whoa what a mind reader! Was just visiting Wikipedia to lookup the naming conventions or whatever on Dell Precision Workstation.
Now if you could only tell me if U.3 NVMe PCIe Gen4 SSDs (Kioxia CM6 V-Series) are backward compatible with the Dell R750 U.2 slots I could retire happy!
I have not tried it, but I believe that Dell sells CM6 SSDs as OEM, right?
What about Sspec or model numbers on CPU,s servers SR2N7 or SR19S.what that mean all this numbers and digits? Thank you.
This is so helpful
WOW FINALLY!
This is Cascade Lake + r grade SKU designation break out seen in WW channel inventory; 1.5 TiB base memory = 93.05%, 2.0 M = 2.51%, 4.5 L = 2.06%, Y speed select = 0.44%, N NFV specialized = 0.05%, V virtual machine density value optimized = 0.11%, T 10 year life thermal package = 0.08%, S search optimized = 0.02%, U 1P = 0.08% and W workstation = 1.67%.
Specific Skylake only breakout; Per core performance = 28.48%, 2x DRAM = 5.29%, Performance per Watt = 63.07%, Omni path attempt = 0.70%, 10 year life = 1.33%, Workstation 31/21xx_ = 1.33%.
Not enough Ice lake data yet to tell. All data is WW open market channel inventory, trades and sales.
Channel VARS/consultants/enterprise who has the applications ability to address specialized Xeon options? Probably waiting for business of compute applications specialization to get out into the contract consulting community associated secondary product sales.
Channel does not see Platinum 92xx or Cooper lake volumes; at least not yet? There is one 5320H that shows up from tome to time and it seems snapped up pretty quick like removed from public availability so unlike confidential Christmas presents.
intel produced approximately 200 M Cascade Lake + r on top of 200 M Skylake.
Skylake and Cascade combined by core grade;
4C = 7.92%
6C = 4.64%
8C = 23.22%
10C = 9.65%
12C = 15.04%
14C = 4.75%
16C = 9.95%
18C = 6.9%
20C = 5.18%
22C = 1.54%
24C = 4.19%
26C = 1.88%
28C = 5.14%
LCC = 45.43% or 187,716,644 units of production over four years
HCC = 36.64% or 146,573,165 units
XCC = 17.93% or 77,710,191 units
Dividing by 18 processors per rack on power envelope 2 system control and 16 data processing;
10,095,369 fully configured racks of LCC
8,142,954 racks of HCC
3,983,900 racks of XCC
All up 22,222,222 racks of servers which is the largest TAM installed market in computing history currently being targeted for GPU compute cluster refurbishment (to salvage) optimization including by FPGA and innovation replacements. Intel knows where its installed and dGPU lags? Nvidia chomping at the bit. AMD will sacrifice consumer for commercial margins to get out from in between two goliaths in consumer GPU? FPGA like Nvidia knows where it is among hyperscale / cloud; what about enterprise premise? ARM wants a cut. Innovations what a stake.
On the mystical magical Intel and certain commercial analysts so said 20 M servers per year that is totally a misrepresentation concealing DCG Xeon laundering theft; 100 M units running Xeon annual / XCC 17.93% = 17,927,547 in this example 20/24/28 core processors for system configuration.
20 M servers, noooo, how about 22 M fully configured racks of servers that do not include v4/v3 enterprise legacy and v4/v3/v2 MP in this example are still sought after on the cheap for hyperscale.
mb
Time for playing games is over. Intel better get their act together. AMD SKUs require less head scratching while delivering overall better performance.
upvote for the dad jokes
At least someone saw this :-)
It would be nice to buy a XEON with a single thread performance around 3000 passmark.
Patrick, like all the rest of us trying to make sense out of life's Intel CPU naming conundrum, 9:44.
Now we know where all of Intel's marketing brain power goes: trying to keep all those SKUs straight!
I'm getting my slide rule out of the cupboard. I know where I am with it. This is all just too confusing.
no even Intel has a naming overview as comprehensive as this one
intel has more lakes than minnestota
If only Mikrotik ran the world......or was at least responsible for naming stuff.
Just have a generation number followed by core count. If you need a letter afterward to differentiate SKUs, keep it simple and straightforward.
Let's make even more letter monsters XD Like Intel Xeon Platium 9843HQGLOPBX :D How's that? :D
Intel markets CPUs the same way they move to smaller die sizes.
That Y sounds like you're buying a Mercedes and they're locking it down to a Honda Civic because you didn't pay them enough
set -3db
Xeon naming is completely baffling. It doesn’t seem like Intel is making it easier at all
It's irrelevant. There isn't a single role in SMB all the way to up Enterprise that justifies buying intel over AMD outside of adding servers to a preexisting intel VMware cluster. The architecture is archaic and loses to Milan in every way from performance, security, cost, and now even compatibility is a non issue.
Wait a second… so… Intel’s highest end 40 core CPU only has 60MB or L3 cache? Is this a joke?
Hello :p
Amd all the way
OMG
Intel is compensating for bad performance with marketing. Which they are very goo... no, wait..... very bad at. :/
FIRST
I think someone may have beaten you by a few seconds this time!
@@ServeTheHomeVideo thanks for reading the comments keep up your good work