Sunday, April 30th 2023

Intel to Introduce Core Ultra Brand Extension with "Meteor Lake," iGPU Packs 128 EU

Intel is planning a major change in its client processor brand extensions with its next-generation mobile processors codenamed "Meteor Lake." The company is working to introduce the new Core Ultra brand extensions, where "Ultra" replaces the "i" in extensions such as i3, i5, i7, and i9 in some processor models. An example of such a brand extension would be the "Core Ultra 5 1003H." Ashes of the Singularity benchmark leaks of the processors surfaced on social media.

The benchmark also detects 128 EU (1,024 unified shaders) for the iGPU powering "Meteor Lake." If true, this iGPU could offer performance that's in the league of an Arc A380 discrete GPU, with some performance lost to the shared memory setup compared to the A380 with its dedicated graphics memory. The iGPU clock speed is detected to be 2.10 GHz, and having 4 MB of L2 cache, the last-level cache local to the Graphics Tile. The detection string for the iGPU as reported by its OpenCL ICD reads "Intel(R) Graphics i gfx-driver-ci-master-13736 DCH RI (1024S 128C SM3.0 2.1GHz, 4MB L2, 12.7GB)."
Source: BenchLeaks
Add your own comment

48 Comments on Intel to Introduce Core Ultra Brand Extension with "Meteor Lake," iGPU Packs 128 EU

#26
Daven
I assume the purpose of the name change is to try and hide the lower core counts in ML versus RL. At least the lower counts at launch.
Posted on Reply
#27
Wirko
Maybe Apple finally sued Intel for the use of the letter "i", in lower case, with one round or square dot above.
Posted on Reply
#28
john_
It's funny reading the objections here about Celeron be a junk, when half of the site's posts are "AMD is junk", "Radeon is junk", "FSR is junk".

Anyway, Celeron is the lowest of the CPUs from Intel and when compared with everything else Intel, it will look bad. Especially those models based on Atom cores only. But it is not Celeron's mistake when companies take this CPU and throw it on a 15,6'' chassis and pretend they have a competing option to Ryzen 3 or i3 based models.
Posted on Reply
#29
ExcuseMeWtf
Tek-CheckThe second quote is not mine.
It wasn't a good joke in this occasion, but you tried.
I also misfire with jokes. It doesn't always work.
Yes, it def didn't land well with the crowd, so I take that egg on my face.
Posted on Reply
#30
TheinsanegamerN
john_It's funny reading the objections here about Celeron be a junk, when half of the site's posts are "AMD is junk", "Radeon is junk", "FSR is junk".
"guy what about our lord and saviour AMD? It have its blood for YOUR sins!" :laugh: :roll: :laugh:
john_Anyway, Celeron is the lowest of the CPUs from Intel and when compared with everything else Intel, it will look bad. Especially those models based on Atom cores only. But it is not Celeron's mistake when companies take this CPU and throw it on a 15,6'' chassis and pretend they have a competing option to Ryzen 3 or i3 based models.
Celerons are just plain bad. Pentiums were significantly faster, barely drew any more power, and were maybe $5-10 more. I've never seen someone build a system with one of those abominations. Pentiums? Yeah, those pop up from time to time. But not celerons.
Posted on Reply
#31
P4-630
Intel 5 Ti
Intel 5 Super
Intel 5 GTX
Intel 5 Turbo

Nah.. Just leave it at "i5" intel....
Posted on Reply
#32
AusWolf
When you thought Intel's naming system couldn't be a bigger clusterfuck.
Posted on Reply
#33
ChettManly
Once again Intel proves they have the worst naming team in the industry.

If I was them I'd replace the Core brand with Meteor Lake. They are moving to MCM, they are changing the architecture dramatically. Its been many years. Spice things up with a new name!

Then implement a simple system where the first number is the core count, the second number is the year. Then K if high performance.

Intel Meteor 2223K
Intel Meteor 1623K
Intel Meteor 823

Higher is always better, you can tell the launch year so you know if you are getting an old one, you can designate K/KS... to denote variations of the same core count. And because it uses the year its solid through 2122.
Posted on Reply
#34
AusWolf
ChettManlyOnce again Intel proves they have the worst naming team in the industry.

If I was them I'd replace the Core brand with Meteor Lake. They are moving to MCM, they are changing the architecture dramatically. Its been many years. Spice things up with a new name!

Then implement a simple system where the first number is the core count, the second number is the year. Then K if high performance.

Intel Meteor 2223K
Intel Meteor 1623K
Intel Meteor 823

Higher is always better, you can tell the launch year so you know if you are getting an old one, you can designate K/KS... to denote variations of the same core count. And because it uses the year its solid through 2122.
That's too logical. Why not just throw around random numbers and letters in a meaningless order, like i7 1165G7 for example... oh wait... :D
Posted on Reply
#36
AnotherReader
The GPU size is impressive, but it would need more cache to mitigate memory bandwidth based bottlenecks. AMD should take note; Intel will have a 33% larger IGP and more cache for it.
Posted on Reply
#37
ymdhis
Can't wait until Intel Atom gets rebranded to INTEL SUPER PRO or something.
Posted on Reply
#38
Dr. Dro
john_It's funny reading the objections here about Celeron be a junk, when half of the site's posts are "AMD is junk", "Radeon is junk", "FSR is junk".
All four are objectively and demonstrably true :roll:

*this is satire
ymdhisCan't wait until Intel Atom gets rebranded to INTEL SUPER PRO or something.
That will deeply anger Jensen Huang who might raise prices in RTX cards even more due to his bad mood. Please no! Oh God no! Hahaha.
Posted on Reply
#39
Wirko
ChettManlyThen implement a simple system where the first number is
Thats characteristic of some Russian processors. Looks like Russians haven't yet fully mastered the trade of modern marketing.
Posted on Reply
#40
ymdhis
Dr. DroThat will deeply anger Jensen Huang who might raise prices in RTX cards even more due to his bad mood. Please no! Oh God no! Hahaha.
With igpus becoming so powerful lately, who cares? There are even rumors of an upcoming (in 2024) APU that uses a 40CU chiplet rdna3 gpu, which would make anything south of a geforce 3070 obsolete. Unfortunately it is rumoured to be a halo product.
Posted on Reply
#41
Dr. Dro
ymdhisWith igpus becoming so powerful lately, who cares? There are even rumors of an upcoming (in 2024) APU that uses a 40CU chiplet rdna3 gpu, which would make anything south of a geforce 3070 obsolete. Unfortunately it is rumoured to be a halo product.
That was a very bad joke, but you definitely have my interest. Halo product or not, imagine a computer in the form factor of a Mac mini with those capabilities? Man, I want one. :toast:
Posted on Reply
#42
Wirko
mechtech128 European Unions??? ;)
But they partially overlap.
ymdhisWith igpus becoming so powerful lately, who cares? There are even rumors of an upcoming (in 2024) APU that uses a 40CU chiplet rdna3 gpu, which would make anything south of a geforce 3070 obsolete. Unfortunately it is rumoured to be a halo product.
Think about the memory bandwidth needed. Would four channels of LPDDR5-7500 suffice?
Posted on Reply
#43
Lew Zealand
ymdhisWith igpus becoming so powerful lately, who cares? There are even rumors of an upcoming (in 2024) APU that uses a 40CU chiplet rdna3 gpu, which would make anything south of a geforce 3070 obsolete. Unfortunately it is rumoured to be a halo product.
Going from their best iGPU matching a 6400 in CU count (12) to one matching a 6700XT in one year? I'd love to see it but frankly matching the 6600 (28 CUs) would get them well over the line to fully usable GPU for gaming without going budget overkill, except:
WirkoBut they partially overlap.


Think about the memory bandwidth needed. Would four channels of LPDDR5-7500 suffice?
2 channels of LPDDR5-7500 (94 GB/sec) comes up well short of the 6400's measly 128 GB/s. So even some currently never-marketed 256-bit 4-channel LPDDR5-7500 only get you 186 MB/sec, coming up well short of the 6600 (224GB/s) but kinda close enough to supply those 28CUs with some reasonable bandwidth. Maybe increasing the L3 cache to 64-128MB could help?

But 40CUs would be wasted CUs/Wattage/Die Space on this.
Posted on Reply
#44
AusWolf
Lew ZealandGoing from their best iGPU matching a 6400 in CU count (12) to one matching a 6700XT in one year? I'd love to see it but frankly matching the 6600 (28 CUs) would get them well over the line to fully usable GPU for gaming without going budget overkill, except:



2 channels of LPDDR5-7500 (94 GB/sec) comes up well short of the 6400's measly 128 GB/s. So even some currently never-marketed 256-bit 4-channel LPDDR5-7500 only get you 186 MB/sec, coming up well short of the 6600 (224GB/s) but kinda close enough to supply those 28CUs with some reasonable bandwidth. Maybe increasing the L3 cache to 64-128MB could help?

But 40CUs would be wasted CUs/Wattage/Die Space on this.
Not to mention die size. The 40 CU Navi 22 is 335 mm2 large at 7 nm. Even at 5 nm, it would be around 240 mm2. There's no room for that on a CPU. Whoever thought up a 40 CU APU was obviously joking.
Posted on Reply
#45
Minus Infinity
Tek-CheckThis is nonsense. Celeron CPUs power millions of low power useful devices, for example my NAS Asustor. Great media support with QuickSync too.
Well a Celeron based on the new Crestmont or even better Skymont E-cores would actually be a pretty strong cpu as both are large improvements over Gracemont with 15% IPC uplifts and other changes.

Would be very powerful for NAS, small home theatre box, and ultrathin laptops etc.
Posted on Reply
#46
ymdhis
AusWolfNot to mention die size. The 40 CU Navi 22 is 335 mm2 large at 7 nm. Even at 5 nm, it would be around 240 mm2. There's no room for that on a CPU. Whoever thought up a 40 CU APU was obviously joking.
GPU is supposed to be on a separate chiplet with 96MB cache and 256bit lpddr5x controller. Unfortunately it is also rumored to be a laptop apple M chip competitor. So even if it's not just a fake rumor, it's probably nothing for desktops for the next 1-2 years.
Posted on Reply
#47
Lew Zealand
ymdhisGPU is supposed to be on a separate chiplet with 96MB cache and 256bit lpddr5x controller. Unfortunately it is also rumored to be a laptop apple M chip competitor. So even if it's not just a fake rumor, it's probably nothing for desktops for the next 1-2 years.
Could you point me to that rumor, it sounds cool especially the 256-Bit lpddr5x memory system. That's unusual to the point of being unique, at least for the past decade or so.
Posted on Reply
#48
ymdhis
Lew ZealandCould you point me to that rumor, it sounds cool especially the 256-Bit lpddr5x memory system. That's unusual to the point of being unique, at least for the past decade or so.
Just google "AMD Strix Halo".
Posted on Reply
Add your own comment
Jun 3rd, 2024 07:26 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts