• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Imagination launches IMG RTXM-2200 - its first real-time embedded RISC-V CPU

TheLostSwede

News Editor
Joined
Nov 11, 2004
Messages
17,597 (2.41/day)
Location
Sweden
System Name Overlord Mk MLI
Processor AMD Ryzen 7 7800X3D
Motherboard Gigabyte X670E Aorus Master
Cooling Noctua NH-D15 SE with offsets
Memory 32GB Team T-Create Expert DDR5 6000 MHz @ CL30-34-34-68
Video Card(s) Gainward GeForce RTX 4080 Phantom GS
Storage 1TB Solidigm P44 Pro, 2 TB Corsair MP600 Pro, 2TB Kingston KC3000
Display(s) Acer XV272K LVbmiipruzx 4K@160Hz
Case Fractal Design Torrent Compact
Audio Device(s) Corsair Virtuoso SE
Power Supply be quiet! Pure Power 12 M 850 W
Mouse Logitech G502 Lightspeed
Keyboard Corsair K70 Max
Software Windows 10 Pro
Benchmark Scores https://valid.x86.fr/yfsd9w
Imagination Technologies announces IMG RTXM-2200, its first real-time embedded RISC-V CPU, a highly scalable, feature-rich, 32-bit embedded solution with a flexible design for a wide range of high-volume devices. IMG RTXM-2200 is one of the first commercial cores in Imagination's Catapult CPU family, previously announced in December 2021. Accelerating the expansion of its RISC-V offering, Imagination's IMG RTXM-2200 can be integrated into complex SoCs for a range of applications including networking solutions, packet management, storage controllers, and sensor management for AI cameras and smart metering. Together with its market-leading GPU and AI accelerator IP, Imagination's new CPU cores offer customers access to innovative heterogeneous solutions.

This real-time embedded core features up to 128 KB of tightly coupled memories (both instruction and data) for deterministic response and Level 1 cache sizes of up to 128 KB for robust performance. The new CPU offers a range of floating-point formats including single-precision and bfloat16. The latter enables manufacturers to deploy AI applications through this core without the need for an additional chip. This reduces silicon area, for a cost-effective and optimised design in AI cameras and smart metering applications.




Chris Porthouse, Chief Product Officer, Imag128 KBon, says: "We are excited to announce IMG RTXM-2200, our first real-time RISC-V CPU, which underpins our continued commitment to driving growth for the RISC-V ecosystem. Imagination now has a unique and broad portfolio of compute IP, including GPU, AI accelerators, and now CPU, designed for this innovative architecture."

Calista Redmond, CEO, RISC-V International, says: "RISC-V International is excited to witness Imagination maintaining its momentum by launching its first licensable 32-bit embedded real-time processor, since the announcement of its entry into the RISC-V CPU IP market. This underlines the expansion of the RISC-V architecture as developers take advantage of this flexible technology to address multiple market requirements. We are pleased to see Imagination's commitment to drive ecosystem growth with the delivery of its SDK and the support of open-source operating systems."

James Hodgson, Principal Analyst, ABI Research, says: "As automakers look to take advantage of the potential innovation and time to market advantages of RISC-V, they need an IP partner with proven ability to deliver in the automotive market, meeting the stringent power and safety requirements of the vertical. With the launch of IMG RTXM-2200, Imagination lays the foundation for future automotive RISC-V CPUs for this market, rounding out its portfolio of GPU and AI acceleration IP. This positions Imagination well to underpin the heterogeneous compute that will prove essential to meeting the connected infotainment and autonomous vehicle ambitions of the automotive industry going forward."

Steve Leibson, Principal Analyst, TIRIAS Research, says: "It's great to see an established IP provider such as Imagination jumping onto and giving even more momentum to the RISC-V bandwagon. With an initial emphasis on real-time embedded applications, the configurable IMG RTXM-2200 CPU, the company's first commercial RISC-V core in the previously announced Catapult family, should appeal to SoC design teams working in IoT and other embedded markets."

Leveraging Imagination's 20 years plus of experience in delivering complex IP solutions, the new CPU is supported by the rapidly expanding open-standard RISC-V ecosystem, which continues to shake up the embedded CPU industry through the democratisation of compute, unlocking innovative and open technological development for companies at all levels.

Software
Imagination real-time embedded cores come with a fully capable software SDK and tools package at launch. They work out of the box across multiple platforms to help develop and support the RISC-V ecosystem.

Shreyas Derashri, VP of Compute, Imagination, says: "For more than 20 years Imagination has been creating market-leading GPU, AI accelerator, and CPU IP. This puts us in a unique position where we can offer our customers a robust and competitive choice when it comes to next-generation SoC development. The IMG RTXM-2200 RISC-V CPU brings together years of IP design experience and comprehensive software support to enable instant and easy access for developers, further enhancing our heterogeneous offering."

The Catapult Studio Integrated Development Environment (IDE) is a modern, cutting-edge IDE based on Visual Studio code providing a familiar environment that is better to use for up-to-date developers and designers. The IDE;

runs on Windows, Ubuntu, CentOS, and macOS;
offers full Linux support (including reference bootloaders, kernel, and filesystem);
is compatible with gem5 software, unlocking simulation environments for enhanced power and energy-efficiency testing.
The IMG RTXM-2200 CPU is part of Imagination's heterogeneous compute solution, offering full hardware, software, and debug support for Imagination IP - complementing its market-leading GPU, AI accelerator, and Ethernet Packet Processor (EPP) cores.


View at TechPowerUp Main Site | Source
 
Joined
Nov 4, 2005
Messages
11,980 (1.72/day)
System Name Compy 386
Processor 7800X3D
Motherboard Asus
Cooling Air for now.....
Memory 64 GB DDR5 6400Mhz
Video Card(s) 7900XTX 310 Merc
Storage Samsung 990 2TB, 2 SP 2TB SSDs, 24TB Enterprise drives
Display(s) 55" Samsung 4K HDR
Audio Device(s) ATI HDMI
Mouse Logitech MX518
Keyboard Razer
Software A lot.
Benchmark Scores Its fast. Enough.
Another ARM implementation for…….

Fridge, Washer, smart home, Roku, Firestick what exact application?

That will undoubtedly have a small market presence for its astounding AI ability GPU that supports codecs for a few current services that will become obsolete in the next 3 years.


Not trying to be a downer, but I foresee a market crash on these products as a few bigger players in the AI market finally settle on a industry standard that renders most other devices obsolete by virtue of adoption and performance.
 
Joined
Nov 4, 2005
Messages
11,980 (1.72/day)
System Name Compy 386
Processor 7800X3D
Motherboard Asus
Cooling Air for now.....
Memory 64 GB DDR5 6400Mhz
Video Card(s) 7900XTX 310 Merc
Storage Samsung 990 2TB, 2 SP 2TB SSDs, 24TB Enterprise drives
Display(s) 55" Samsung 4K HDR
Audio Device(s) ATI HDMI
Mouse Logitech MX518
Keyboard Razer
Software A lot.
Benchmark Scores Its fast. Enough.
They lost me at 32bit...
Exactly, in current homogeneous systems the idea is to allow access to larger amounts of memory so there is less overhead. This implementation sounds good for a 2005 system.

Maybe it will be able to AI when veggies go bad by comparing pixel colors, or some photo app.
 

TheLostSwede

News Editor
Joined
Nov 11, 2004
Messages
17,597 (2.41/day)
Location
Sweden
System Name Overlord Mk MLI
Processor AMD Ryzen 7 7800X3D
Motherboard Gigabyte X670E Aorus Master
Cooling Noctua NH-D15 SE with offsets
Memory 32GB Team T-Create Expert DDR5 6000 MHz @ CL30-34-34-68
Video Card(s) Gainward GeForce RTX 4080 Phantom GS
Storage 1TB Solidigm P44 Pro, 2 TB Corsair MP600 Pro, 2TB Kingston KC3000
Display(s) Acer XV272K LVbmiipruzx 4K@160Hz
Case Fractal Design Torrent Compact
Audio Device(s) Corsair Virtuoso SE
Power Supply be quiet! Pure Power 12 M 850 W
Mouse Logitech G502 Lightspeed
Keyboard Corsair K70 Max
Software Windows 10 Pro
Benchmark Scores https://valid.x86.fr/yfsd9w
Another ARM implementation for…….
RISC-V, not arm.
Fridge, Washer, smart home, Roku, Firestick what exact application?
More like SSD controllers, cars and other things that need a real-time chip that runs an RTOS.
That will undoubtedly have a small market presence for its astounding AI ability GPU that supports codecs for a few current services that will become obsolete in the next 3 years.
It's just IP.
Not trying to be a downer, but I foresee a market crash on these products as a few bigger players in the AI market finally settle on a industry standard that renders most other devices obsolete by virtue of adoption and performance.
See above. Someone has to license the IP and make a product around it. But it'll most likely appear in some car or even airplane chips down the road.

They lost me at 32bit...
It's more of an MCU rather than a CPU and it will need an RTOS, so it's not for for typical computer applications.
 
Joined
Jul 10, 2017
Messages
2,671 (0.99/day)
RISC-V, not arm.

More like SSD controllers, cars and other things that need a real-time chip that runs an RTOS.

It's just IP.

See above. Someone has to license the IP and make a product around it. But it'll most likely appear in some car or even airplane chips down the road.


It's more of an MCU rather than a CPU and it will need an RTOS, so it's not for for typical computer applications.
5G RRU and BBU can already utilize 64-bit. What are those for? EV-DO?
 
Joined
Aug 31, 2021
Messages
21 (0.02/day)
Processor Ryzen 9 5900X
Motherboard Gigabyte Aorus B550i pro ax
Cooling Noctua NH-D15 chromax.black (with original fans)
Memory G.skill 32GB 3200MHz
Video Card(s) 4060ti 16GB
Storage 1TB Samsung PM9A1, 256GB Toshiba pcie3 (from a laptop), 512GB crucial MX500, 2x 1TB Toshiba HDD 2.5
Display(s) Mateview GT 34''
Case Thermaltake the tower 100, 1 noctua NF-A14 ippc3000 on top, 2 x Arctic F14
Power Supply Seasonic focus-GX 750W
Software Windows 10, Ubuntu when needed.
Joined
Aug 6, 2020
Messages
729 (0.46/day)
RV32 is a 32-bit instruction with with 32-bit registers. IDK about its memory interface width but we're talking about maximum address space.
Yeah, its going to be a long time before most risc 5 custom cores graduate up to 64-bit!

it's not lack of 64-bit in the standard, it's the required baby-steps while understanding a new CPU arch.
 
Joined
Nov 4, 2005
Messages
11,980 (1.72/day)
System Name Compy 386
Processor 7800X3D
Motherboard Asus
Cooling Air for now.....
Memory 64 GB DDR5 6400Mhz
Video Card(s) 7900XTX 310 Merc
Storage Samsung 990 2TB, 2 SP 2TB SSDs, 24TB Enterprise drives
Display(s) 55" Samsung 4K HDR
Audio Device(s) ATI HDMI
Mouse Logitech MX518
Keyboard Razer
Software A lot.
Benchmark Scores Its fast. Enough.
RISC-V, not arm.

More like SSD controllers, cars and other things that need a real-time chip that runs an RTOS.

It's just IP.

See above. Someone has to license the IP and make a product around it. But it'll most likely appear in some car or even airplane chips down the road.


It's more of an MCU rather than a CPU and it will need an RTOS, so it's not for for typical computer applications.
RISC has lost its way in the woods of application specific acceleration, making them ASIC and kind of RISC, but even that nomenclature is misleading as RISC was fundamentally closer to metal on Power PC architecture, which made it fast on older hardware resources, but a modern day architecture would emulate a whole
RISC system with ease.
 
Joined
Jul 10, 2017
Messages
2,671 (0.99/day)
RISC has lost its way in the woods of application specific acceleration, making them ASIC and kind of RISC, but even that nomenclature is misleading as RISC was fundamentally closer to metal on Power PC architecture, which made it fast on older hardware resources, but a modern day architecture would emulate a whole
RISC system with ease.
In general, I agree. My only problem is RTOS.

Where for example the 5G gNB scheduler needs to make sub-1ms decisions for dozens of users simultaneously, while also taking into account hundreds of constantly changing variables. You need a halluva OS and chips to run it .
 
Joined
Jul 16, 2014
Messages
8,198 (2.17/day)
Location
SE Michigan
System Name Dumbass
Processor AMD Ryzen 7800X3D
Motherboard ASUS TUF gaming B650
Cooling Artic Liquid Freezer 2 - 420mm
Memory G.Skill Sniper 32gb DDR5 6000
Video Card(s) GreenTeam 4070 ti super 16gb
Storage Samsung EVO 500gb & 1Tb, 2tb HDD, 500gb WD Black
Display(s) 1x Nixeus NX_EDG27, 2x Dell S2440L (16:9)
Case Phanteks Enthoo Primo w/8 140mm SP Fans
Audio Device(s) onboard (realtek?) - SPKRS:Logitech Z623 200w 2.1
Power Supply Corsair HX1000i
Mouse Steeseries Esports Wireless
Keyboard Corsair K100
Software windows 10 H
Benchmark Scores https://i.imgur.com/aoz3vWY.jpg?2
RV32 is a 32-bit instruction with with 32-bit registers. IDK about its memory interface width but we're talking about maximum address space.
I think the max amount should be the same limit that windows had when it was 32bit, 4Gb limit.
 
Joined
Jul 10, 2017
Messages
2,671 (0.99/day)
I think the max amount should be the same limit that windows had when it was 32bit, 4Gb limit.
Windows ran on x86 ISA at that time. While x86 has 32-bit general purpose registers, it's a CISC after all, meaning it has some fancy extensions.
One such is the SSE. SSE for example works with 128-bit registers, so... I hope you get the idea.
 
Joined
Aug 31, 2021
Messages
21 (0.02/day)
Processor Ryzen 9 5900X
Motherboard Gigabyte Aorus B550i pro ax
Cooling Noctua NH-D15 chromax.black (with original fans)
Memory G.skill 32GB 3200MHz
Video Card(s) 4060ti 16GB
Storage 1TB Samsung PM9A1, 256GB Toshiba pcie3 (from a laptop), 512GB crucial MX500, 2x 1TB Toshiba HDD 2.5
Display(s) Mateview GT 34''
Case Thermaltake the tower 100, 1 noctua NF-A14 ippc3000 on top, 2 x Arctic F14
Power Supply Seasonic focus-GX 750W
Software Windows 10, Ubuntu when needed.
RV32 is a 32-bit instruction with with 32-bit registers. IDK about its memory interface width but we're talking about maximum address space.
ok, but who cares about address space if it should be application specific, 4 GB are more than enough, I think this will compete with microcontrollers like the ones from infineon or ST. if you are concerned about AI consider that AI models are not that memory hungry when talking about inference and probably there will be instructions to load vectors from memory. then you said 32 bit registers, but we don't know how many nor we know if there are special registers. that was my point.
 

TheLostSwede

News Editor
Joined
Nov 11, 2004
Messages
17,597 (2.41/day)
Location
Sweden
System Name Overlord Mk MLI
Processor AMD Ryzen 7 7800X3D
Motherboard Gigabyte X670E Aorus Master
Cooling Noctua NH-D15 SE with offsets
Memory 32GB Team T-Create Expert DDR5 6000 MHz @ CL30-34-34-68
Video Card(s) Gainward GeForce RTX 4080 Phantom GS
Storage 1TB Solidigm P44 Pro, 2 TB Corsair MP600 Pro, 2TB Kingston KC3000
Display(s) Acer XV272K LVbmiipruzx 4K@160Hz
Case Fractal Design Torrent Compact
Audio Device(s) Corsair Virtuoso SE
Power Supply be quiet! Pure Power 12 M 850 W
Mouse Logitech G502 Lightspeed
Keyboard Corsair K70 Max
Software Windows 10 Pro
Benchmark Scores https://valid.x86.fr/yfsd9w
RISC has lost its way in the woods of application specific acceleration, making them ASIC and kind of RISC, but even that nomenclature is misleading as RISC was fundamentally closer to metal on Power PC architecture, which made it fast on older hardware resources, but a modern day architecture would emulate a whole
RISC system with ease.
Which is why this is RISC-V, which is not the same as the previous RISC based chips.
 
Top