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

Gcn = hd7000

Joined
Jul 10, 2010
Messages
1,229 (0.25/day)
Location
USA, Arizona
System Name SolarwindMobile
Processor AMD FX-9800P RADEON R7, 12 COMPUTE CORES 4C+8G
Motherboard Acer Wasp_BR
Cooling It's Copper.
Memory 2 x 8GB SK Hynix/HMA41GS6AFR8N-TF
Video Card(s) ATI/AMD Radeon R7 Series (Bristol Ridge FP4) [ACER]
Storage TOSHIBA MQ01ABD100 1TB + KINGSTON RBU-SNS8152S3128GG2 128 GB
Display(s) ViewSonic XG2401 SERIES
Case Acer Aspire E5-553G
Audio Device(s) Realtek ALC255
Power Supply PANASONIC AS16A5K
Mouse SteelSeries Rival
Keyboard Ducky Channel Shine 3
Software Windows 10 Home 64-bit (Version 1607, Build 14393.969)
http://www.nordichardware.com/news/...slands-will-be-an-all-fresh-architecture.html

It's official based on this source

:toast:

What you need to know:

It throws out the VLIW design and replace it with what it calls a Compute Unit. The new units will sport several elements from different design concepts, like being able to execute a larger number of instructions and a MIMD design; Multiple Instruction stream, Multiple Data stream. Simultaneous Multi-Threading (SMT) is available in the architecture and many will recognize this from Intel's processors where it has enabled eached core to handle 2 threads, also known as HyperThreading. It has also implemented a genuine cache structure, something NVIDIA has used for a long time.

Its VLIW design offers really high FLOPS if you compare HD 6970 to GTX 580, AMD's alternative is almost twice as capable with 2.7 TeraFLOP, but in most cases GTX 580 wins in parallel calculations where GPGPU triumphs. The reason for this is a design that is much simpler to program for, but also NVIDIA's own programming language CUDA. What AMD does with its Compute Unit is to make this bit a lot batter and simpler to work with for developers, which has complained for long about how hard it is to program for AMD's GPUs.

Another big news is that architecture will support x86-64 memory addressing, which will make it possible for both CPU ang GPU to share memory. This will make it faster to load new textures in games and allow developers to get access to "really virtualized memory" according to AMD.

The new support for memory addressing will come to great use with its coming APUs. Llano and Ontario acts just like before when it comes to memory where it has to address a certain portion (128 MB, 256 MB, 512 MB etc.) in BIOS for the GPU and thus lose a bit of the memory regularly available to the processor. Instead both CPU and GPU can share all of the memory in the system or use it all when needed.

The new architecture also has full support for C, C++ and other high leverl languages to make it even easier for developers to use the power of the GPU and the list of support programming languages will grow with every generation says AMD.

http://en.wikipedia.org/wiki/Southern_Islands_(GPU_family)
 
Last edited:
Joined
Jul 19, 2006
Messages
43,585 (6.74/day)
Processor AMD Ryzen 7 7800X3D
Motherboard ASUS TUF x670e
Cooling EK AIO 360. Phantek T30 fans.
Memory 32GB G.Skill 6000Mhz
Video Card(s) Asus RTX 4090
Storage WD m.2
Display(s) LG C2 Evo OLED 42"
Case Lian Li PC 011 Dynamic Evo
Audio Device(s) Topping E70 DAC, SMSL SP200 Headphone Amp.
Power Supply FSP Hydro Ti PRO 1000W
Mouse Razer Basilisk V3 Pro
Keyboard Tester84
Software Windows 11
Joined
Jul 10, 2010
Messages
1,229 (0.25/day)
Location
USA, Arizona
System Name SolarwindMobile
Processor AMD FX-9800P RADEON R7, 12 COMPUTE CORES 4C+8G
Motherboard Acer Wasp_BR
Cooling It's Copper.
Memory 2 x 8GB SK Hynix/HMA41GS6AFR8N-TF
Video Card(s) ATI/AMD Radeon R7 Series (Bristol Ridge FP4) [ACER]
Storage TOSHIBA MQ01ABD100 1TB + KINGSTON RBU-SNS8152S3128GG2 128 GB
Display(s) ViewSonic XG2401 SERIES
Case Acer Aspire E5-553G
Audio Device(s) Realtek ALC255
Power Supply PANASONIC AS16A5K
Mouse SteelSeries Rival
Keyboard Ducky Channel Shine 3
Software Windows 10 Home 64-bit (Version 1607, Build 14393.969)

Check the comments

What AMD has essentially done is broken down the CPU into compute units that can either act as x86+AVX scalar units or as GPU compute units.

the vector units are 16-wide, which means they can be used as SIMD units for GPU tasks or as AVX-SIMD units for the x86 scalar unit.

It is likely that we will see this first 512-bit AVX implementation (16-wide) as we would see half the units being idle if AMD stuck with a 256-bit AVX implementation (8-wide).

We can expect to see a very efficient use of silicon with this architecture as the CPU can tune the ratio of x86 vs compute units depending on the workload, potentially hundreds if not thousands of times per second.

- Tom Hammond

Be interesting if that works out
 
Top