Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [wccftech] AMD Jaguar based Beema And Mullins APUs Confirmed For 2014 – Feature HSA Enhancements
New Posts  All Forums:Forum Nav:

[wccftech] AMD Jaguar based Beema And Mullins APUs Confirmed For 2014 – Feature HSA Enhancements - Page 5

post #41 of 43
Quote:
Originally Posted by Newbie2009 View Post

Where are these APU's performance wise at the moment? Faster tan say a 9800GT?

depends which one you're talking about. And how fast the ram is and how overclocked the igpu is. At the absolute top end with ddr3 2133mhz ram, a Richland a10-6800k overclocked is about as fast as a discrete HD 6670 with ddr3 ram (not the gddr5 version). A trinity a10-5800k with high speed ram tops out about 77% of the performance of the above 6670, or something like (or a little better then) a HD 4650 or a GT 440, or an old x1950.

So yes, a richland a10-6800k with high speed ram and under a good overclock will match up well with a 9800GT.
 
Deep-6
(14 items)
 
 
CPUMotherboardGraphicsRAM
Core I5-4690K Asus Maximus VI Gene MSI GAMING 4G GTX 970 gskill sniper 
Hard DriveHard DriveCoolingOS
Samsung 850 evo  Samsung 850 evo Thermalright HR-02 Macho Rev.B Windows 10 Pro 
MonitorKeyboardPowerCase
Acer K242hl Corsair Raptor K30 Seasonic X-650 Fractal Design Arc Mini R2 
MouseAudio
Logitech MX518 Sennheiser - MOMENTUM Over-the-Ear Headphones 
  hide details  
Reply
 
Deep-6
(14 items)
 
 
CPUMotherboardGraphicsRAM
Core I5-4690K Asus Maximus VI Gene MSI GAMING 4G GTX 970 gskill sniper 
Hard DriveHard DriveCoolingOS
Samsung 850 evo  Samsung 850 evo Thermalright HR-02 Macho Rev.B Windows 10 Pro 
MonitorKeyboardPowerCase
Acer K242hl Corsair Raptor K30 Seasonic X-650 Fractal Design Arc Mini R2 
MouseAudio
Logitech MX518 Sennheiser - MOMENTUM Over-the-Ear Headphones 
  hide details  
Reply
post #42 of 43
Quote:
Originally Posted by Particle View Post

the delay between issuing a read command and data first becoming available is lower in the case of DDR3 than it is for GDDR5.

That type of latency is insignificant.
PC
(9 items)
 
  
CPUMotherboardGraphicsRAM
4790K 4.6G 1.24v/1.74v, 5.1G 1.35v validation. GA-Z97X-Gaming 7 Tri-X R9 290 1100/1350 +0.012v G.Skill 2400 c10 
CoolingMonitorPowerCase
Noctua NH-D15 24EA53 IPS 76Hz OC EVGA 1000 G2 Phanteks Enthoo Pro 
Audio
SoundMagic E10 
  hide details  
Reply
PC
(9 items)
 
  
CPUMotherboardGraphicsRAM
4790K 4.6G 1.24v/1.74v, 5.1G 1.35v validation. GA-Z97X-Gaming 7 Tri-X R9 290 1100/1350 +0.012v G.Skill 2400 c10 
CoolingMonitorPowerCase
Noctua NH-D15 24EA53 IPS 76Hz OC EVGA 1000 G2 Phanteks Enthoo Pro 
Audio
SoundMagic E10 
  hide details  
Reply
post #43 of 43
Quote:
Originally Posted by fateswarm View Post

You don't deserve a response with that attitude but anyway. I'm generous.
To both of you, I'd say that what I said is simple to understand though I'm not sure if it's easy to implement: The PCI-E bus is too slow to utilize VRAM for general processing directly and too slow for the GPU to use system's RAM directly. What I believe would greatly improve the PC platform is the "P4 approach": just use VRAM for both the main system and the graphics card. That of course currently has that bottleneck of the PCI-E bus. Another philosophy concerning that is needed entirely and then it's set.

Motherboards with a socket for a GPU and a CPU or a bus that allows VRAM-type bandwidth. Anything that would avoid this inefficiency of having the system divided into two parts and rarely achieving the juggling-quality software required to utilize both optimally.
To clarify a little more about what HSA is intended for. The sole purpose of HSA is not entirely to do with PCI-E latency or another. As them are insignificant to how OpenCL works today. The point behind HSA is to give developers a break. That break being a much easier and faster tool set to build GPU accelerated applications. Programming current OpenCL accelerated applications is a feat due to having to work with two completely separate memory spaces right from the start. What HSA does is eliminates the extra memory space that would normally be allocated in your discrete graphics cards vram. It does so by using the APU's processor and graphics core, plus the system memory both are tied to. The iGPU and the CPU inside the APU both can access the same memory space at the same time. This makes programming the software a million times easier as we don't have to work with two completely separate memory spaces (hard as you have to pipe the resources from one to another). Instead this allows us to write OpenCL applications the same way you would write a normal Win32 application. It doesn't even need to be threaded to be insanely fast over a standard Win32 application.

Standard OpenCL Application
Code:
CPU ←→ Memory

↑
↓

GPU ←→ Memory

HSA OpenCL Application
Code:
CPU ←→ Memory ←→ GPU
Quote:
Originally Posted by Newbie2009 View Post

Where are these APU's performance wise at the moment? Faster tan say a 9800GT?
The A10-6800k iGPU performs roughly around a discrete HD 6670.
The Beast
(13 items)
 
Sero 7 Pro
(8 items)
 
 
CPUMotherboardGraphicsRAM
A10-6800k ASRock FM2A85X Extreme6 ASUS HD 5870 8GB AMD Radeon™ RP1866 
Hard DriveOptical DriveCoolingOS
60GB SSDNow V300 LG 24X FX Liquid Loop Windows 8 x64 
MonitorKeyboardPowerCase
HP 2009m Logitech K120 Fractal Tesla R2 650W Fractal Arc Midi R2 
Mouse
HP Modguo 
CPUGraphicsRAMHard Drive
Nvidia Tegra3 ULP GeForce 1 GB 8 GB 
OSMonitorPowerAudio
Android 4.2 (Jelly Bean) 7" IPS @ 1280×800 4000 mAh Stereo Speakers 
  hide details  
Reply
The Beast
(13 items)
 
Sero 7 Pro
(8 items)
 
 
CPUMotherboardGraphicsRAM
A10-6800k ASRock FM2A85X Extreme6 ASUS HD 5870 8GB AMD Radeon™ RP1866 
Hard DriveOptical DriveCoolingOS
60GB SSDNow V300 LG 24X FX Liquid Loop Windows 8 x64 
MonitorKeyboardPowerCase
HP 2009m Logitech K120 Fractal Tesla R2 650W Fractal Arc Midi R2 
Mouse
HP Modguo 
CPUGraphicsRAMHard Drive
Nvidia Tegra3 ULP GeForce 1 GB 8 GB 
OSMonitorPowerAudio
Android 4.2 (Jelly Bean) 7" IPS @ 1280×800 4000 mAh Stereo Speakers 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Rumors and Unconfirmed Articles
Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [wccftech] AMD Jaguar based Beema And Mullins APUs Confirmed For 2014 – Feature HSA Enhancements