HCW does quite a few overclocking reviews for both Intel and AMD processors. This time, Carl Nelson got a hold of the high-end AMD A10-7850K and gave it a pretty healthy boost in frequencies. By the time he was done with it, the CPU was operating a whole gigahertz above stock simultaneous with a 300 MHz boost to its integrated graphics.
Image Credit: HCW
3DMark 2013 Fire Strike scores gained 27%.
One again, they break down tests along a suite of different games of varying engines and add some OpenCL tests to round things out. In real-world applications, the increase was not quite as dramatic as the one seen in 3DMark but still significant. This overclock allowed certain games to jump from 720p to playable at 1080p. Apparently this silicon is a decent little overclocker.
So if AMD can get their
So if AMD can get their process yield gremlins in order they should be able to increase the default clocks on Kaveri, with that much overclocking headroom on this chip the fab process variability kinks should be able to be worked out, and this chip has some potential! AMD needs very badly to get this chip working with its descret GPUs in crossfire, or some combination of HSA GPGPU enhanced physics with the Integrated GPU/CPU and gaming graphics on AMD’s descrete GPUs! This APU, using all of its HSA computational ability, matched with a Highend GPU could potentially be an i5 killer(certainly for integrated graphics) when piared with an AMD high end GPU! AMD needs to make this APU work with their most powerfull descrete GPUs, even if The APU’s integrated GPU can only be utilized for CPU type gaming loads and can not crossfire with the top end AMD GPUs. AMD’s HSA really has the potential to make its intergrated GPU act like an extra few CPU cores, but I am waiting to see just what descrete AMD SKUs can be paired with this APU in a corssfire arrangment.
I have extremely high hopes
I have extremely high hopes for GPGPU tasks but until there are more programs that take advantage of this then HSA and hUMA are void until that point. This is a good way to start laying down the bricks and mortar to get HSA and hUMA adopted but unless they start cranking out chips that compete with an i5 IPC-wise then mass adoption won’t kick in for these chips. I’m definitely going to be picking one up for HTPC and regular living room computer stuff but don’t have high hopes for this version in particular. Maybe in the next couple of generations will we finally be able to see the benefits of HSA/hUMA being integrated with more programs I use on a daily basis.
I duno man, these days a i5
I duno man, these days a i5 is alot of power, to be frank more power than most people need. The i3 realy is the sweet spot for average consumers, pure performance for price this chip does kick a little ass
hUMA is in the hardware, and
hUMA is in the hardware, and works through the shared memory controller on the APU. hUMA works through the underlying hardware, in a manner simular to the way virtual memory is implemented (the hardware hides the implementation of virtual memory from the application software, same goes for hUMA) the OS deals with the virtual memory, and the OS deals with hUMA( it’s one hell of a lot easier for the OS/APU software drivers to pass a 64 bit pointer than gigabytes of memory between GPU and CPU). HSA can be as little as making sure the OpenCL drivers are offered for any GPU hardware, to AMD’s more hardware optimized version of some HSA features, along with a standard HSAIL(VM) running on any hardware( the HSA foundation is working towards utilizing HSAIL, HSA Intermediate Language software(VM) That has the ability to run HSAIL on CPUs and GPUs that host the standard HSA VM). Yes Kaveri is waiting on application software venders to adopt more HSA features, but openCL, took a while to work its way into the application ecosystem also, and openCL is as HSA as any other HSA, AMD does not own the term HSA, its marketing monkeys may be slinging the term around like it is an AMD only feature, but HSA will benifit the entire CPU/GPU industry, and users too.
As far as the i5 and kaveri, did you read the review that Scott linked to.
OpenCL (and OpenGL and
OpenCL (and OpenGL and DirectX) were designed before HSA was a thing. To take advantage of the HSA performance improvements you’d need to re-write an OpenCL app in the brand new OpenCL 2.0
The part of HSA I find most exciting is its integration into LLVM. The tight integration with the CPU and memory interface opens up the possibility of compilers and language runtimes speeding up app execution automatically much the way SSE extensions do now. THIS is what I’ve been waiting for.
HSA is 3 words, and not a
HSA is 3 words, and not a thing OpenCL does general purpose Computation on GPU, so it fits the 3 words! OpenCL is a driver HAL that enables HSA functionality on ANY GPU that has the openCL drivers available, do not let AMD’s marketing monkeys bastardize generic terminology, Intel’s Ivy bridge CPU/GPU based SKUs have OpenCL drivers and OpenCL enables HSA on the Intel product, ARM has OpenCL for its mali GPU, as well as others. OpenCL is mentioned by the HSA foundation as being HSA, HSA is not a trademark, that is owned by AMD! OpenCL and other Driver/APIs/frameworks that allow the utilization of GPUs for general purpose compute, are the very beginnings of HSA on the PC, hUMA is just Uniform memory access between processing units(shareing the same memory address space), and Heterogeneous computing, has been around for a good while, just read the wikipedia on Heterogeneous computing and see companies like Cray, and IBM, and others. hUMA is just AMD’s marketing trying to differentiate AMD’s products, just like the laundry soap manufacturers try to differentiate their product to make it stand out to the average consumer. Most of the stuff you see on PCs today was on mainframes in the 1960s and 1970s into the 1980s, VMs hosting multiple OSs(IBM in the mid to late 1960), multitasking OSs(same thing 50 years ago), vector processors(GPUs) Cray, and others. The only thing new is LCD displays, and More circuit density on microprocessors, and a few incremental programming language improvments. THE UCSD P-System(1978) was the first LLVM(precursor to JAVA), code was compiled into P-code IL and run across different computing platforms for UCSD pascal on IBM, orginal PC, CP/M OS, etc. HSAIL is just a more standard up to date VM/LLVM implemntation, the the HSA foundation is proposing as a industry standard VM for cross platform and cross computing device(CPU, GPU, OTHER) HSA aware systems. All these things that marketing monkeys try to push as new, for the sake of selling to Vern in Podunk. AMD’s Push towards HSA in the PC world, is a good thing, but those marketing folks make me LOL.
Add PDP-11 to the UCSD
Add PDP-11 to the UCSD P-system, as the P-system was initially developed for the DEC PDP-11 minicomputers.
Big Omition, That’s a spanken!
I am just waiting to jump
I am just waiting to jump into Kaveri because we all know that the ps4 has 1152 shaders built into its APU. 1152 shaders is almost the amount of shaders that the 7850 GPU has! That being said we all know AMD has way more potential for their APU lineup in the future.
I am interested in the APU’s because I wanna build a super small Mini ITX build and be able to play all my games at medium to high settings. I mean right now I cannot build a super compact PC that has more GPU potential than the consoles but I bet in a year or 2 I will be able to with an Integrated GPU.
I do not think that HSA or hUMA will help AMD to prove better or even as good as an Intel cpu but we also know that the xbox one has an 8 core processor inside so maybe that is also hinting at an 8 core APU? 8 cores plus HSA and hUMA and a iGPU as powerful as an 7850 Descreet GPU may prove to be a contender.
Rumor has it that the XBox
Rumor has it that the XBox One chip does not have HSA and is a more traditional APU. The PS4 chip is at least hUMA and probably full HSA. Sony and AMD are founding members of HSA Foundation and Microsoft is not. If the PS4 chip has full HSA support I find it MUCH more likely that we will see developer tools and game engines getting HSA optimizations which could give AMD a big advantage once these improvements emerge into the PC space.
What I notice in the image,
What I notice in the image, on the highest OC they say DDR3 2400 where as rest they don’t say anything? Was that cause they used slower ram on other 2 test or something?
The highest overclock had the
The highest overclock had the RAM at 2400 MHz.
i dunno. the graphics
i dunno. the graphics benefits at this price point seemed negated by the almost mandatory need for expensive high seed RAM. Why not just spend that on a less powerful CPU and go with a low cost discrete card. At least for gaming you will come out ahead most of the time.
Actually fast ddr3 is almost
Actually fast ddr3 is almost the exact same price as “regular” ddr3
Yep, and it’s still the case
Yep, and it’s still the case today. The price gap between 1333-1600mhz ddr3 and 2133-2400mhz ddr3 is about $10 on average now.