• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Desktop Kaveri APU With 13 CUs Enabled Radeon R5 M200 (832 Stream Processors) GPU Spotted

The old rumour threads I read seemed to think a Nvidia solution was more likely due to them having a better relationship with Valve and better Linux driver support, how much truth is in that though I don't know.

Both will sort their drivers out. There's no doubt that if this is successful manufacturers will want to sell Kaveri APU systems running SteamOS. Valve's hardware might be something a bit higher if it's to compete with the next gen consoles.
 
Agreed, taking full advantage of more cores can only be a good thing. Look forward to seeing games being developed with this in mind and how they run on AMD's Steamroller architecture. My last AMD chip was a 1055T, would love to have a reason to go back to an AMD CPU. 8 core Steamroller may be it...

Game engines don't really lend themselves well to that style of multithreading its all well and good saying game developers should better thread their code but by its nature its not something that works that way very well compared to say media encoding or other bulk data processing.

While the nature of the game sitting on top of the engine can have some effect on how well additional cores are utilised the engine itself will have large amounts of code that needs to be processed in a specific order and within very rigid time frame not lending itself to multi-threading very well.
 
The old rumour threads I read seemed to think a Nvidia solution was more likely due to them having a better relationship with Valve and better Linux driver support, how much truth is in that though I don't know.

That would mean one of two things, either ARM CPU and separate Nvidia discrete GPU (unlikely)

Or a Tegra Chip, which compared with AMD / Intel APU's is an absolute joke, Bejeweled 2 is probably not what Steam have in mind for a gaming Box

It may well be a Tegra Chip, but a much better solution is an APU for a proper gaming box.
 
Last edited:
This is what a game developer said when talking about 8 core AMD CPUs vs quad i5s.

"This (Sony) approach of more cores, lower clock, but out-of-order execution will alter the game engine design to be more parallel. If games want to get the most from the chips then they have to go 'wide'... they cannot rely on a powerful single-threaded CPU to run the game as first-gen PS3 and Xbox 360 games did. So, I would probably go for the AMD as well, as this might better match a console port of a game... based on what we know so far."
 
This is what a game developer said when talking about 8 core AMD CPUs vs quad i5s.

"This (Sony) approach of more cores, lower clock, but out-of-order execution will alter the game engine design to be more parallel. If games want to get the most from the chips then they have to go 'wide'... they cannot rely on a powerful single-threaded CPU to run the game as first-gen PS3 and Xbox 360 games did. So, I would probably go for the AMD as well, as this might better match a console port of a game... based on what we know so far."

Its Sony who are also the first to realise that Multithreading is the only way to go now if you want to be on the PS4.

http://attackofthefanboy.com/news/planetside-2-dev-calls-ps4-amd-chip-challenge-work/

Despite Sony making the PlayStation 4 an easier system to develop for, and having been praised by developers across the industry, there still could be some hurdles for developing on the new console.
According to Matt Higby, a developer on Sony’s Planetside 2, the Forgelight Engine and AMD chips that rely on multi-threading, are posing a problem.
“People who have AMD chips have a disadvantage, because a single core on an AMD chip doesn’t really have as much horsepower as they really require,” said Higby in a recent interview with Eurogamer.
Higby’s problem is one that stems from a game that was developed for a different type of system. The problems that they are facing right now with Planetside 2 is that they are needing to re-engineer their game for the new platform.
“It’s very challenging to split those really closely connected pieces of functionality across in multiple threads. So it’s a big engineering task for them to do, but thankfully once they do it, AMD players who’ve been having sub-par performance on the PC will suddenly get a massive boost – just because of being able to take the engine and re-implement it as multi-threaded.”
While not all developers are likely to experience this type of challenge, Higby says that “Planetside 2 is a prime example” of challenge of making a game designed to work on “one really big thread” to be compatible with Sony’s multi-threaded AMD hardware.

This could very well mean re-engineering game development tools and engines.

mmm'good! :D
 
Thing is you can't take a game engine and force it to "go wide", I see a lot of this talk about re-engineering engines from people who really should know better.

While increasingly complex physics, AI, etc. will help uptake of utilisation of additional cores theres just so much that you simply can't apply out of order execution to there simply is no way to re-engineer for it.

Kaveri however being based on Steamroller will side step this issue as its been designed in to be able to boost 1 and 2 thread priorities when needed, without having to wait until other threads become idle (which is what holds piledriver, etc. back) and in theory Steamroller based CPUs should be competitive in games that depend highly on single threaded performance.
 
These are game developers talking though, not commentators.

For certain, what holds back Piledriver is typically one core stuck at 100% and the other cores, even if they're being used by other threads, often as low as 30%. That's why Piledriver gains so much with a 4.5GHz+ overclock as it frees up the overloaded core.
 
Thing is you can't take a game engine and force it to "go wide", I see a lot of this talk about re-engineering engines from people who really should know better.

While increasingly complex physics, AI, etc. will help uptake of utilisation of additional cores theres just so much that you simply can't apply out of order execution to there simply is no way to re-engineer for it.

Kaveri however being based on Steamroller will side step this issue as its been designed in to be able to boost 1 and 2 thread priorities when needed, without having to wait until other threads become idle (which is what holds piledriver, etc. back) and in theory Steamroller based CPUs should be competitive in games that depend highly on single threaded performance.

Yeah I get it, it may be difficult for some, but what I don't believe in is the word "impossible" or "can't be done"

In my mind, and I don't for a second pretend to be in any way an expert, devs have been far to comfortable to do it the easy way.

I think I have said it before, this is exactly the sort of "no option kick up the back side" that has been needed for far to long.

There is only so much faster a single thread can get, even Intel are stagnating on that front, the only way forward now to gain performance is to go wide.

Where money, time and plain bone idleness is concerned nothing moves forward unless its forced to, Multicore CPU's have been around for 10 years tempting devs to do more, the carrot hasn't worked, here comes the stick.
 
Is this ever actually going to be of use to gamers? Right now it seems a total waste to have iGPUs that no high end gamer will ever use taking up huge numbers of transistors on high end CPUs. You simply don't buy a high end CPU and then use integrated graphics. It would make more sense to make high-end CPUs have no iGPU and have low-mid end CPUs have the best iGPU possible for those on a budget. At least today that is, but I can't see iGPUs ever being more powerful than discrete GPUs anyway and hence having any place on high end CPUs.
 
Last edited:
Not a great fan of the notion that devs haven't pushed threading more out of laziness or being too comfortable - the lengths some of the big names behind idtech, unreal, etc. engines go to eek out extra performance, optimisations, etc. certainly don't deserve to be branded that way heh let alone deserve the stick.
 
Not a great fan of the notion that devs haven't pushed threading more out of laziness or being too comfortable - the lengths some of the big names behind idtech, unreal, etc. engines go to eek out extra performance, optimisations, etc. certainly don't deserve to be branded that way heh let alone deserve the stick.


Agreed, but those are not necessarily who i'm referring to, the biggest problem is with MMO's who still use DX9 engines.

Later games built on Frost Byte, Unreal, Cryengine.... do rather better in Multithreading, Crysis 3, BF3, Tomb Raider.... even Farcry3 don't bottleneck the GPU, often even on AMD CPU's. a lot of MMO DX9 games do, even on Intel CPU's, and you can see why when you watch just one thread maxed with the rest sitting virtually idle, Planet Side 2 does exactly that.

Having said that, there is always room for improvement even with the best engines.
 
Last edited:
Is this ever actually going to be of use to gamers? Right now it seems a total waste to have iGPUs that no high end gamer will ever use taking up huge numbers of transistors on high end CPUs. You simply don't buy a high end CPU and then use integrated graphics. It would make more sense to make high-end CPUs have no iGPU and have low-mid end CPUs have the best iGPU possible for those on a budget. At least today that is, but I can't see iGPUs ever being more powerful than discrete GPUs anyway and hence having any place on high end CPUs.

depends on what can be don't with it, there are already tools out there that make use of the iGPU to boost the Discrete GPU, granted they don't work all that well, yet. but you see my point :)

AMD themselves could also do something along those lines with this, Hybrid Cross-Fire, iGPU paired up with any AMD Discrete GPU to boost it.
 
MMOs are definitely an area that lend themselves towards mulithreading and definitely could be improved from where they are.
 
I want an 8 core Steamroller CPU - I just don't know when it's coming and in what form it'll be.

I was thinking this too, but am holding back to see if they are releasing a new socket now before planning the rest of the build.
 
On the other hand 8 low performance cores is a good idea for game development, just coding everything on one or two thread wont do with it, it will force game devs to spread the workload across many more threads / cores and the result of that is no more bottlenecking the GPU on AMD and Intel CPU's.
Hopefully it'll pan out like people want.
 
Last edited:
Thing is you can't take a game engine and force it to "go wide", I see a lot of this talk about re-engineering engines from people who really should know better.

Rroff your not seeing the wood from the trees. If you dont force some people to change their ways then they will continue coding like they have always done.

It may be a challenge now but hopefully it will breed a new generation of parallel development. Keyboards have thier place but look how everyone embraces the touchscreen now it on lots of devices. Hell my 14 month old girl can navigate round the ipad well - resistance is futile!!! :D
 
Thing is you can't take a game engine and force it to "go wide", I see a lot of this talk about re-engineering engines from people who really should know better.

That's like a low league football coach talk who plays down the middle! :p

There is only so much faster a single thread can get, even Intel are stagnating on that front, the only way forward now to gain performance is to go wide.

This is apt. Look how the overall speed has tapered off, the core 2 range could hit 4 Ghz and they are still releasing better chips but around 3 Ghz out of the box. Raw single core grunt has had it's day but until developers start utilising more threads/cores it will stagnate not innovate.
 
Last edited:
On silicon perhaps.
But Intels stagnating could be down to the crazy things they're doing with power consumption as that's the focus right now.
 
Back
Top Bottom