Multi GPU Thoughts - Micro Stuttering Redux

Introduction

First I would like to mention that it is not the first time that someone brings problems with Multi GPU implementations into the spotlight, as you may already have heard about "micro stuttering". Micro stuttering was described as an uneven frame rate when using two GPUs, something that may happen for other reasons like lack of RAM, software accessing the HDD at the same time, software using the CPU(s), etc. Many things can disturb frame rate but the issue with micro stuttering was that the uneven frame rate seemed to only occur when SLI or CrossFire are in use. PCGH had a look at the problems and published some data a while back and it was an interesting look at the problem. The issue is that they didn't notice the problem's origin. They noticed that frames were arriving at uneven intervals, which mitigates the purpose of a higher frame rate providing better responsiveness, which is what is expected from these setups. Keep this in mind: better frame rate must be equal to better responsiveness, or it's a moot point.

Still, an uneven interval should still provide better responsiveness nonetheless, if the frame rate is going to almost 2x. This is where it gets tricky: looking at just the frame times doesn't show you the real problem that's going on with contemporary multi GPU solutions, the big problem is that the responsiveness is in fact worse!

Before we dwell any further into this issue, we need to look at multi GPU rendering modes and have a more detailed look at how the issue presents itself to the user.

AMD Releases Radeon Mobility HD 6990M

And it's that time of the week, where AMD releases a product that has such an absurd model number that one has to wonder what floats in the air of Sunnyvale. Take a wild guess at the performance and specifications of the new Radeon HD 6990M, which is targeted at performance oriented laptops. Please, go ahead. For sure your mind isn't ready for the amount of bullshit coming your way.

The Radeon HD 6900M series currently features "Barts" chips that have 960 shaders which - for those keeping score - is shy of the 1120 that are present in the Radeon HD 6870 and even slower clocks. Obviously, performance of a 6970M is nowhere near the performance of the HD 6970.
The desktop Radeon HD 6990, on the other hand, is a dual GPU card, so you would expect some kind of "Cayman" chip enough of a performance bump to justify sticking it a HD 6900M model number to boot. So, no, what you get is paltry fully enabled "Barts" chip that is present in the Radeon HD 6870 but which is clocked 20% lower. That's right, for the geniuses working at AMD marketing, it is correct to market a graphics device with 1/3 of the performance as if it was the real thing. Perhaps this was the motive behind ditching the "X2" suffix on dual GPU cards earlier.

Intel has been doing this, Nvidia has been doing this, AMD has been doing this for some time now and yet it doesn't get any more excusable for each passing day. It is absolutely disgusting that no consumer advocate association steps up and take them to court on accounts of consumer fraud, or misleading marketing or whatever you call it. Slapping an "M" to a card that has nothing to do with another very similar part is purportedly misleading and someone should be held accountable for this. Nvidia has been doing this for a long time but I think AMD is actually the first to blatantly sell a product that has a dual GPU equivalent on the desktop as a single GPU in the mobile space.

Now, don't interpret me wrong on this, I know it is impossible, or practically impossible, to fit the same parts on something that can still be called a laptop or that has still some mobility of a kind. That is mostly impossible but if shader counts, memory and core clocks are being cut at will to fit TDPs, model numbers must reflect that. It obviously gives the consumer the information beforehand that the laptop is and always will be inferior to a desktop, instead of a "hey, we have a laptop here with the same GPU!".

To make matters worse, big publications that only some god knows how they get the traffic they do, completely miss such blatant facts about new product launches, instead focusing on the ongoing war between the two companies.

If you know someone who may help put end to this, do leave me a comment below. This situation has existed for too long of a period. It makes a lot of people waste their time with this bullshit for no good reason. It is not easy to counsel someone a new piece of hardware with this kind of marketing. I can just imagine what people that want to do a decent job at a hardware store must go through.

We all need to buy computers from time to time, do make us a favor and make our life easier so one can actually look forward to the next upgrade.

I leave the best for last, a must read from the press release:

There’s always been a belief that when it comes to mobile computing you need to make performance compromises. Today AMD demolishes that myth. The AMD Radeon HD 6990M GPU, which not only packs AMD Eyefinity technology with unprecedented specs, also provides full Microsoft DirectX 11 and Stereo 3D support. Bottom line, this processor is epic and it’s here - now.

It is not a myth, it is a fact. This guy is either lying through its teeth or doesn't know one thing of what he's talking about. Either way, this is unacceptable.

Acer Laptops With AMD C-50

The model is the Aspire 5253 and comes equipped with one screen that is... wait for it... 15.6" diagonal. Nice...

Zotac FUSION350-A-E Announced - AMD E-350 Inside

Giddy up for another AMD "Bobcat" based E-350 media center hardware with Zotac. Specs after the jump.

Choosing Laptop Processors - Core Series

Intel processors have been among the most confusing lines of CPUs yet, since they feature completely different naming schemes compared to the hardware on Desktop CPUs. For instance:

Core i5 760  in desktop: 2.8GHz, 3.33GHz Turbo, 4 cores, no HyperThreading

Core i5 450M in laptops: 2.4GHz, 2.66GHz Turbo, 2 cores, HyperThreading

So, it should come as no surprise that the desktop CPU is more than two times faster. HyperThreading helps a bit but not enough to really turn a laptop into a power machine. General rule of thumb, if you get a desktop nowadays, you get twice the CPU performance for slightly less money -  always go for quad cores, i5 700+ and i7 800+.
Still, even in laptops Intel doesn't make it clear what they're selling to the customers, so I'll try to be of some help here. This is what I've been able to gather.

• Intel Core i3       - 4 threads, 2 cores
• Intel Core i5       - 4 threads, 2 cores, Turbo
• Intel Core i7 non-Q - 4 threads, 2 cores, Turbo, +1MB L3 cache(4MB), Hardware AES, VT-d, TXT
• Intel Core i7 Q     - 8 threads, 4 cores, low clocks, 6MB L3 cache on 700 models, 8MB on 800 models

Intel Core i3/i5/i7 U series have the same hardware as the other processors but lower base clock and power consumption. A Core i3-380UM is much slower than a Core i3-380M since it runs at 1.33GHz vs 2.53GHz base clock and there's no Turbo.

The Core i7 Q processors are not very fast compared to Core i7s 600 with higher clocks(the 640M is basically the same speed as the 720QM in heavy load) since they run at speeds not higher than 2GHz in base clock. In my book this makes the Core i7-600 a better choice for high performance mobile processors.

The Core i3s are the best choice all around, with a special mention to Core i5s only if you use a lot of virtualization software since some already allow VT-d to be used, providing better graphics performance in guest operating systems by allowing virtualized access to the GPU. The Core i3 still supports VT-x, which means it will still be very fast at the rest. The lack of Turbo isn't particularly problematic since most software is already threaded to some degree and in those situations the use of Turbo will be limited and i3s and i5s clocked at the same base speed will converge to a similar performance.

If you need help deciding do drop by the forums and I'll be sure to help you look at some options.

OLPC XO-1.75 News

Thoughts On ''Bulldozer'', "Sandy Bridge"

I have already spilled some beans about the two upcoming architectures from the big players in the x86 game:

• AMD's Bulldozer
• Intel's Sandy Bridge

But it was not until just now, while doing some research, that I actually stumbled on some clarifications from Intel about AVX support in "Sandy Bridge". Well, it turns out that while Intel will have 256 bit FPUs per core but one big problem with first generation processors:

Hi Igor,

Sandy Bridge will not have FMA, it's targeted for a future processor. 

...

It sounds like you are an FMA supporter - beyond the raw FLOPS improvement, do you have any sensitivity to the numerical advantages FMA can provide? There are obviously a lot of tradeoffs in the implementations we can provide, and having some data to understand how you would use it would be very helpful.

Regards,

Mark Buxton

AMD Radeon HD 6800 Series Thoughts

AMD Ontario, Zacate Die Sizes - Take 2

Earlier I asked if AMD's leaked Bobcat based Fusion processor die shots were real or another PR show off, this update sets some ideas straight.

Intel Shows Off Ray Tracing In Wolfenstein

Intel's Daniel Pohl once again delivers exciting new prospects about real time ray tracing on regular CPUs, this time with another id Software title.