Tag Info

Hot answers tagged

225

TL;DR answer: GPUs have far more processor cores than CPUs, but because each GPU core runs significantly slower than a CPU core and do not have the features needed for modern operating systems, they are not appropriate for performing most of the processing in everyday computing. They are most suited to compute-intensive operations such as video processing ...


173

What makes the GPU so much faster than the CPU? The GPU is not faster than the CPU. CPU and GPU are designed with two different goals, with different trade-offs, so they have different performance characteristic. Certain tasks are faster in a CPU while other tasks are faster computed in a GPU. The CPU excels at doing complex manipulations to a small set ...


54

GPUs lack: Virtual memory (!!!) Means of addressing devices other than memory (e.g. keyboards, printers, secondary storage, etc) Interrupts You need these to be able to implement anything like a modern operating system. They are also (relatively) slow at double precision arithmetic (when compared with their single precision arithmetic performance)*, and ...


19

A CPU is like a worker that goes super fast. A GPU is like a group of clone workers that go fast, but which all have to do exactly the same thing in unison (with the exception that you can have some clones sit idle if you want) Which would you rather have as your fellow developer, one super fast guy, or 100 fast clones that are not actually as fast, but all ...


17

Because GPUs are designed to do a lot of small things at once, and CPUs are designed to do a one thing at a time. If your process can be made massively parallel, like hashing, the GPU is orders of magnitude faster, otherwise it won't be. Your CPU can compute a hash much, much faster than your GPU can - but the time it takes your CPU to do it, your GPU could ...


11

Are you really asking why are we not using GPU like architectures in CPU? GPU is just a specialized CPU of a graphics card. We lend GPU non graphics computation because general purpose CPU are just not up to par in parallel and floating point execution. We actually are using different (more GPU-ish) CPU architectures. E.g. Niagara processors are quite ...


9

I might be horribly mistaken here, and am speaking from little or no authority on the subject, but here goes: I believe each GPU execution units ("core") have a very limited address space compared to a CPU. GPU execution units can't deal with branching efficiently. GPU execution units don't support hardware interrupts in the same way CPUs do. I've always ...


6

It isn't unless nVidia provides a VMware-compatible paravirtualized driver for that purpose. This discussion on the nVidia forums explains why. Now, newer CPUs and chipsets support "IOMMU", which could serve the similar function as "PCI-E passthrough" that they discuss on that forum. However, this still requires support and cooperation from both VMware's ...


6

This looks like what you need, although you probably need to recompile, and the SDK linked from there appears specific to Intel CPUs; this link to AMD's documentation appears to describe the equivalent for AMD CPUs.


3

It does support OpenCL. I have same GPU in my HP laptop. This is how to get OpenCL support: Download AMD app SDK short for AMD Accelerated Parallel Processing (APP) SDK (formerly ATI Stream) you get it from here Supported drivers are available here Download and install your driver using the latest Catalyst driver installer for your GPU device After ...


3

This is nothing about clock speed or purpose. They are both equally able to complete most, if not all tasks; however some are slightly better suited for some tasks then others. There has been a very old argument about whether it's better to have lots of dumb cores or a small group of very smart cores. This goes back easily into the 80's. Inside a CPU there ...


2

The whole point of there being a GPU at all was to relief the CPU from expensive graphics calculations that it was doing at the time. By combining them to a single processor again would be going back to where all started.


2

It isn't a GPU if it dosen't do graphics ;) it would be a GPGPU then. You'd want to go for a design that's massively multithreaded and with very good floating point performance. Cell processors are really very specialised power architecture based systems, so that's one option alternately, load up a system with GPGPUs such as the tesla


2

I'm not aware of any real GPU-only devices, I think the closest you can probably get is the playstation 3 with its cell processor. To my understanding, GPUs are very poorly optimized for memory management (no free lunch!) so they're not really suitable for doing anything other than repetitive simple mathematics.


2

AFAIK, AMD/ATi begun supporting OpenCL starting with their DirectX 11 cards. Your card is DirectX 10. Here is the list of supported GPUs: http://developer.amd.com/gpu/AMDAPPSDK/pages/DriverCompatibility.aspx


2

There is a point at which you will saturate the resources of your CPU, and the GPU's will be sitting idle. There is also a point at which you could run out of bus resources. Since it is a bus, there is a maximum amount of transferrable data per unit time, which could again cause GPU's to sit idle. That being said, adding GPUs should not decrease ...


2

How much a GPU can accellerate depends on the code you run. GPUs are extremely good in running simple, massively paralel instructions. Programs which can use that can gain a massive performace boost. Code which is single treaded or complex will fair quite poorly.


1

I've done a bit more research and I'm going to answer this one myself in case anyone finds themselves looking to do something similar. AWS (and other vendors) provide GPU cloud compute services. This works great for certain applications, but certainly not all. As best I can tell, the virtualized GPU clusters tend to be slower than the actual hardware ...


1

Most server chassis are not designed for the amount of heat the 4 cards are going to put out. And it sounds like your applications will be pushing the cards to peak levels. You will definitely need to closely monitor the individual cards temperatures, as well as the processor. You very well might need additional fans, or even go to a liquid cooling ...


1

If you need single floating point performance, go with the GTX680, it is also more power efficient, you won't pay as much for electricity. If you need double floating point performance go with GTX 580. Take a look at this blog post. Also renderstream had a blogpost about an 8GPU 4U server. It's 14k$. The fixed the BIOS to get it working with all 8 cards. I ...


1

I'm not sure how useful or ready for prime time this is (looks like it isn't ready for serious use yet) but I found KGPU interesting. According to this Slashdot article it was used to speed up AES operations for filesystem crypto. Again I currently have no idea how to use it.


1

http://codereflect.com/2008/09/04/how-to-enable-syntax-highlighting-for-cuda-files-in-visual-studio-2005/


1

As far as I'm aware there are none. The reason being that GPUs are really only efficient for certain math related operations. They also lack the instructions necessary to do the things required by an Operating System. Consequently there are no operating systems. However there are frameworks for having code run on the GPU. These are called "GPGPU" Microsoft ...


1

If you are ready to use Xen instead of VMware Workstation, you can try out Xen VGA Passthrough and see if your hardware configuration is supported. This would give you full control over the graphic card in the VM. Here is an example of what you can accomplish: http://www.youtube.com/watch?v=Gtmwnx-k2qg


1

It is important to keep in mind that there is no magical dividing line in the architecture space that makes one processor the "central" one and another the "graphics" one. (Well, some GPUs may be too crippled to be fully general, but those are not the ones we are are talking about here.) The distinction is one of how they are installed on the board and what ...


1

I've found an interesting solution: Freemake Video Converter. It looks like it supports GPU, and I'm currently converting a 2 hours long video (1.36GB). It got to 50% in 10 minutes. Going to share the results in a comment later.



Only top voted, non community-wiki answers of a minimum length are eligible