MQL5 Cloud Network: Are You Still Calculating? - page 6
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
And perhaps, there will be some solution allowing to bypass this limitation.
They have no future for the average user right now. I mean, at all.
Let's compare. Performance for double
7970 -- 0.95 TFlops
k20x -- 1.3 TFlops.
Now pay attention to the price )))
7970 -- 400
k20 -- 3200 (!) k20x costs unreal money.
Apparently your decision to organise agents as services was wrong.
Tesla by Nvidia certainly has no future. What they are doing is beyond good and evil. They artificially slaughter the performance of gaming cards for the sake of "high performance solutions" and charge unrealistic price tags. Even outdated and weak Tesla cards cost several times more than top-end graphics cards, but they lose to them.
Fortunately, GPU development is going very fast, AMD/ATI have overtaken NVIDIA in the compute part, Intel is accelerating its integrated solutions, and CPUs are easily becoming (with AMD and Intel drivers) OpenCL devices. The only problem is the limitations of the video driver model.
We are betting that the market will still come to massively exploit the parallelism of GPU devices. In the next version we will enable fuller OpenCL support in remote agents, allow CPU usage and measure the statistics of OpenCL device distribution in MQl5 Cloud Network. The statistics will be publicly available.
Then it will be clear what to do next.
Or perhaps there will be some solution to get around the current restriction.
I hope so. And you will get in addition to the current one a part of a completely different market, which by the way is now rapidly developing.
We bet on the fact that the market will still come to the mass use of parallelism of GPU devices.
Still, despite the fact that this (in my opinion) is a marketing restriction, someday hardware manufacturers will make access to OpenCL functions from services.
And perhaps, there will be some solution allowing to bypass this limitation.
Well, if a terminal launched by a user can launch local agents that have access to all OpenCL goodies, then maybe it is worth making a separate agent (I think not many people have 2 or more video cards) for OpenCL launched from the "Autoload" folder?
And let it hang quietly in the tray.
And leave the service agents, which already exist, for ordinary calculations, they start even without user login.
So already :) crypto coins play not the least role in this :) now a large number of top cards (AMD of course) for the user segment are dismantled by miners.
Not anymore. They have passed the easiest, then reasonable part of the way and now there is no sense at all in bitcoin mining.
In fact, only games play the only meaningful role on a global scale.
Not anymore. They are past the easy part, then the sensible part and now there is no point in mining bitcoins at all.
There's still some, I'm in that field all the time right now. But on the fringe.
Basically the only meaningful role on a global scale is in gaming.
Here is how OpenCL startup and initialisation looks like in service agents on the example of NVIDIA Tesla:
Everyone will have it working from the next build.Yay! Finally my Tesla C2070 will work! I'm always counting rainbow tables on it, and picking up passwords :)
Dear programme specialists! Can you explain why in the log:
<(30 units, 1440 MHz,...> .
As I remember, this Tesla (C1060) has 240 cores. Or does 30 units mean what? Will all cores be used in the calculations?
Will there be optimisation for CUDA?
<(30 units, 1440 MHz,...> .
I didn't understand how to correct my own post, so I'm answering myself :) 30 are multiprocessors (pixel pipelines).
On C2070 there are 48 of them, on the neighbouring Quadro4000 -32 pcs. Can I ask when the next release of Meta-tester is planned?
I didn't understand how to correct my own post, I'm answering myself :) ....
And this is a bug of the forum. It's a feature :)
Usually on my own posts there is a sign "edit", and on other people's posts "reply".
And now on all posts "reply".
It turns out that you can edit your post only if it is the last one. If it's not the last one, then that's it, the train has left.
And this is a bug of the forum. It's a feature :)
Usually on my own posts there is a sign "edit", and on other people's posts "reply".
And now on all posts "reply".
It turns out that you can edit your post only if it is the last one. If it's not the last one, it's gone.