AMD or Intel as well as the memory brand - page 72

 

To be honest, I'm not interested in the subject of "hardware".

However, I would like to share my observations about the i7 965. I bought it solely for analysis and forecasting of time series in any markets.

  1. My first impression was that I wondered why I paid so much money for it. I expected an order of magnification of the analysis speed but I didn't notice it during the first acquaintance.
  2. I searched on the forum for comments on the MT4 performance increase. The answer - the multiterminality.
  3. I divided the task into 6 parts and started each of them on a separate terminal. My optimism increased at once. Why did I need 6, but not 8 or more? I've left a reserve for other tasks not to waste time. By the way, tried to raise priority to high and real time - no effect, so I left medium. Also, I tried to set MT terminals to a certain CPU. I didn't detect any acceleration effect.
  4. I decided to transfer my analysis to MATLAB. Vector algebra is good enough for me. I ran the test. The results: MT4 takes about 30 seconds, while MATLAB takes about 0.3 seconds. The result was encouraging, but not for a long time. It turned out that MATLAB does not want to handle matrixes of the size I need. And again, splitting and gluing together the result reminded me of something.
  5. As a result I've reached a conclusion - I need faster methods (algorithms) of analysis.
  6. Experts in "hardware" advise me to switch to vista64 or w7. So far I'm thinking about it.

If anyone is interested, I can post performance screenshots in next posts.

 
DC2008 >> :

If anyone is interested, in future posts I can post screenshots of the performance......

Apart from me and you, no one else seems to be interested in script-like tests... Or, to be more precise, tests are not directly related to optimizer.

 
DC2008 >> :
  1. My first impression is - what do they charge such money for? I expected to increase the analysis speed by an order of magnitude, but as a result of my first acquaintance I didn't notice it.

An order of course is a lot, but if you look at the first table, you get this picture:

benik

Celeron 325 @ 2.53 GHz, cache 256 K L2

DDR2 PC-6400 1GB

103.3*2.53=261.35

438*2.53=1108.14

begemot61

Xeon W5590 @ 3.47 GHz, cache 4x256 KB L2 + 8 MB L3

DDR3 PC-10670 12GB

27,53*3,47=95.53

62*3.47=215.14


This is if you compare the weakest and the strongest system by ff criterion. The difference is more than 5 times (on optimization; less noticeable on script). Progress is evident (by the way, increased number of cache in PIV doesn't help much: the result is still only slightly better than in Celeron). On the other hand, I agree with you that more can be done with software.

If anyone is interested, I can post performance screenshots in the next few posts.

I doubt that the results will be significantly better than the figures shown by Xeon W5590 and Core i7 920 (in ff-performance, of course), but purely for statistics it's still interesting. If you make up your mind, drop me a line and I'll give you detailed instructions on what to do.

 
joo >> :

Apart from me and you, no one else seems to be interested in script-like tests... Or to be more precise, tests that have no direct relation to optimizer.

"And don't give me that tragic voice" - (That Munchausen)

It's just that everyone will choose a configuration specifically for MT4 (5), focusing on the performance of the Optimizer. And the fact that this performance is more than enough for tasks solved by scripts... or so: if the performance is not enough for your scripts, it will NOT be from the fact that the choice was made on the basis of optimization test. You just can't buy such a config for reasonable money.

 

Svinozavr писал(а) >>

You just can't buy a config like that for reasonable money.

That's the idea right there!

If a 100% config will cost us 100,000 rbl. and for a config that provides only 75%

but the price, well, let's say 50,000r. then it already makes sense to reconsider the price/speed for today.


It is clear that the best is the best, however, and a reasonable limit must be...

 
Mathemat писал(а) >>

I doubt the results will be significantly better than the figures shown by the Xeon W5590 and Core i7 920 (in terms of ff-performance, of course), but purely for statistics it's still interesting.

Matlab has a performance test. Here are the results.

These results are obtained with 6 tests running. You can see the computer load in the following screenshot.

And this is a test on a non-calculating PC.

 
Svinozavr >> :

It's just that the configuration specifically for MT4 (5) will all be chosen based on performance specifically in the Optimiser.

No. Everybody just do not see resource-intensive tasks in MT apart from optimizer's work. And even if they do, they don't use it in their daily work. At least most of them do. But never mind. I will wait for MT5. The speed of code there can be seen with the naked eye. And there is also CUDA. I've downloaded from nVidia site toolkits, will study it now. And it's no problem to transfer the code into dll.

 
joo >> :

Nah nah. Just all do not see resource-intensive tasks in MT apart from optimizer's work. And even if they see them, they do not use them in their daily work. At least most of them do. But never mind. I will wait for MT5. The speed of code there can be seen with the naked eye. And there is also CUDA. I've downloaded from nVidia site toolkits, will be studying them. And it's no problem to transfer the code into dll.

... if there is not enough performance for your scripts, it will NOT be from the fact that the choice was made based on an optimisation test.

Taken out of context and banging on the open door. I made it clear. Haven't you read to the end?)))

And the 5's tester - yes, if that's how I imagine/dream about it))). that would be great. But "our sorrowful labour will not be lost")))

 
joo >> :

Apart from you and me, no one else seems to be interested in script-like tests... Or to be more precise, tests not directly related to optimizer.

Why. I'm also very interested in the speed of calculation of serious things

 
Svinozavr >> :

... if the performance is not good enough for your scripts, it will NOT be from the fact that the choice was made based on an optimization test.

Taken out of context and banging on the open door. I made it clear. Haven't you read to the end?)))

And tester 5 - yes, if that's how I imagine/dream about it))). it will be great. But "our sorrowful labour will not be lost")))

Your point I understand well. But my point is that we're not loading the tester the way it could have been loaded. My point, on the other hand, you don't seem to have understood it. But it doesn't matter, by and large. For orientation, so to speak, "on the ground", that last expert will also do.