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
P.S. Post the improved results in the table?
Post it, of course. And I'll still try it with external graphics card (my graphics card is not so good, MSI GF6600 256 MB, but maybe it will help). Maybe this factor will prove to be sensitive as well.
P.S. The graphics card had another slight improvement, to 8:01 (481s). Though it's not too clear why: I've got plenty of memory.
Xeon, original configuration, XP64
Single run to assess correctness and number of trades
Same Xeon, 3G sclerosis (3 channels x 1Gig 1333) HP32
The result is almost identical to the previous one
Opteron-Istanbul, original configuration, HP64
I hope I tested correctly.
Thank you, Eugene.
Zeon... Yeah, unattainable (not counting the possible Core i7 tests).
And the Istanbul... On the level of a decent Core 2 Duo.
P.S. Another stupid question arises: how is it possible that the budget Phenom II X3 720 BE @ 2.813 GHz from Belford is faster (or more exactly, "better") than Istambul on one core? Is it really all about faster memory?
Thank you, Eugene.
Zeon... Yes, it's out of reach (not counting possible Core i7 tests).
And Istanbul... On the level of a decent Core 2 Duo.
P.S. Another stupid question arises: how is it possible that the budget Phenom II X3 720 BE @ 2.813 GHz from Belford is faster (or more exactly, "better") than Istambul on one core? Is it really all about faster memory?
The speed of the memory and the number of controller channels involved will certainly have an effect. But of course not directly proportionally.
Also, there is the question of how the OS will load this memory. If I have time, I will try to find slower memory and run one of the systems.
By the way, I don't recommend using asymmetric allocation if memory needs to be ramped up.
I.e. for Xeon, 2+1+1 may significantly differ in speed from 1+1+1 or 2+2+2. Well of course if the task requires memory handling.
A single stream tester may not notice this. But if there are several serious tasks, the difference should be noticeable.
Phenom II X3 pleasantly surprised me. And Istanbul - that's the problem, it uses outdated memory. Although it is the newest brick for workstations.
This was going on 4 years ago, until Intel changed the P4 architecture to Core. With Core- well it was still possible to compete, losing out a bit.
Some may not agree but Phenom II X3 is the newest one in AMD and the Core is already the previous generation.
So, you have to overpay to Intel, if you want to be faster. Marketing, dammit!
P.S. The vidya has slightly improved the result to 8:01 (481s). Although it's not too clear why: you've got plenty of memory.
The point here is not too much or not enough memory, but situations may occur (and do occur!) when one and the same physical memory module is accessed by both processor and "graphics card" - so someone has to wait his turn. And of course the processor has to wait in turn - which slows it down accordingly.
At my time I have taken the cheapest video card (the same GF6600 - for 900 rubles) to eliminate this effect. However as evident by test acceleration is less than 2% - I can't estimate error of measurements, but it's about the same.
The speed of the memory and the number of controller channels involved will certainly have an effect. But of course not directly proportionally.
Thanks for the tests! Very interesting.
By the way, since we apparently don't have Core i7/i5 owners, maybe you could find a way to test Xeon with 1 and 2 memory channels?
Got the data in the table.
I think it's quite obvious now "who's the boss" - Intel Core ix (represented by Xeon).
By the way, since we apparently have no Core i7/i5 owners
Core i7 is available. I'll email his owner to look into it here. I'll give him a link to the second script so he won't get confused and explain the testing and results presentation conditions.
A Core i7 is available. I'll email his owner to look into it here. I'll give him the link to the second script right away, so he doesn't get confused, and explain the test conditions and presentation of results.
Great, thanks!