Need a new CPU ??? - page 8
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
Apparently Win7 does distinguish them - for optimum planning.
As far as I know, no.
>> It doesn't matter at all whether it is virtual core or not. Most importantly, the virtual core does not give nearly as much performance boost as the physical one. Moreover, there are programs where Hyper Trading even reduces performance. So even if we divide a program by 8 threads, additional 4 virtual cores will be of little use. Although much depends on optimization of your program for multithreading calculations. But in any case, the virtual core will not give you a great gain. If it will not slow down too...
Intel, at the beginning of next year, is already promising six-core processors, but under socket 1366. I hope they will not be too expensive.
http://www.ixbt.com/cpu/archspeed-2009-3.shtml
http://www.ixbt.com/cpu/archspeed-2009-4.shtml
Here you can read in detail about the results and usefulness of Hyper Trading. You should not expect to get at least 8 cores and get a significant performance gain. The average performance gain from Hyper Trading on a 4 core CPU is only 10%. This is exactly average, in 11 programs Hyper Trading decreased CPU performance. So you still need to look at what programs you are using for Hyper Trading as if it were not even worse...
When HT is enabled, ALL cores become virtual. It's no longer there that this core is physical and this one is virtual.
But the fact that HT behaves differently is true. Then, don't forget that caches of all levels are halved per core, which is also not good. Anyway, it all depends on a concrete code. In one of threads we have already discussed cache is important for MT4 and according to modest statistics HT mode slows down. But this is 4, which is single threaded.
In short, there is no subject for discussion - we are talking about nothing.
Intel, at the beginning of next year, is already promising six-core processors, but under socket 1366. I hope they will not be too expensive.
The 1366 socket is by definition expensive by itself. Especially if you consider the cost of the platform as a whole. Expensive X58 chipset motherboards, you have to install three RAM modules for a three-channel memory controller. The CPUs themselves are positioned as top models and therefore are the most expensive in the line of manufactured CPUs. Besides any hardware, which has just been released, comes with high prices. So I think ... even sure that the price they charge to the full. Well, as always, the presence of 6 cores does not automatically mean a significant increase. You need to optimize the software for the 6 cores. Not all programs are optimized for 4 cores either.We are well aware of the fakiness of hyperthreading.
Don't worry, we will run practical tests and if it turns out to be realistically better (quite expected) to run no more tester threads than the number of physical cores, then that is what the terminal will recommend by default.
The 1366 socket is by definition expensive by itself. Especially if you consider the cost of the platform as a whole. Expensive X58 chipset motherboards, you would need three RAM modules for a three channel memory controller. The CPUs themselves are positioned as top models and therefore are the most expensive in the line of manufactured CPUs. Besides any hardware, which has just been released, comes with high prices. So I think ... even sure that the price they charge to the full. Well, as always, the presence of 6 cores does not automatically mean a significant increase. You need to optimize the software for the 6 cores. Not all programs are optimized for 4 cores.The point is that the optimisation is perfectly parallel. Roughly speaking, one run is one thread. So I'm not expecting any tricks here. The more cores the better. If it's cheaper then it's better to have a look at AMD 6-core units. Not considering Intel HT, it's faster than 4 cores of AMD, it's not 1.5 times difference.
And the 1366 platform is a server platform from the very beginning. Hence 3 channels of RAM and so on. Including the price.
The point is that the optimisation is perfectly parallel. Roughly speaking, one run is one thread. So I'm not expecting any tricks here. The more cores the better. If it's cheaper then it's better to have a look at AMD 6-core units. Leaving aside Intel HT, it's faster than 4 of their cores - not 1.5 times the difference.
The 1366 platform is a server platform from the beginning. Hence the 3 channels of RAM, etc. Including the price.
To AMD it is possible to have a closer look, but when they will appear the 6-core from AMD ... as it seems they will be released later than the Intel's. And taking into account that AMD completely loses in top sector, we can assume that 6-core from AMD will just overtake the current Kore and7 on performance. The same way Phenom2 has caught up with Intel quads which have been on the market for 2 years already. Hence the price...they will probably cost less than Intel's CPUs.When HT is enabled, ALL cores become virtual. It's no longer there that this core is physical and this one is virtual.
But the fact that HT behaves differently is true. Then, don't forget that caches of all levels are halved per core, which is also not good. Anyway, it all depends on a concrete code. In one of the threads we already discussed for MT4 - cache is important and, according to modest statistics, HT mode slows down. But this is 4, which is single threaded.
>> Anyway, there is no subject to discuss, we are talking about nothing.
They become virtual only nominally. The real work is done by the physical cores. Hyper Trading is not really a multi-core technology, it is a technology to optimize the physical core load. In terms of physical effect HyperTrading does not provide any second core, it only provides for optimal loading of the physical core, due to the fact that it can utilize idle core resources in order to launch one more computational thread. Hyper Trading is not a separate kernel per se, it just runs on the kernel. There is no way a kernel can jump over its own performance level. That is to say, if a core is giving peak performance of say 100 units. There is no way Hyper Trading can surpass the physical limit of the core and give you at least 101 units of performance. As I said it can only optimize and maximize the core load, thereby avoiding downtime of individual processor processing units and thus giving a power boost. If a core has, say, 70% utilization then Hyper Trading can help load the remaining 30%. But it will not give more power than core can give. This is why there is only 10% average gain from Hyper Trading. It is precisely for this 10% that it helps to overload cores with work.