chintandhandha :
Intel Launched Xeon Processors as its first step towads multi-core processors. It was only on xeon that intel implemented the six core technology while most of the users were on either core2duo or core2quad. These processors are mainly oriented for servers which require more processing power and also use more cores at the same time.
The reason the first Intel six-core CPUs were Xeons is because Intel at the time had no client dies with more than two cores. The Core 2 Quads were simply two Core 2 Duo CPUs tied together over a shared frontside bus. The client chipsets like the i975 and X38 could not support splitting the frontside bus more than two ways, so there was no way to put more than four cores on one die without making a new die. Nehalem with its monolithic quad-core dies was due to be introduced soon and Intel probably didn't want to go through the trouble of making a new die for the Core 2 line when it was going to be EOLed soon. So they made their six-core CPU as a big monolithic beast with three dual-core Core 2 modules bolted to a huge L3 cache and sold it for thousands of dollars as the Xeon MP 7400 series. Apparently the high selling price and relatively low shipping quantity of the Xeon MP 7400s let Intel make this large and likely difficult to manufacture CPU without losing a lot of money in the process.
Compare this to AMD's approach. They also introduced their first six-core CPU as a server part, but they basically just expanded the current quad-core Phenom II-class die to house two more cores. The resulting die was a lot smaller than Intel's Dunnington and easier to make. They then went on to use that general die in a wide variety of other parts, ranging from all of the currently-selling Opteron 4100 and 6100 CPUs to the Phenom II X6s.
For a server PC speed is all that matters, they do not need any special graphics or need to run any hi-def games. The main aim of these processors is to give better performance to the servers. That is why they are this expensive.
Processor speed is not all that matters to server operators. Reliability is extremely important to server operators, as is energy efficiency and memory capacity. Price is also a consideration as well. Not all server CPUs are all that expensive, either. AMD starts their Opteron line out at $99, and you can get entry-level Xeons starting at about $200 each. The reason why some are expensive is because they are either very difficult to make and/or the maker feels they can charge that sum of money for those parts. Servers are typically bought by companies with fairly large pockets, so the amount buyers are willing to pay is generally much higher than for desktop and laptop CPUs. That's why you see otherwise identical CPUs being sold at a premium based on the number of sockets they can support.
To inform you about how these processors came into being. It all started with HT(hyper-threading) which means actually 1 core but 2 logical cores. Then the idea came that as the temperatures on the CPU are a big limitation to increasing the clock speed, it would be better to increase the number of cores on the CPU.
Wrong, the reason HyperThreading came into use was to try to keep the long pipeline of the NetBurs (P4) based-CPUs utilized. Pipeline stalls and flushes were notoriously painful on these CPUs and having a second thread to utilize unused CPU resources was a way to wring more performance out of the poorly-designed CPU microarchitecture. HyperThreading came about well before the "clock speed wall" was hit as HyperThreading was present in the very first NetBurst chips but Intel didn't realize that its P4s weren't going to hit their projected 10 GHz target until a few years later.
What you are describing is why
multi-core CPUs came into play. That was about four years after HyperThreading first debuted.
This means the processes get distributed on different cores enhancing the performance even at a lower clock speed, the better speed was also given by the increase in cache. The more the cache the faster the speed.
Not necessarily. The performance also depends on the latency of the caches as well. The Pentium 4 Prescotts were available in two versions, one with a 1 MB L2 cache and one with a 2 MB L2 cache. Performance was overall similar since the latency on the 2 MB cache was higher than that on the 1 MB cache.
Now choosing a CPU needs a lot more research than the old times. We have to look at the number of cores, the cache, clock speed and the speed of the memory supported by the CPU.
It really varies by the manufacturer. You also need to be specifically aware of what features are enabled and disabled on Intel CPUs since they like to disable things like hardware virtualization, HyperThreading, and Turbo Boost on their lower-end CPUs. That and they have two different desktop sockets as well.
There are a lot more aspects to consider than just walking in the shop and asking for a Pentium 4 like in old times.
Even back then it wasn't that simple. There are a lot of Pentium 4s, and only one line of them was worth a crap, the Northwood. There were quite a few people that accidentally bought a dog of a Willamette instead of a Northwood because they didn't know how to tell the two apart. The Willamettes stank compared to Pentium III Tualatins and Athlon XPs, while the P4 Prescotts and Cedar Mills were notably inferior to the Athlon 64s. You had to be somewhat intelligent and know what's going on to not get stuck with an inferior CPU even back in the Pentium 4 days.
Now my friend the first thing that is a misconception is that cores will multiply your clock speed or the cache. It is not so. A six core at 3.6 Ghz is clocked at 3.6 GHz and this clock speed is distributed among the six cores so does the cache.
Not quite. All of the cores on a multi-core CPU being run at full load will run at the exact same speed. Core 1 runs at 3.6 GHz, so do cores 2, 3, 4, 5, and 6. You do get more cache with more cores, since each core has its own L2 cache and adding more cores also means you add more cache.
It is not 3.6 X 6 Now the question arises, why is it done this way,even a single core could do it? The answer is that maybe there is a heavy process utilizing one core, the other slows down and so does your computer, whereas when you have more number of cores the process runs on one core keeping the others free and hence allowing you to do multi-tasking. It does not speed up anything, it just enhances the way the processor has to be used. More cores=more multi tasking.
Multi-core CPUs actually do speed up programs that are single-threaded compared to a single-core CPU. The single-threaded program can completely peg one core at 100% and the other cores will run the other tasks the computer has to run, which can use up some CPU time. If you ran it all on a single-core CPU, both the heavy program and the background tasks have to run on that one CPU, so the heavy program will not be able to use 100% of the CPU and thus will run a little slower. Note that this speedup really only happens when going from one core to two cores; running a single-threaded program on a dual-core CPU versus a quad-core CPU of the same clock speed will be identical.
But here comes the sad part if you have a program that can utilize only 1 core at a time and you have a six core the clock speed provided to that program will be only 1/6 of 3.6. Hmmmmm... so that means a quad core with 3.2 will provide more than a six core with 3.6, hence enhancing the speed of the program.
Absolutely incorrect. Like I said above, a six-core CPU rated at 3.6 GHz can run
all of its cores at 3.6 GHz. A single-threaded program running on that 3.6 GHz six-core CPU will run faster than on a 3.2 GHz quad. The one active core on the 3.6 GHz six-core runs at 3.6 GHz, while the one active core on the 3.2 GHz quad-core runs at 3.2 GHz. 3.6 GHz is faster than 3.2 GHz, and everything else being equal, the program will run faster on the higher-clocked six-core.
Now a six core wont be harmful now because there are mny programs that do use multi cores but the question is do you need it? There are CPUs in the market with 12 cores but there is only limited use of them as they are expensive. As i said earlier you need to identify your needs and then decide, consider a budget and then purchase according to your needs. I say this unlike others because i do not know you, i do not know what you are going to do with it, but you do so this is your decision.
The 12-core CPUs are server CPUs, and you can get one of them for less than a six-core Core i7 970. The biggest reason you would not want to use one on the desktop is that their clock speed is low and that most desktop programs don't currently take advantage of more than 2-4 cores right now. The one 12-core unit that costs less than the i7 970 (Opteron 6168, ~$750) runs at 1.90 GHz. It will perform very well at programs that use a lot of threads (such as most server applications) but will perform like a CPU from five years ago on single-threaded programs and perform like a $100 modern CPU on tasks that use only a few threads. There are some consumer programs like video encoding that will happily use all 12 threads and perform very well on the Opteron 6168, but others like most games will only use 2-3 threads and will run more slowly than an $80 Athlon II X3 due to its low clock speed.
Facebook
Twitter
Instagram