Okay, so often we see people come in talking about how great their hexacore Bulldozer chip is, especially when clocked at a whopping 4 GHz. Now, we all know that this really isn't that great. The most common response is something like this:
"Cores != performance, GHz != performance."
Now, this is correct. It just doesn't teach these people why what they say is incorrect. This is what I have a problem with, because if they don't know why Cores and GHz do not equal performance, then they have no clue if the person telling them this actually knows what they're talking about.
I think that a proper response to these kinds of messages should explain why cores and GHz do not directly equate into performance. Here are some examples that I've typed up:
In-depth explanation:
You can't just say that a processor is faster than another because it has a higher clock speed and more cores. Clock speed is purely subjective; if one processor has a higher IPC (Instructions Per Cycle), the other processor's clock speed will need to be higher in order to match the raw performance of said processor. Cores are also an ineffective way to judge the speed of one chip vs the other. Most applications don't even use two cores, so what good does adding eight of them do? Absolutely nothing.
Take for example the quad-core i5-2500k clocked at 3.2GHz, and the hexacore FX-8150 clocked at 3.6GHz. To someone who does not know what they are talking about, the FX-8150 is the obvious choice; it has more GHz and cores!
No. The IPC of Bulldozer (The FX-8150's architecture) is significantly lower than the IPC of Sandy Bridge (the i5-2500k's architecture). Combine this with the fact that most applications don't even use two cores, and you have a recipe for disaster by the name of "FX-8150". AMD is simply packing many cores and a high clock speed with a low IPC into one chip, whereas Intel is getting a very high IPC with a moderate clock speed and an average amount of cores. The i5-2500k is the clear winner in games and every day applications.
The shorthand:
Cores and GHz do not equate to performance when comparing CPUs, because some CPUs can do more work in one clock cycle (where GHz is the clocks per second) than others. Most applications don't use more than one core.
Analogy by BC_Programming:
Think of clock speed as rotation speed. But then to consider the actual execution speed of the processor's instructions as dictating the size of the wheel. A larger wheel will cover more ground in a single rotation than a smaller one.
Analogy by Shpaget:
A person with a shovel can shovel rapidly, but a digger with a backhoe, no matter how slow, will out dig a shovel.
Feel free to copy and paste those paragraphs. Lets actually teach people why they are wrong, rather than telling them that they are wrong.
Telling them what you suggested still doesn't help them though. Only someone who has expertise on the subject will understand what you have said.
It's better than just telling them outright that it doesn't translate directly to performance. This shows that you do know what you're talking about, and you can back up what you're saying.
It's better than just telling them outright that it doesn't translate directly to performance. This shows that you do know what you're talking about, and you can back up what you're saying.
Here is my honest opinion on the matter. If someone believes More GHZ = More performance, I will let them believe so, at the end of the day, it is not me spending the money.
It's far easier to type the shorthand than give an explanation. If the person wants an explanation, we can then tell them why.
This is the most logical choice, since there isn't exactly a "short" explanation to why, aside from "better architecture" which is usually what people add after saying ghz/cores =/= performance.
Well, the interesting thing to note that, while, clock speed and amount of cores do say nothing about performance, the statement "More GHz =/= more performance" isn't correct either. If it were correct, what would the whole point of overclocking be? I think the most scientifically correct thing to say would be "You can't just compare processors by their clock speed or amount of cores, that's not how it works".
CPU performance in a certain application is, quite simply said, difficult to explain. Benchmarking software tells you only part of the story, and I'm going to agree with BC here that it's practically useless, except for allowing people to drool over a number. Real performance is hard to measure. Framerates in games only tell you part of the story, because games usually do not use more than two cores, sometimes only one. If playing games is all you want, then that's not a problem, but some people use applications that are more CPU intensive than games.
Usually only the more experienced people will be able to name the best CPU for a certain situation. As much as all of us woud like to educate the people buying a computer, it's almost impossible to do, because the subject is quite complicated, and the more time you invest in researching it, the more complicated it becomes.
Another analogy would be to think of clock speed as rotation speed. But then to consider the actual execution speed of the processor's instructions as dictating the size of the wheel. A larger wheel will cover more ground in a single rotation than a smaller one.
Not a good example, considering a pentium 4 has 1 core and 2 threads, whereas the i3 has 2 cores and 4 threads.
EDIT : Pentium 4 dual core? Wouldn't that be a pentium D? If so, that has 2 cores and 2 threads
"Cores != performance, GHz != performance."
Now, this is correct. It just doesn't teach these people why what they say is incorrect. This is what I have a problem with, because if they don't know why Cores and GHz do not equal performance, then they have no clue if the person telling them this actually knows what they're talking about.
I think that a proper response to these kinds of messages should explain why cores and GHz do not directly equate into performance. Here are some examples that I've typed up:
In-depth explanation:
The shorthand:
Analogy by BC_Programming:
Analogy by Shpaget:
Feel free to copy and paste those paragraphs. Lets actually teach people why they are wrong, rather than telling them that they are wrong.
Want a place to advertise your Minecraft server? try MyMCStatus.net now!
It's better than just telling them outright that it doesn't translate directly to performance. This shows that you do know what you're talking about, and you can back up what you're saying.
Here is my honest opinion on the matter. If someone believes More GHZ = More performance, I will let them believe so, at the end of the day, it is not me spending the money.
Want a place to advertise your Minecraft server? try MyMCStatus.net now!
This is the most logical choice, since there isn't exactly a "short" explanation to why, aside from "better architecture" which is usually what people add after saying ghz/cores =/= performance.
CPU performance in a certain application is, quite simply said, difficult to explain. Benchmarking software tells you only part of the story, and I'm going to agree with BC here that it's practically useless, except for allowing people to drool over a number. Real performance is hard to measure. Framerates in games only tell you part of the story, because games usually do not use more than two cores, sometimes only one. If playing games is all you want, then that's not a problem, but some people use applications that are more CPU intensive than games.
Usually only the more experienced people will be able to name the best CPU for a certain situation. As much as all of us woud like to educate the people buying a computer, it's almost impossible to do, because the subject is quite complicated, and the more time you invest in researching it, the more complicated it becomes.
so its better than the i3-2100 rite?!?!
Most P4s only had one core one thread
Yeah, there was a P4 with HT, but it was more expensive and there was only one model that had it (IIRC)