OSDev.org

suthers wrote:They may still recover, but its still quite bad how a proc they are releasing now can't out perform clock for clock a one year old proc
Jules

There's more to a processor than it's clock rating.

I was an AMD fan as a teenager, but I do like the Intel's mobile processors (and they keep cool under normal use (playing games != normal use )).

At the end of the day, a similarly marked processor from AMD or Intel will probably perform on par. I know there will be bench marks and people will argue ("but you get an extra 5 frames per second with an AMD" or "compressing a 1GB archive will take 20 second less with an Intel") but the difference is so insignificant (100 vs 105 fps or 30 minutes vs 30 minutes and 20 seconds) that the end user would probably not even care.

What I meant is if you take a 1 year old C2Q proc, if you take a phenom processor and clock them both at the same speed, the C2Q will come out on top even though its architecture is 1 year old and phenom hassent even come out yet...
Also even though it performs better per clock than the phenom it can also achieve higher clock frequencies.
and the price range is about the same...
Jules

I probably should give my oppinion too, so here goes.

I have never cared much for amd and here is one of the reasons.
I have had alot of different cpus in my time and every single amd cpu has, sooner or later, went dead on me, but with intel only one has, and that was really my own fault, safe to say i dont take very good care of my hardware. When you computer keeps crashing duo to a faulty cooler, you should really do something about it, but still i took allmost a year for the cpu to give up and it didnt really give you, it could still boot, but after 5 minutes it crashed, nomatter the temperature.

IMHO the core 2 architecture is a masterpiece and i doubt very much that AMD will ever catch up with intel, but if they will ever go out of buisness i really dont know.

But when all is said there is allso good thing about AMD (i think) and i look forward to the day when intel buys AMD.

Edit:
About overclocking, i dont think its relevant in this context.

I'm not saying it will be a good thing if AMD is bought by Intel because they will stop being driven by competition to do better...
So proc improvement will probably slow down and I don't see why overclocking isn't relevant in this context.
Jules

suthers wrote:I don't see why overclocking isn't relevant in this context.

I dont think it is relevant because you by a product, but expect to get something else.

fx.
i buy a 1000mhz cpu, but what i really want is a 2000mhz cpu, and of course thats just fine, but its no longer tha same product.

Yes its cool that you can tune your car, but if it was what was intented, you woldnt have to do it your self. The factory wold have tuned it for you, and it would no longer be a tunes car, but a faster faster car.
There is usually a reason why they dont do this for you.

Yes i know you can you can overclock some cpus to amazin extends if like puer liquid nitrogen on to it... But what has that got to do with anything? It is not the cpu that are amazing, its the cooling methode.

When all is said i dont even believe its necasery (i just cant spell that word!) to overclock anything anymore. People more or less do it just to show of if you ask me.

Some CPU's can overclock loads with there standard fans...
Im just thinking that we should also look at the maximum ability of there products intel's new penryn CPU's can overclock on water colling to near the maximum frequency ever reached by an AMD proc and still give more performance per clock than AMD procs.
Jules

Hi,

suthers wrote:Im just thinking that we should also look at the maximum ability of there products intel's new penryn CPU's can overclock on water colling to near the maximum frequency ever reached by an AMD proc and still give more performance per clock than AMD procs.

You're right - as OS developers we should look at the maximum reliability of these products, and try to detect potential reliability problems (like overclocking) and either fix the problem, refuse to boot or display huge warnings.

This would be especially useful when some con-artist is trying to sell a computer as something faster than what it actually is; or even when some stupid script kiddie reads a web-site and thinks they can get more bang for their buck without realising there's consequences.

As a bonus, it'd also be nice if the OS could underclock things like RAM, hyper-transport links, etc to save power when the OS is mostly idle.

Of course it'd be extremely difficult to do in practice - you'd need special code for each chipset (but then, I'm considering doing that anyway to avoid the huge "ACPI mess" so perhaps it could be done for some chipsets).

As for deliberately overclocking - why bother? Tweaking the hardware to get dubious "improvements" seems fairly silly considering that most people are running slow and bloated software on top of slow and bloated OSs. Improving the software can make a huge difference in performance without risking unreliability, but then I guess it takes real knowledge to write fast and efficient software (it's beyond the abilities of the average script kiddie).


Cheers,

Brendan

Implementing that in to an os would certainly be something, but im not quite sure that is the argument suthers had in mind. Maybe im mistaken.

As for deliberately overclocking - why bother? Tweaking the hardware to get dubious "improvements" seems fairly silly considering that most people are running slow and bloated software on top of slow and bloated OSs. Improving the software can make a huge difference in performance without risking unreliability

this isnt really true -- most people can get good (20%-60%) overclocks on standard cooling, with 100% reliability (intel CPUs overclock much better than AMD CPUs in general)

this is sort of like saying 'why should i type 80 words-per-minute, when my keyboard only has 104 keys on it... it wasnt meant for that' which really doesnt make any sense -- the only thing the 'rating' marked on the chip says, is how much the worst chip produced on the assembly line can do -- in fact, when intel sells a 3GHz chip, they test it at 3.6GHz first, if it fails, then they sell it to the large computer sellers (such as dell, hp, and lenovo)

and all of the large players in sales of high-performance PCs overclock the computers significantly (falcon northwest, alienware, voodoo (now hp) -- basically anyone who sells computers over $2000)

Hi,

JAAman wrote:

As for deliberately overclocking - why bother? Tweaking the hardware to get dubious "improvements" seems fairly silly considering that most people are running slow and bloated software on top of slow and bloated OSs. Improving the software can make a huge difference in performance without risking unreliability

this isnt really true -- most people can get good (20%-60%) overclocks on standard cooling, with 100% reliability (intel CPUs overclock much better than AMD CPUs in general)

You're definition of "good" (fast as possible and not too likely to crash?) is probably a lot different to mine (fast enough and guaranteed not to crash, even in the most unusual corner cases).

JAAman wrote:this is sort of like saying 'why should i type 80 words-per-minute, when my keyboard only has 104 keys on it... it wasnt meant for that' which really doesnt make any sense -- the only thing the 'rating' marked on the chip says, is how much the worst chip produced on the assembly line can do -- in fact, when intel sells a 3GHz chip, they test it at 3.6GHz first, if it fails, then they sell it to the large computer sellers (such as dell, hp, and lenovo)

Intel test it at 3.6 GHz and sell it as 3 GHz if just in case there's problems that their testing didn't detect. The same probably applies to other manufacturers (RAM, GPUs, etc).

Some manufacturers overclock it, then test to see if it works by installing Windows on it, then sell it to the nearest sucker.

JAAman wrote:and all of the large players in sales of high-performance PCs overclock the computers significantly (falcon northwest, alienware, voodoo (now hp) -- basically anyone who sells computers over $2000)

These aren't high-performance PCs - they're low reliability games machines for people who's data is worthless. For high performance (and reliability) you'd want to look at servers from companies like IBM and HP. You can usually tell the difference by the cases (e.g. boring looking plain grey vs. bright colours intended to attract teenagers with the "extremely important" interior lighting).


Cheers,

Brendan

Brendan wrote:You can usually tell the difference by the cases (e.g. boring looking plain grey vs. bright colours intended to attract teenagers with the "extremely important" interior lighting).

...and lots of places to place loud, whirring fans to make your workplace warm and cozy.

I'd like to add another definition of a "good" CPU: performance / heat ratio. I've never employed a CPU that required special cooling, or cooling that could be heard over a casual conversation, in a system that sits on / under the desk as opposed to sitting in the basement. I haven't had a server system in the basement yet, but when I do (which is about 4 months from now ), it won't be "the fastest" either because of cost, longevity, and reliability considerations.