| 1 |
On 10/11/2015 20:37, Stanislav Nikolov wrote: |
| 2 |
> |
| 3 |
> |
| 4 |
> On 11/10/2015 08:17 PM, Mick wrote: |
| 5 |
>> On Tuesday 10 Nov 2015 17:47:08 Stanislav Nikolov wrote: |
| 6 |
>>> Dear Gentoo users, |
| 7 |
>>> I'm building a new PC. I have a budget of ~$550-$650. No GPU, no special |
| 8 |
>>> case (I may use a card box), not even a hdd or ssd. So, as you can see, |
| 9 |
>>> it's pretty much "get the best CPU and mobo/ram that are compatible with |
| 10 |
>>> it". The problem is, which is the best one. By "best" I mean to compile |
| 11 |
>>> shit fast. My laptop with 3rd gen i5 compiles firefox for 40 minutes on |
| 12 |
>>> average. |
| 13 |
>>> |
| 14 |
>>> The most expensive Intel CPU is the skylake i7-6700k. But is it the best? |
| 15 |
>>> Is there something from AMD that will perform even better? I can't find |
| 16 |
>>> any benchmarks with AMD/Intel CPUs. And how much does the mobo matter? |
| 17 |
>>> Will a cheap $30 400W PSU power that thing? |
| 18 |
>>> |
| 19 |
>>> Thanks |
| 20 |
>> I don't (yet) own a i7-6700k, but my 6 year old laptop with (1st generation) |
| 21 |
>> i7 Q720 @1.60GHz takes slightly less than yours: |
| 22 |
>> |
| 23 |
>> Sat Oct 3 14:35:40 2015 >>> www-client/firefox-38.3.0 |
| 24 |
>> merge time: 36 minutes and 53 seconds. |
| 25 |
>> |
| 26 |
>> Fri Nov 6 09:10:06 2015 >>> www-client/firefox-38.4.0 |
| 27 |
>> merge time: 38 minutes and 8 seconds. |
| 28 |
>> |
| 29 |
>> |
| 30 |
>> In contrast a year old AMD A10-7850K APU is significantly faster: |
| 31 |
>> |
| 32 |
>> Sat Oct 3 19:40:48 2015 >>> www-client/firefox-38.3.0 |
| 33 |
>> merge time: 17 minutes and 42 seconds. |
| 34 |
>> |
| 35 |
>> Fri Nov 6 08:41:02 2015 >>> www-client/firefox-38.4.0 |
| 36 |
>> merge time: 18 minutes and 18 seconds. |
| 37 |
>> |
| 38 |
>> |
| 39 |
>> I would also be interested to see compile times of more modern i7s and FXs, |
| 40 |
>> but bear in mind that in single core operations Intel is these days |
| 41 |
>> significantly better than AMD. |
| 42 |
>> |
| 43 |
> So, I shouldn't prepare for a 8x times faster compile time... :( |
| 44 |
> |
| 45 |
|
| 46 |
|
| 47 |
|
| 48 |
I can't help but think you are approaching this from the wrong perspective. |
| 49 |
|
| 50 |
Why exactly are you using compile times as your sole criterion? Are you |
| 51 |
building a compile farm for Ubuntu? Running continuous integration tests |
| 52 |
for LibreOffice [on a $600 budget in a cardboard box :-) ]? |
| 53 |
|
| 54 |
Or do you want emerge world to get it over with quicker? |
| 55 |
|
| 56 |
If the latter, you better rethink your priorities. In computing terms, |
| 57 |
compilation is a rare event; launching apps is a common event; and |
| 58 |
writing to the disk happens all the time. Optimize for the common case. |
| 59 |
|
| 60 |
A CPU never works in isolation, it is always part of a much larger |
| 61 |
system, like disks, RAM and all possible kinds of I/O. The best CPU on |
| 62 |
the market plugged into a POS motherboard will perform on emerge world |
| 63 |
like a piece of shit - it will follow the weakest link. |
| 64 |
|
| 65 |
If you want to build a compiling machine, buy the best collection of |
| 66 |
stuff that works together well and still fits the budget. If you want a |
| 67 |
machine that you can use and be happy with, ignoree the temptation to |
| 68 |
must have the biggest baddest fastest CU (you will never get to use all |
| 69 |
that big bad fast) and invest rather in gobs of RAM and an SSD. Remember |
| 70 |
that apps are launched many times more than they are compiled. Or put |
| 71 |
another way, sacrifice compilation times t get something you can use. |
| 72 |
|
| 73 |
|
| 74 |
|
| 75 |
-- |
| 76 |
Alan McKinnon |
| 77 |
alan.mckinnon@×××××.com |
Archives