Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
AMD Bulldozer 8 core cpus and Gentoo/gcc/ potential
View unanswered posts
View posts from last 24 hours

Goto page 1, 2  Next  
Reply to topic    Gentoo Forums Forum Index Kernel & Hardware
View previous topic :: View next topic  
Author Message
wrc1944
Advocate
Advocate


Joined: 15 Aug 2002
Posts: 3435
Location: Gainesville, Florida

PostPosted: Fri Oct 14, 2011 2:54 pm    Post subject: AMD Bulldozer 8 core cpus and Gentoo/gcc/ potential Reply with quote

I was eagerly awaiting the release of the new AMD Bulldozer Zambezi 8 core "FX" cpus, but all the hardware/performance/reviewing sites and forums are trashing it big time.

The problem seems to be that performance is outstanding with multi-threaded apps, even beating the intel i7-2600k flagship cpu.

However, with single-threaded apps it sucks badly, not even competing with the previous generation AMD x6 Thuban cores, and badly losing to even the 2500k intel in single threaded apps. Power consumption is also way too high. Even the price/performance comparison is losing, according to most reviews I've seen (hundreds). There's enough of them to get a good idea of what the FX's are capable of- and the consensus is very disappointing, so far. Nobody is recommending this long-awaited chip, as opposed to investing in an intel in the same price range- in fact, at this point most would recommend X6 thubans over any FX zambezi.

Basically, the Zambezi seems to be a server chip, being marketed as a "desktop/general purpose" cpu, but the initial reports really seem to argue this is not working at all. The AMD zone and many other sites have long detailed technical discussions of all the details, pros and cons, regarding L1, L2, and L3 caching problems causing huge cache misses and having to re-fetch from the cache. Some are saying the entire arch is "completely broken" in this respect.

All that said, most of the above mentioned reviewing has apparently been done on Windows 7 systems, and no real mention of Linux systems, yet. There's talk that the windows scheduler has problems with the new zambezi arch, and that currently most common desktop software is really not optimized for multi-threads.

This brings me to my Gentoo (and Linux in general) questions:
Since in Gentoo we compile everything ourselves using gcc, and have scheduler/kernel/USE and cflag flag options, I'm wondering if the performance problems with single-thread apps might be less of a problem in Linux in general, and especially with Gentoo?

I realize that just re-compiling an app not specifically coded by its creator for multi-thread probably can't be completely transformed by gcc and multi-thread kernel options/patches, but this type of thing was being mentioned as being potential but limited "fixes" for windows7 and the zambezi cpus.

I've always run AMD cpus, and am currently using the athlon II X4 640 propus core (basically a phenom II just without the L3 cache). I've noticed that during long emerges all 4 cores are always pegged at 100% usage for the duration, which I assume means portage/emerge is highly designed for multi-threading.

Anyone using or planning to get a new AMD FX cpu, and have any thoughts/experiences on all this?

I'd really like to know if recompiling actually makes any difference as regards to single thread vs. multi-thread behavior/performance. In other words, would recompiling a linux app originally coded for single-thread but using march=native, USE="threads", and any kernel threading options/patches render the code more suited for the 8 core zambezi which has superb multi-thread performance, thus justifying getting a zambezi cpu instead of an X6 thuban, or simply forgetting AMD and moving to the intel Sandy/Ivy bridge i7-2600k platform. Guess I could keep the Athlon II X4 and wait for the AMD "Piledriver" and hope the new arch problems are fixed then. :roll:

If I really thought running Linux and Gentoo on these cpus would perform better than what's being reported on Windows7, I'd probably get one, as I'm using Gentoo Linux 99% of the time anyway.
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11


Last edited by wrc1944 on Sat Oct 15, 2011 2:00 pm; edited 1 time in total
Back to top
View user's profile Send private message
DaggyStyle
Watchman
Watchman


Joined: 22 Mar 2006
Posts: 5909

PostPosted: Fri Oct 14, 2011 4:16 pm    Post subject: Reply with quote

AMD decided to follow the "more core, better" notion, that is why single core performance sucks, today, most of the apps are in transition from single core to multi core.
basically, the bulldozer arch is more server oriented the desktop oriented, this gives them better performance in the server section with less price against intel's sb server, add to the fact that intel doesn't seems to be able to solve the chipset but in their servers (same one which they solved for the desktop), thus then are shipping diskless systems.

the bottom line is, what are your needs? if you can live with the crappy performance of the programs you need which aren't mt optimized, then got for bulldozer, if not, then either wait for programs to finish the transition, wait for newer AMD chips or got intel.

that my 0.02 cents.

answering your question, compilation method usually doesn't affects the way a single threaded app works, different compilers can affect it.

imho, if some dev doesn't plan to port his sw to mt, then it isn't worth while to use it.
also, I wouldn't trust much windows benchmarks, most of the programs are either compiled with icc or are intel oriented.
_________________
Only two things are infinite, the universe and human stupidity and I'm not sure about the former - Albert Einstein
Back to top
View user's profile Send private message
krinn
Watchman
Watchman


Joined: 02 May 2003
Posts: 7470

PostPosted: Fri Oct 14, 2011 11:52 pm    Post subject: Reply with quote

there's way too much things bad in that cpu gen:
- real too high power eaten
- real too high tdp (seriously who will use that as server except maybe real high server with dedicated cooling system, but you must be mad to use more than one cpu of that type with a fan system)
- real too weak performance, that cpu cannot even seriously manage to beat even other amd cpu !
- that's not real 8 cores, bullshit marketing : these are just 4 cores with a "name it as you wish : advanced, tweak, dirty..." HT clone
- the "windows is not ready for it" explain nearly made me pee in my pants :
-->someone might build a scheduler that will be aware of the turbo problem and try to assign more tasks to one core instead of spliting them to more cores so the cores might get higher turbo benefits : but this tweak will not proof the arch concept is good, just that running a task on fewer cores made them run at higher clock speed, looks like amd discover more clock speed increase calc speed :p
(this could be improve by scheduler yes, and this will of course also improve corei7 class cpu that also could turbo their core speed when fewer are in use)
-->and because of the 2 units in one core if you gave a task that depend on another task amd guys expect the sheduler to gave them both at the same core because it will be done faster, yeah they expect the scheduler to be the magic guesser of what you are about to ask to your cpu.
This is a great feature, but this feature should be thread as it is: if it work a bonus, if not then you see the real power of each core and this should be the default value use to speak about that cpu power, just because the bonus won't happen that much.
What amd is trying to do would be like if intel says that any HT cpu using HT is showing its power, and when not using HT it's because of the scheduler/os/bad luck... and it underrate our cpus, not our fault.
- the price is a joke, and anyone bying that cpu will take a lost in a week, as the cpu prize will fall faster than light.
- when gcc will be aware of the cpu, it might be able to optimize more the branching of thread knowing cache size & cpu specs, but it will always be the scheduler task to assign them. So gcc impact will be low while you can expect random results with tasks assignations.

And as daggystyle said, comparing benchmark in windows programs that are nearly always build with icc compiler isn't a good thing to do vs intel cpu.
The more laughting part is that if you compare program build with icc vs another amd cpu, this time it's legit as both cpu will have the same arms to battle, and this show the bulldozer is more a shovel than anything.

Many people compare that arch with netburst arch that intel design for high clock speed and at end gave weak cpu perf, high tdp, caching missery (too low cache at first, then a big cache help a bit, but too much cache miss and long branching gave high latency ending in poor results).
It's a sad news as the new intel cpu will then get out at an even lower clock speed (yeah weaker perf so they will keep a big marge in perf in case amd start to trouble them later) and they will get out at a higher prize now
Back to top
View user's profile Send private message
DaggyStyle
Watchman
Watchman


Joined: 22 Mar 2006
Posts: 5909

PostPosted: Sat Oct 15, 2011 5:34 am    Post subject: Reply with quote

the intel i7 isn't a real 8 core too... most of intel's cpu are pseudo multi cores
_________________
Only two things are infinite, the universe and human stupidity and I'm not sure about the former - Albert Einstein
Back to top
View user's profile Send private message
<3
Veteran
Veteran


Joined: 21 Oct 2004
Posts: 1081

PostPosted: Sun Oct 16, 2011 2:22 am    Post subject: Reply with quote

DaggyStyle wrote:
the intel i7 isn't a real 8 core too... most of intel's cpu are pseudo multi cores

Yeah but Intel does not market it's virtual cores as a logical one. For Example a core i7-2600K is advertised as a quad core chip even though it has 4 logical cores and 4 virtual ones.
Back to top
View user's profile Send private message
DaggyStyle
Watchman
Watchman


Joined: 22 Mar 2006
Posts: 5909

PostPosted: Sun Oct 16, 2011 5:51 am    Post subject: Reply with quote

<3 wrote:
DaggyStyle wrote:
the intel i7 isn't a real 8 core too... most of intel's cpu are pseudo multi cores

Yeah but Intel does not market it's virtual cores as a logical one. For Example a core i7-2600K is advertised as a quad core chip even though it has 4 logical cores and 4 virtual ones.


can you explain to me the diff between virtual core and logical core?

maybe Intel doesn't but Intel isn't selling the product to the customer, where I live, the i7 are advertised as 8 cores.

also, Intel did it with the first C2D, why can't AMD do that?
_________________
Only two things are infinite, the universe and human stupidity and I'm not sure about the former - Albert Einstein
Back to top
View user's profile Send private message
krinn
Watchman
Watchman


Joined: 02 May 2003
Posts: 7470

PostPosted: Sun Oct 16, 2011 9:41 am    Post subject: Reply with quote

they didn't :)
look at cpu specs from intel, they tell you the #cores and #threads, no lie (this time)
http://www.intel.com/content/www/us/en/processors/core/core-i7-processor/Corei7Specifications.html
Back to top
View user's profile Send private message
DaggyStyle
Watchman
Watchman


Joined: 22 Mar 2006
Posts: 5909

PostPosted: Sun Oct 16, 2011 10:25 am    Post subject: Reply with quote

krinn wrote:
they didn't :)
look at cpu specs from intel, they tell you the #cores and #threads, no lie (this time)
http://www.intel.com/content/www/us/en/processors/core/core-i7-processor/Corei7Specifications.html


I didn't said they lied this time, I said that vendors are selling i7 as octa-cores.
_________________
Only two things are infinite, the universe and human stupidity and I'm not sure about the former - Albert Einstein
Back to top
View user's profile Send private message
wrc1944
Advocate
Advocate


Joined: 15 Aug 2002
Posts: 3435
Location: Gainesville, Florida

PostPosted: Sun Oct 16, 2011 2:37 pm    Post subject: Reply with quote

Perhaps a better way to ask the question might be: Does anyone know which of the main Linux apps are coded for multi-thread, and which are not?

Apparently portage/emerge is multi-threaded at least up to 4 cores, as all of my 4 cores of my Athlon II X4 640 are always at a constant 95-100% usage when doing "emerge whatever."

Thus, so as far as Gentoo linux and all the compiling we do is concerned, the AMD FX Bulldozer 8 core might be an ideal system.

In my case, in addition to the usual Office apps, web Browsers, and Multi-media apps, I'm particularly interested in knowing definitively if any of the Linux pro audio apps like Rosegarden, Ardour, Audacity, Hydrogen, Qsynth, etc., etc. are indeed multithreaded. Anyone know, or know how to find out if any app is single or multi-threads?

That would definitely be a deciding factor in whether or not I'd buy a Bulldozer system right now, or maybe just drop in an on sale AMD Phenom II X6 1100T in my current AM3 system, and wait until Piledriver's out, where hopefully some of the Bulldozer problems will be revised/fixed.
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11
Back to top
View user's profile Send private message
DaggyStyle
Watchman
Watchman


Joined: 22 Mar 2006
Posts: 5909

PostPosted: Sun Oct 16, 2011 3:58 pm    Post subject: Reply with quote

wrc1944 wrote:
Perhaps a better way to ask the question might be: Does anyone know which of the main Linux apps are coded for multi-thread, and which are not?

Apparently portage/emerge is multi-threaded at least up to 4 cores, as all of my 4 cores of my Athlon II X4 640 are always at a constant 95-100% usage when doing "emerge whatever."

Thus, so as far as Gentoo linux and all the compiling we do is concerned, the AMD FX Bulldozer 8 core might be an ideal system.

In my case, in addition to the usual Office apps, web Browsers, and Multi-media apps, I'm particularly interested in knowing definitively if any of the Linux pro audio apps like Rosegarden, Ardour, Audacity, Hydrogen, Qsynth, etc., etc. are indeed multithreaded. Anyone know, or know how to find out if any app is single or multi-threads?

That would definitely be a deciding factor in whether or not I'd buy a Bulldozer system right now, or maybe just drop in an on sale AMD Phenom II X6 1100T in my current AM3 system, and wait until Piledriver's out, where hopefully some of the Bulldozer problems will be revised/fixed.


that is what I've said...

btw, I run gentoo with 8 cores so I can up the bar to 8 cores, I wanted to install gentoo at work but me boss doesn't allow my, shame, I wanted to see how gentoo behaves on a 104 cores, 0.5TB mem beast...
_________________
Only two things are infinite, the universe and human stupidity and I'm not sure about the former - Albert Einstein
Back to top
View user's profile Send private message
mattst88
Developer
Developer


Joined: 28 Oct 2004
Posts: 422

PostPosted: Sun Oct 16, 2011 4:35 pm    Post subject: Reply with quote

wrc1944 wrote:
Apparently portage/emerge is multi-threaded at least up to 4 cores, as all of my 4 cores of my Athlon II X4 640 are always at a constant 95-100% usage when doing "emerge whatever."


The process of compiling is very easily parallelizable since projects consists of a large number of independent C files, which can be compiled independently. It's not portage that's multithreaded, it's that make is able to invoke gcc multiple times, ie: make -j4.
_________________
My Wiki page
Back to top
View user's profile Send private message
krinn
Watchman
Watchman


Joined: 02 May 2003
Posts: 7470

PostPosted: Mon Oct 17, 2011 1:33 am    Post subject: Reply with quote

mattst88 wrote:
wrc1944 wrote:
Apparently portage/emerge is multi-threaded at least up to 4 cores, as all of my 4 cores of my Athlon II X4 640 are always at a constant 95-100% usage when doing "emerge whatever."


The process of compiling is very easily parallelizable since projects consists of a large number of independent C files, which can be compiled independently. It's not portage that's multithreaded, it's that make is able to invoke gcc multiple times, ie: make -j4.


It is because of --jobs that will spawn multiple make :)
Back to top
View user's profile Send private message
mattst88
Developer
Developer


Joined: 28 Oct 2004
Posts: 422

PostPosted: Mon Oct 17, 2011 1:41 am    Post subject: Reply with quote

krinn wrote:
mattst88 wrote:
wrc1944 wrote:
Apparently portage/emerge is multi-threaded at least up to 4 cores, as all of my 4 cores of my Athlon II X4 640 are always at a constant 95-100% usage when doing "emerge whatever."


The process of compiling is very easily parallelizable since projects consists of a large number of independent C files, which can be compiled independently. It's not portage that's multithreaded, it's that make is able to invoke gcc multiple times, ie: make -j4.


It is because of --jobs that will spawn multiple make :)


That's true. I'd forgotten about --jobs.
_________________
My Wiki page
Back to top
View user's profile Send private message
Ant P.
Watchman
Watchman


Joined: 18 Apr 2009
Posts: 6920

PostPosted: Mon Oct 17, 2011 7:23 pm    Post subject: Reply with quote

You're better off waiting for a FX II model. The 1st-gen Phenoms sucked too. A Phenom II X6 T is the best value for money you can get from AMD for the time being.
Back to top
View user's profile Send private message
Chiitoo
Administrator
Administrator


Joined: 28 Feb 2010
Posts: 2564
Location: Here and Away Again

PostPosted: Mon Oct 17, 2011 9:26 pm    Post subject: Mmm. Reply with quote

Well at least it's holding a world record (for now). That must count for something, right? Right?!

Anyhoo, I've been using a Phenom(tm) II X6 1090T for a little over a year now, and it's been serving me quite well and emerge will gladly use it so much (all cores at around 100% at certain points) that 2GiB of RAM is not enough for it with -j7 or even some lower, resulting into a grand slow-down, to a point where the screen is updates every 10-30 seconds or so, until the build fails or I manage to kill it otherwise.

I had to actually reduce it to around -j3 or 4 to prevent this when building some larger packages.
Why only 2GiB you ask? Well, one of the sticks went bad (used to have 4 in total), so I had to deal with that for a while. Now I have 12GiB from which 8 is installed and all is well again. ^^

Only thing annoying me about it is that the price dropped a LOT just a while after I bought it. But then again, that tend sto happen me anyways, no matter when I buy stuff, so I'm used to it. I think I will do fine with this for many more years to come. Unless... I mean until I win lots from a lottery.
_________________
Kindest of regardses.
Back to top
View user's profile Send private message
Anon-E-moose
Watchman
Watchman


Joined: 23 May 2008
Posts: 6097
Location: Dallas area

PostPosted: Tue Oct 18, 2011 9:53 am    Post subject: Reply with quote

The x6's are reasonably priced, and decent performance to them.
_________________
PRIME x570-pro, 3700x, 6.1 zen kernel
gcc 13, profile 17.0 (custom bare multilib), openrc, wayland
Back to top
View user's profile Send private message
wrc1944
Advocate
Advocate


Joined: 15 Aug 2002
Posts: 3435
Location: Gainesville, Florida

PostPosted: Tue Oct 18, 2011 6:29 pm    Post subject: Reply with quote

Thanks to all for the feedback! :)

Here's a link I ran across for some new benchmarks of Bulldozer running with Gentoo, with more to come: http://www.phoronix.com/scan.php?page=news_item&px=MTAwMTg

For example, the FX-8150 on Gentoo, with gcc-4.6.1, kernel 3.0.6-gentoo: http://openbenchmarking.org/result/1110168-GR-BULLDOZER88


Guess I'll try running the Phoronix Test Suite on my current athlon II x4 640, and see how that compares to the 1100T x6 and Bulldozer 8150/8120.

Main AMDzone Bulldozer thread- lots of new feedback coming in now.
http://www.amdzone.com/phpbb3/viewforum.php?f=532&sid=4aa5c037b946489a74cbb488dd4e9fa1

Newegg has some more Bulldozer user's real-world reviews now trickling in (58 so far).
http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007671%20600213781&IsNodeId=1&name=Socket%20AM3%2b

Also, FWIW a good 3 page thread with early discussions of tech details of Bulldozer/Linux from March 2011. I thought it was worth a look for more background.
I really need to do my homework and wait for some more real-world user results before making any decision.
http://www.amdzone.com/phpbb3/viewtopic.php?f=532&t=138486
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11
Back to top
View user's profile Send private message
Dominique_71
Veteran
Veteran


Joined: 17 Aug 2005
Posts: 1875
Location: Switzerland (Romandie)

PostPosted: Wed Oct 19, 2011 6:38 pm    Post subject: Reply with quote

wrc1944 wrote:
In my case, in addition to the usual Office apps, web Browsers, and Multi-media apps, I'm particularly interested in knowing definitively if any of the Linux pro audio apps like Rosegarden, Ardour, Audacity, Hydrogen, Qsynth, etc., etc. are indeed multithreaded. Anyone know, or know how to find out if any app is single or multi-threads?


All the Linux pro audio apps are using jack as sound server. A requirement for a good audio app, and especially with a real time sound server like JACK, is multihreading. The audio thread must be separated from the control thread. You will also get a GUI thread, and so on. The consequence is than the audio pro apps are all multithreaded. Even one of the oldest audio app, the alsaplayer, is heavily multithreaded.
_________________
"Confirm You are a robot." - the singularity
Back to top
View user's profile Send private message
wrc1944
Advocate
Advocate


Joined: 15 Aug 2002
Posts: 3435
Location: Gainesville, Florida

PostPosted: Thu Oct 20, 2011 2:37 pm    Post subject: Reply with quote

Dominique_71,
Thanks very much for the info! :D

This is very encouraging regarding the Bulldozer FX cpus, as audio app performance is one of my main criteria.

This, with Bulldozers apparently being superb with multi-threaded apps, and the Gentoo/gcc compiling process apparently utilizing all cores, things are looking a little better for my specific usage patterns. Especially since I always compile/patch all my own kernels, even in other binary distros. Could anyone point me towards more info on the topic of audio apps being multi-threaded?

I'd think that even with the lack-luster single thread bulldozer performance, with web browsers, email, and normal office apps bulldozer would still offer "cpu overkill" performance. Not sure on other multimedia and/or xorg/video app performance (being single or multi threaded?).

Anyone with a Bulldozer running any Linux audio apps, such as multi-tasking with for example Rosegarden/Ardour/Jack along with multiple soft-synths and multiple VST instruments, soundfonts, and effects? Sure would be nice to get some user feedback on this audio usage pattern using Bulldozer.

UPDATE: GCC 4.5 vs. 4.6 On AMD's FX-4100 Bulldozer, Published on October 20, 2011.
http://www.phoronix.com/scan.php?page=article&item=amd_fx4100_gcc&num=1

From end of P.5, the basic conclusion, but the individual tests are still worth a look- some are significantly better with gcc-4.6.1, but it is a mixed bag. :roll: However, as is mentioned, this is with the stock compiler flags.
Quote:
More Bulldozer Linux benchmarks are on the way now that there's the AMD FX-8150 at hand for benchmarking, but as these benchmarks show, simply upgrading from GCC 4.5 to GCC 4.6 won't yield any real improvements for the new AMD processors when using the stock compiler flags. Some tuned compiler tests, a look at AVX on Bulldozer, and other articles are among what's coming up for the FX-8150 system.

_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11
Back to top
View user's profile Send private message
wrc1944
Advocate
Advocate


Joined: 15 Aug 2002
Posts: 3435
Location: Gainesville, Florida

PostPosted: Thu Oct 20, 2011 10:13 pm    Post subject: Reply with quote

I'm currently wondering how the Con Kolivas BFS scheduler might behave (positively or negatively) with the Bulldozers as opposed to stock kernel versions, as regards to its current issues.

Here's the latest version, where the concepts are explained in detail (commented out in the patch itself) in segments like Goals, Design summary, Design reasoning, Design details, Task insertion, Data protection, Virtual deadline, Task lookup, Scalability, etc. Seems like these items might have some effect on what people are discussing about Bulldozer issues, or is this unlikely to make any difference at all?

http://ck.kolivas.org/patches/bfs/3.0.0/3.0-sched-bfs-413.patch

I'm a long-time user of the Kolivas desktop responsiveness patches, but certainly no expert, but over the years they have always seemed to perform pretty well compared to stock cfq or deadline sinmce the earliest kernel 2.6 days.

Maybe someone with a lot more knowledge than I have would like to take a look at this, and offer comments and/or opinions?
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11
Back to top
View user's profile Send private message
Atle
Apprentice
Apprentice


Joined: 16 Sep 2004
Posts: 179

PostPosted: Sat Oct 22, 2011 7:39 am    Post subject: Reply with quote

I think too much emphasis is put on "single threaded apps". How often do you sit and wait for one of those? How often is the CPU the bottle neck here? Me: never.

Even web browsers (at least Google Chrome) are multithreaded these days. And common usage for me is running several programs at once. At the moment, Windows XP in VirtualBox, Chrome, Firefox (two different users), Android emulator etc. Cores and RAM are for me more important than really fast cores.

A quick google seems to indicate this CPU draws 125W, which is in the same range as Intel Core i7 which seems to be common at 95W or 130W. (You can get low power 65W i7 soon).

It really comes down to how you use your computer. If you're playing games on Windows, I'd go for fewer faster cores. If low power consumption is important I'd go for i5 or i7 65W. If you run a server, many virtual machines, many browsers etc., the new Bulldozer seems like a strong choice to me.
Back to top
View user's profile Send private message
John R. Graham
Administrator
Administrator


Joined: 08 Mar 2005
Posts: 10587
Location: Somewhere over Atlanta, Georgia

PostPosted: Sun Oct 23, 2011 11:39 am    Post subject: Reply with quote

Split off nontechnical conversation and moved it to Intel processors are manufactured into the Zionist country in Off the Wall.

- John
_________________
I can confirm that I have received between 0 and 499 National Security Letters.
Back to top
View user's profile Send private message
drescherjm
Advocate
Advocate


Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA

PostPosted: Sun Oct 23, 2011 3:16 pm    Post subject: Reply with quote

Quote:
I think too much emphasis is put on "single threaded apps". How often do you sit and wait for one of those?


Every single day of my life for the last 30+ years. Well almost..

Quote:
How often is the CPU the bottle neck here? Me: never.


Even my 12 threaded i7 processor with 18GB of memory is a bottleneck at what I do at work.
_________________
John

My gentoo overlay
Instructons for overlay
Back to top
View user's profile Send private message
wrc1944
Advocate
Advocate


Joined: 15 Aug 2002
Posts: 3435
Location: Gainesville, Florida

PostPosted: Sun Oct 23, 2011 5:08 pm    Post subject: Reply with quote

I finally decided to order a x6 1090T from newegg.

With a revised AMD Bulldozer version and/or Ivy Bridge chips probably released within 6 months or so, I just couldn't justify getting any of the current Bulldozers plus a new motherboard that would likely be replaced so soon. All things considered, a simple x6 1090T drop-in replacement into my current 8GB ram system seemed the best option, and wait and see what Q2/Q3 2012 brings.

I'll admit the 8-core Bulldozers were very tempting (especially for Gentoo emerging), but reading many hundreds of user and website reviews on the x6 1090T vs. Bulldozers convinced me I easily could live with the x6, and save most of my money for a future "Bulldozer II/Piledriver" or Ivy Bridge setup on a future upgraded motherboard/cpu combo that had the bugs worked out. Plus, later on, I'll still have a great x6 system for my backup box.


FWIW: Here's a thread just started on AMDzone "Recommended compiler optimizations for Zambezi and Interlago" Has 2 links (from amd devs) for gcc-4.6.x and Open64 compilers, that mention things like -ftree-vectorize, avx flags, and march=bdver1 (interlagos), among others.
http://www.amdzone.com/phpbb3/viewtopic.php?f=532&t=138898
_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11
Back to top
View user's profile Send private message
wrc1944
Advocate
Advocate


Joined: 15 Aug 2002
Posts: 3435
Location: Gainesville, Florida

PostPosted: Tue Oct 25, 2011 8:45 pm    Post subject: Reply with quote

New 13 pages of in depth Bulldozer FX-8150 benchmarking on Linux. (Unfortunately, on Ubuntu and not Gentoo, but still good stuff).

http://www.phoronix.com/scan.php?page=article&item=amd_fx8150_bulldozer&num=1

Quote from the last page, which is encouraging:
Quote:
Fortunately, for Linux users, most open-source software is well multi-threaded. If you are running Gentoo, Arch, or just doing a lot of compiling in general, the AMD Bulldozer CPUs should be able to prove their value very well. Beyond that, with open-source software that you may be building, GCC and Open64 already have Bulldozer (version 1) optimizations in place.

Linux users will be able to take full advantage of the Bulldozer architecture sooner than Microsoft Windows customers, which will primarily see the real potential when Windows 8 is released. In the Linux world, there's still some Bulldozer kernel work that's not yet merged and presumably more compiler/kernel optimizations coming, but we will hopefully see all of that merged and ready in time for next spring when Ubuntu 12.04 LTS, Fedora 17, and other Linux distributions are pushing out their new versions. If you are so inclined, you can always pull the patches yourself, tune your compiler options, and make other tweaks today to take greater advantage of these new AMD processors. The upcoming FX-8150 Linux articles have more revealing information.

_________________
Main box- AsRock x370 Gaming K4
Ryzen 7 3700x, 3.6GHz, 16GB GSkill Flare DDR4 3200mhz
Samsung SATA 1000GB, Radeon HD R7 350 2GB DDR5
OpenRC Gentoo ~amd64 plasma, glibc-2.36-r7, gcc-13.2.1_p20230304
kernel-6.8.4 USE=experimental python3_11
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Kernel & Hardware All times are GMT
Goto page 1, 2  Next
Page 1 of 2

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum