1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

which is the better chip

Discussion in 'General Hardware' started by trt740, Jul 29, 2006.

  1. trt740

    trt740

    Joined:
    May 12, 2006
    Messages:
    10,935 (3.53/day)
    Thanks Received:
    1,113
  2. wazzledoozle

    wazzledoozle New Member

    Joined:
    Aug 30, 2004
    Messages:
    5,414 (1.46/day)
    Thanks Received:
    161
    Location:
    Seattle
  3. trt740

    trt740

    Joined:
    May 12, 2006
    Messages:
    10,935 (3.53/day)
    Thanks Received:
    1,113
    Im wondering

    I have a 939 3000+ now overclocked to 2.43ghz or a 3800+ and Im wondering if now is the time to buy a upgrade I waited on the old xp3200 to become cheaper and it never did it acutally got more expensive. So I don't want to miss out if they stop making the 939 chips. So what do you think i should do wait or buy now!!! :D
     
  4. trt740

    trt740

    Joined:
    May 12, 2006
    Messages:
    10,935 (3.53/day)
    Thanks Received:
    1,113
    okay crack head I neither wish to buy intel or build a new system LOL. I would rather die than buy from intel. That company lost me years ago when they constantly made me buy a new motherboard for every new chip while at the same time Amd allowed up to 9 different chips in their motherboards.
     
  5. trt740

    trt740

    Joined:
    May 12, 2006
    Messages:
    10,935 (3.53/day)
    Thanks Received:
    1,113
    oh and if there wasn't an Amd we would still be using Pent Pros, and the price would be crazy crazy.
     
  6. trt740

    trt740

    Joined:
    May 12, 2006
    Messages:
    10,935 (3.53/day)
    Thanks Received:
    1,113
    Intel is a greedy bastard of a company,
     
  7. wazzledoozle

    wazzledoozle New Member

    Joined:
    Aug 30, 2004
    Messages:
    5,414 (1.46/day)
    Thanks Received:
    161
    Location:
    Seattle
    Well if you refuse to get the best deal on the best consumer level processor ever, then ok. The 3800 X2 is only $150 here-
    http://www.mwave.com/mwave/viewspec.hmx?scriteria=BA21810

    The extra 512k of L2 on the Opteron's/4600 are not going to make any difference. Spend the cash you will save on something else.
     
  8. trt740

    trt740

    Joined:
    May 12, 2006
    Messages:
    10,935 (3.53/day)
    Thanks Received:
    1,113
    also all Amd has to do is up the on chip cache no wonder Conroe is faster it has L2 of 4 mbs And AMD has 1MB and some have 2 MBS. Also AMD can still open its bus more. I once read the 64 can reach 3 ghz on its frontside bus
     
  9. trt740

    trt740

    Joined:
    May 12, 2006
    Messages:
    10,935 (3.53/day)
    Thanks Received:
    1,113
    I refuse they are trying to kill AMD and if they do watch the prices jump to stupid levels.Then we will only have one choice intel and i remember when they were the only choice and it was horrible
     
  10. wazzledoozle

    wazzledoozle New Member

    Joined:
    Aug 30, 2004
    Messages:
    5,414 (1.46/day)
    Thanks Received:
    161
    Location:
    Seattle
    L2 cache has no effect on the performance of high IPC and short pipeline cpu's. Just a marketing gimmick.
     
  11. trt740

    trt740

    Joined:
    May 12, 2006
    Messages:
    10,935 (3.53/day)
    Thanks Received:
    1,113
    whatever you say bro just remeber what it was like with no AMD. Ask Cyrix and Alfa how they like intel in the 90's
     
  12. _33

    _33 New Member

    Joined:
    Mar 25, 2006
    Messages:
    1,248 (0.40/day)
    Thanks Received:
    31
    Location:
    Quebec
    Opteron 165 is the better deal.
     
  13. Alec§taar New Member

    Joined:
    May 15, 2006
    Messages:
    4,677 (1.51/day)
    Thanks Received:
    94
    Location:
    Someone who's going to find NewTekie1 and teach hi
    You guys are under the impression that L2 cache is of "no benefit" here I've seen...

    See my subject-line/title above first, & then some points I'm going to make next below, do drink it in & digest it (and I invite discussion on it as well)!

    (And I have made these same points in other threads here recently & over time)

    It's much like how when I came here it was being stated here on these forums largely that DualCore/SMP was of no value either!

    (That is easily proven to be a falsehood, especially on the modern OS' of today, & with multithreaded apps (which is MOST of what you run nowadays & for years now, and taskmgr.exe can show you this in its PROCESS tab, w/ the THREAD column visible))

    AND, yes gains can be seen in games that leverage multiple thread design & Quake 4 SMP shows up to 87% boosts over its singlethread early designs as well indicating this, & it's the "game design of the future" (of which you can use taskmgr.exe to see you run MOSTLY apps (not games, not yet @ least, a pity) of that nature):

    Gaming, you only see like 2% from what I have heard HERE, but @ other sites online they say 5% (on average from both, supposedly) but, if ALL you do on a PC is game?

    Get a gaming console.

    Still, nowadays, if you note? They too, use multi-cpu core construction & multithreaded OS & code designs as well! I wonder why that is... maybe because it produces overall gains there as well?

    I imagine like anything else? This varies game, by game & how they are designed as well. Games DO pull a lot of screen refresh, data level loads, but they still have a main loop for animation & this imo, is WHERE you see the gain (instruction-wise, more than data-wise as far as caching is concerned). Quake 4 SMP however, shows that tremendous gain IS possible, again up to 87% has been illustrated as fact.

    Office type apps gain by it typically, so if you use your computer in THIS capacity (home, work, or school) it can help.

    Video or photo editing, afaik, is MOSTLY system-memory (your RAM chips on your mobo) bandwidth constrained though, but from what I understand, it GAINS via L2 cache being larger too (L2 cache IS memory after all).

    E.G.-> Applications like SETI@Home & probably Folding@Home, L2 cache levels can change things radically from what I read on the SETI forums etc. (which is a large part of WHY I went w/ the CPU I have (larger amounts of L2 cache onboard CPU)).

    There are other apps (server-side types) that gain by larger L2 cache onboard the CPU as well, but if you are only into gaming on your rig? Then, you may wish to go with lower L2 cache bearing CPU types for the games of today, but you're shorting yourself for the games of tomorrow (of which Quake 4 SMP is an illustration of gain using multilple threads on an SMP rig)... or again: Just get a gaming console.

    MOST of today's games (Quake 4 SMP being a HUGE exception, again 87% gains are noted in it vs. its older singlethread designed early design) may not show a benefit, but the games of tomorrow, will!

    (A lot of talk here centers around being "future-proofed" & imo? You DO get that on SMP/DualCore rigs due to multithread program design being more & more prevalent (taskmgr.exe shows us all this for Win32 GUI apps (usermode threads iirc), & kernels today have re-entrant code & do multithread design as well (kernel mode threads, needed for SMP afaik))).

    L2 cache is really good when you have a set of program code &/or data that repeats itself constantly, so if you have apps you use a great deal of THAT type of nature - you gain.

    Still, most ANYTHING gains by it, to some degree (again, it varies depending on the application) because above all, you have to remember it is FAR FASTER RAM, than your system motherboard RAM chips are AND, "nearer" to the CPU core itself!

    (Due to "less travel time/latency" typically & the CPU has a faster internal 'bus' speed as well, so does its cache iirc vs. that of the system RAM on the mobo, HyperTransport memory controller designs or not + DMA (they do help though), notwithstanding))...

    & thus, the MORE L2 cache you have?

    The more data &/or instruction code you can keep cached in it, for ANY application that repetitively executes the same instructions &/or data over and over repeatedly (which is what caching is largely about & it is a PROVEN concept - albeit, somewhat imperfect in certain conditions, better than not having it done @ all overall. In Windows VISTA, Microsoft is working to improve L2 cache efficacy & imo, based on what is in my P.S. below).

    APK

    P.S.=> There is an effect known as "cache pollution" that CAN be 'offset' to a large degree & help improve L2 cache efficacy!

    (I've written about it here before in the Folding@Home thread in the GENERAL SOFTWARE section)

    The way to do that? Use Taskmgr.exe "processor affinity" settings BY PROCESS, to stop new code from entering into a particular CPU core's L2 cache, upsetting a process' cache purity (for lack of a better term here) & its already cached data &/or code resident there...

    You can get more info., validating my statement above, here:

    http://www.windowsitpro.com/Article/ArticleID/3873/3873.html

    & here (more technical in nature):

    http://www-jcsu.jesus.cam.ac.uk/~cam84/osIIsI.txt

    "Processor Affinity - Each process saves which processor it was run on last and if it is possible the same processor will be used for it's next scheduled CPU time to help cache locality."

    That's done typically in STRICTLY explicitly coded SMP programs, via API calls like Win32's "SetThreadMaskAffinity" after detection of 2-N actual CPU's present... on multithreaded code, the OS process scheduler manages this & is being improved upon by MS in VISTA especially, & imo, in regards MOSTLY to cache pollution stoppage...

    EDIT PART - now, I will question myself!

    And, it has to do w/ multithreaded code on SMP rigs...

    Will this cache pollution measure I use (weak one imo, a rig job vs. hard affinity settings, this is soft affinity afaik)... does it hold true on multithreaded code? I know it will on single threaded processes, but not sure on SMP!

    Not sure, would like feedback! apk
     
    Last edited: Oct 1, 2006
  14. Tatty_One

    Tatty_One Senior Moderator Staff Member

    Joined:
    Jan 18, 2006
    Messages:
    16,769 (5.23/day)
    Thanks Received:
    2,557
    Location:
    Worcestershire, UK
    Agreed! Tests show that the L2 Cache difference is minimal across the whole range of benchmarks, about 2% but no advantage in games for example. I know you dont mention it but if You wont go for the conroe I can recommend mine and if you do that voltmod you acheive that on air BUT you will be able to overclock to 2.6 on the volts you have I would think.
     
  15. Alec§taar New Member

    Joined:
    May 15, 2006
    Messages:
    4,677 (1.51/day)
    Thanks Received:
    94
    Location:
    Someone who's going to find NewTekie1 and teach hi
    That's how games are designed, today... but, Quake 4 SMP?

    It's shown up to 87% gain over its original single-thread design... it IS the game design of the future, & folks here and elsewhere talk a lot about "future proofing" themselves?

    :)

    * DualCore will & does do that for you! Quake 4 SMP illustrates it PERFECTLY (do read again above, I edited in a LOT of information in this reply as I went to make the point strike home & reinforce it with backnig info. etc. et al).

    See here:

    http://www.2cpu.com/story.php?id=3960

    "id Software and Raven have worked with Intel to dramatically improve QUAKE 4 performance on systems with HT Technology, multiple processors, or dual core processors. Players who have a system with one of these configurations will notice performance gains in QUAKE 4 of anywhere from 25% - 87% depending on processor type".

    Feel free to correct me where I am "wrong" on this account, because I did use the 'iirc' & could be off... but, here is what I am citing when I cite those numbers:

    http://www.phoronix.com/scan.php?pag...item=455&num=1

    "According to id Software, systems with a single physical core but those capable of benefiting from Hyper-Threading Technology should experience a 25% performance boost. Systems with two or more physical cores should see up to an 87% improvement. The true benefits may vary greatly depending upon the CPU and GPU limitations as well as other factors."

    Also, here is a BETA test done on it, before the FINAL model released, just so you have some "solid" numbers (somewhat, because beta code is slower due to debug routines in it still) to look @ as well:

    http://www.amdzone.com/modules.php?...s&file=index&req=viewarticle&artid=225&page=2

    :)

    Generally, I won't cite numbers unless I have SOMEKIND of backing from others online or from tests etc., but those number stuck out & it appears I cited them correctly from "off the top of my head" which usually, I am NOT "wont to do"... but, there ya are!

    On Win32 OS it's not up to 87% usually, but more like 44% on average (and on "HyperThreaded" CPU's, which is largely not nearly as effective as true SMP/DualCore setups are) from what I understand in practice. On Linux, they're not seeing as much though in tests.

    APK
     
    Last edited: Jul 29, 2006
  16. Tatty_One

    Tatty_One Senior Moderator Staff Member

    Joined:
    Jan 18, 2006
    Messages:
    16,769 (5.23/day)
    Thanks Received:
    2,557
    Location:
    Worcestershire, UK
    Alec we will have to stop meeting like this!........

    "Get a gaming console.".....why? will it send and receive e mail for me?....burn DVD's or CD's for me I wont go on....u take my point. I think you and I may be a slightly older generation than many on here, in the UK, 90% of all male PC users (in the home) under the age of 25 use a PC almost exclusively for gaming and for gaming currently L2 Cache or dual core processors make litttle difference especially when taken into account with the average higher costs.

    In fact in many cases a dual core processor running at the same speed as a single core will be slower in a game. I know you programme and do various other stuff, I do a little video editing and number crunching but game some too and I like the fact that I can edit some Video and watch a dvd in a window at the same time for example (hence I have a x2)....but the majority will not be interested in that I think.

    Edit: "it IS the game design of the future, & folks here and elsewhere talk a lot about "future proofing" themselves?"

    Yes of course but who buys the tyres before they have bought the car? To use one or two games as a reason a gamer should buy a dual core TODAY TBH is not an arguement, he should buy it tomorrow when there are 50 games and the CPU is 50% of the price, just look how much dual cores have come down in price in the last 6 weeks alone! I bet you could almost get 2 of yours today for the price you paid and how many games will really fully utilise it?
     
  17. Alec§taar New Member

    Joined:
    May 15, 2006
    Messages:
    4,677 (1.51/day)
    Thanks Received:
    94
    Location:
    Someone who's going to find NewTekie1 and teach hi
    It's been my point all along in fact... you don't get a PC just to play games, not by a longshot!

    And, that means what, vs. what I stated above?

    It doesn't change the fact that SMP & multithreading is & has been the design of softwares and OS' for more than a decade now & it shows definite overall gains (even on consoles, which for example, the xbox360 uses an NT-based OS core, AND consoles today use multiple CPU/SMP designs also - why is that?)

    Did I say that wouldn't happen? No. With single core apps it is how it is... but, note the Quake 4 SMP gains over its early single-thread designs (25%-87% depending on CPU type used (HT vs. true SMP/DualCore)... again - there's talk here of "future proofing" one's self... & if you think SMP ready/multithreaded games are NOT the future?

    Well, see the Quake 4 SMP stuff above.

    I think they will be... especially if buying CPU's today, for the games of tomorrow (of which Quake 4 SMP truly IS the prototype view/look @ the future)... and for tasks such as you mention, it helps as well! I've noted it for YEARS now, nearly a decade running SMP (true dual physically present CPU's setups) & always could multitask smooth as glass because of it.

    :)

    You do it, so you don't have to buy again, tomorrow basically & get MORE out of your existing codebase you already own as well (future proofing yourself to a good extent, because who says you have the money for a "State-of-the-Art" PC in the future?) and that code (see for yourself in taskmgr.exe)?

    It's MOSTLY multithreaded on what you run in usermode apps, & @ the OS kernel level as well (otherwise, w/out re-entrancy for multithreading? We wouldn't have SMP ready OS period!)

    APK
     
  18. Tatty_One

    Tatty_One Senior Moderator Staff Member

    Joined:
    Jan 18, 2006
    Messages:
    16,769 (5.23/day)
    Thanks Received:
    2,557
    Location:
    Worcestershire, UK
    Completely right, so buy it tomorrow then if your a gamer which was my point and the fact that the majority use their PC mainly for gaming. I have no doubts as to the strenghts of dual core.....I have one after all but I do have doubts as to if it is worth buying TODAY for someone who almost exclusively uses their PC for games and I think you will find that there are a lot of them out there.

    I didnt buy my dual core on that particular day because I wanted to video edit in 3 months time, I bought it cause I wanted to video edit right then!
     
  19. Tatty_One

    Tatty_One Senior Moderator Staff Member

    Joined:
    Jan 18, 2006
    Messages:
    16,769 (5.23/day)
    Thanks Received:
    2,557
    Location:
    Worcestershire, UK
    Ohhh yeah and I would go for the Opty. :D
     
  20. Alec§taar New Member

    Joined:
    May 15, 2006
    Messages:
    4,677 (1.51/day)
    Thanks Received:
    94
    Location:
    Someone who's going to find NewTekie1 and teach hi
    They're wasting a PC then, & should have gaming consoles... but note & affirm this for me (since you guys like to seem to evade my points about this & circle back to gaming all the time):

    The XBox360 or other modern gaming consoles: Do they, or do they not, use SMP designs (multiple CPU cores present) first of all?

    Secondly, in the case of the XBox360 specifically: Does it use an NT-based OS core, & is that core SMP ready??

    Third & lastly on this account/note: Why then, is SMP physical CPU design used on consoles as well as PC's, and with OS + code ready for it???

    APK

    P.S.=> This is "diverging" from L2 cache bennies though, still important to point out where SMP will benefit even gamers to some extent, else why design consoles of SMP nature otherwise?

    See more of what I wrote above about L2 cache in my first post's P.S. & cache coherency/cache pollution, once you answer the points here on SMP & gaming consoles... thanks! apk
     
    Last edited: Jul 29, 2006
  21. Tatty_One

    Tatty_One Senior Moderator Staff Member

    Joined:
    Jan 18, 2006
    Messages:
    16,769 (5.23/day)
    Thanks Received:
    2,557
    Location:
    Worcestershire, UK
    Yes of course they are wasting a PC I agree, but that was not my point, they choose to hence my comments in my first post, will a console send e mails(some probably) burn DVD's/CD's etc probably not so thats why in part they have a pc. All your comments are valid without question but I just think you were looking at it more from your own perspective than that of many others. Lol we are hijacking this thread so I am gonna shut up now and let someone else answer your last question.
     
  22. Alec§taar New Member

    Joined:
    May 15, 2006
    Messages:
    4,677 (1.51/day)
    Thanks Received:
    94
    Location:
    Someone who's going to find NewTekie1 and teach hi
    Absolutely.

    They might, because the "trend" in gaming consoles today IS online access... as well as SMP designs @ the physical CPU level + OS cores & code used on them.

    Well, then imo @ least, and it seems yours as well? Not only are they wasting their PC gaming only on it, but their own personal gain by learning things on it other than gaming...

    Plus, wasting their cash spent on a PC really, as most of its code it runs IS multithreaded & has been for decades first of all, & even in gaming? They are shorting themselves for the future of games (finally "catching up" & being multithreaded, & faster - Quake 4 SMP illustrates it, see the URL below & my statements regarding it being ONLY beta debug riddled code laden).

    Think about it.

    No, we're actually MAKING THIS THREAD better!

    Simply by pursuing why SMP/DualCore + multithreaded OS & application design rocks on SMP rigs (where it actually slows up on single cpu rigs) vs. single threaded ones on said SMP/DualCore rigs (moreso than on HyperThreaded ones)...

    As well as going to the initial premise I laid out about L2 cache coherency vs. cache pollution in my first posts' P.S., pasted here for reference on editing, again:

    =================================================================

    P.S.=> There is an effect known as "cache pollution" that CAN be 'offset' to a large degree & help improve L2 cache efficacy!

    (I've written about it here before in the Folding@Home thread in the GENERAL SOFTWARE section)

    The way to do that? Use Taskmgr.exe "processor affinity" settings BY PROCESS, to stop new code from entering into a particular CPU core's L2 cache, upsetting a process' cache purity (for lack of a better term here) & its already cached data &/or code resident there...

    You can get more info., validating my statement above, here:

    http://www.windowsitpro.com/Article/...3873/3873.html

    & here (more technical in nature):

    http://www-jcsu.jesus.cam.ac.uk/~cam84/osIIsI.txt

    "Processor Affinity - Each process saves which processor it was run on last and if it is possible the same processor will be used for it's next scheduled CPU time to help cache locality."

    That's done typically in STRICTLY explicitly coded SMP programs, via API calls like Win32's "SetProcessAffinity" after detection of 2-N actual CPU's present... on multithreaded code, the OS process scheduler manages this & is being improved upon by MS in VISTA especially, & imo, in regards MOSTLY to cache pollution stoppage... apk

    EDIT PART - now, I will question myself!

    And, it has to do w/ multithreaded code on SMP rigs...

    Will this cache pollution measure I use (weak one imo, a rig job vs. hard affinity settings, this is soft affinity afaik)... does it hold true on multithreaded code? I know it will on single threaded processes, but not sure on SMP!

    Not sure, would like feedback!

    =================================================================

    The last point on consoles being SMP & bearning multithread ready OS (and quite possibly code as well in the games on them, like Quake 4 SMP illustrates on PC's as showing 25% (H/T cpu's imo) to 87% (true DualCore/SMP rigs) gains possible).

    Benchmarks results are the BEST way to show this, see here (pictures say a 1,000 words - better than, lol, reading thru my 1,000++ word posts, eh?):

    Also, here is a BETA test done on it, before the FINAL model released, just so you have some "solid" numbers (somewhat, because beta code is slower due to debug routines in it still) to look @ as well:

    http://www.amdzone.com/modules.php?...s&file=index&req=viewarticle&artid=225&page=2

    :)

    The BEST part of that is? It's BETA code, & beta code (this I know, because I do coding for a living & know what beta code bears for performance-hits) bears a lot of debug symbol bulk & extra code for errtraps etc. as well, making it slower!

    A final copy removes that in its builds, & makes it faster 9/10 times, only!

    APK

    P.S.=> I've been using Dual CPU/SMP machines since, oh, the Pentium I days (started out with Pentium I 233mmx's) & have had them ever since thru every OS & processor generation... & I've been writing multithreaded code since 1996 or so. It's ALL in how efficiently you design using threads, some things however, are NOT meant for it & are too "linear" in nature to benefit from them, e.g.:

    A=A+B
    B=A-1
    C=A

    B can't complete itself, until it gets A's result (&, vice-a-versa, + all the way to C result too, may not be BEST example, but point is here)...

    This is not a case where I would put A's & B's code on diff. threads to get C... because A has to wait on B, etc. - et al...

    Commonly shared data is another - something called a "race condition" can result if you put 2 threads onto the same datablock to process, so you have to watch it.

    It's "better" to do 2 - N DIFFERENT tasks with diff. data - as in the Quake 4 SMP example, I would wager they put Sound processing on 1 thread, the main body thread is the game animation loop on its own thread, & networking code on yet another thread for example... safe & there is little arguing with the results (up to 87% increases!!!).

    Above all - Multiple thread code is better on SMP/Dual or more CPU systems, & actually has MORE OVERHEADS on it on single cpu rigs, & runs SLOWER on them because of it!

    However, again, do note - most of what you run in APPS, and certainly your OS (Linux, Unix, Windows, all ready for it now)? IS MULTITHREADED & has been for years... taskmgr.exe above can & should have showed you this, & if done right? The Quake 4 SMP gains (up to 87%, which IS huge imo) certainly does... apk
     
    Last edited: Jul 29, 2006
  23. trt740

    trt740

    Joined:
    May 12, 2006
    Messages:
    10,935 (3.53/day)
    Thanks Received:
    1,113
  24. Moose1309 New Member

    Joined:
    May 7, 2006
    Messages:
    67 (0.02/day)
    Thanks Received:
    1
    Not a bad choice at all!

    I would have saved $75 and gone with the 3800+. But, the 4400 would have been my second choice. Let us know how it OC's. BTW, that's a much better deal than the Opty 165 IMO. 11x multiplier gives you much greater flexibility when OC'ing. Because of my 9x on my Opty I'm stuck at 2.466 GHz

    Regarding the whole L2 cache thing, yeh I didn't read the flame war there, but it does have a ~2-10% impact overall, IIRC. Extra cache on the Intel Conroes has a little more of an impact, because of the memory controller design etc.
     
  25. bigboi86

    bigboi86 New Member

    Joined:
    Apr 8, 2006
    Messages:
    1,450 (0.46/day)
    Thanks Received:
    35
    Location:
    techPowerUp!
    I personally want one of those opteron 165's :)
     

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page