Jump to content


Photo
- - - - -

tight timings? loose timings? dividers? 1T or 2T? 754 or 939? (read he


  • Please log in to reply
41 replies to this topic

#13 Angry_Games

Angry_Games

    Honorary Member

  • Members
  • PipPipPipPipPipPipPipPipPip
  • 27999 posts

Posted 18 August 2006 - 01:49 PM

ok...

against my better judgement...

I have decided to once again open this thread for discussion

I'm already tired of everyone that is going to come in here and nitpick every little last half-point of a benchmark gained and then arguing that this thread is stupid and useless as they know more than the combined forum does...so if you are one of those, buzz off as you will just get deleted from this thread.

This thread is for discussion of real-world testing and the benefits of overclocking the cpu and how dividers and timings make very little difference since with AMD64 cpu's, the cpu Mhz is the KING.

Yes, all you know-it-alls, we all ALREADY know that you are superior with your 1:1 300x10 @ 2-2-2-2-2-2-2-2-2 timings so you can also buzz off as this thread is not for you.

This thread is for those that want to know that using a divider or looser timings to gain more cpu Mhz.

This thread is for anyone that doesn't quite understand or is still living in P4/AMD AthlonXP days when memory controllers were not built into the cpu's die, and made an extraordinary amount of difference in performance.

All of the graphs and charts and tests that are in this thread prove that while 1:1 @ tightest possible timings are always best, it is still only a small percentage increase over loose timings + a divider @ same cpu Mhz.

This thread is for you, who wants to overclock the cpu to it's maximum and is worried that dropping to a DDR180 or DDR166 divider, or moving from 2-2-2- timings to 3-4-4 timings is going to be an ugly, performance-harming affair....

IT IS NOT

It is better to rack up the cpu Mhz as far as you can than it is to have less cpu mhz and tighter timings and no dividers in 99% of the tests you will run (especially games that are cpu dependent etc).


So

keep the discussion going...this thread has schooled MANY of you guys who have shown up over the last 2 years, and I see that schooling paying off as you have all begun to teach everyone else around you the benefits of AMD64 cpu Mhz over memory bandwidth etc (and yes, even Super Pi as Super Pi has ZERO real-world bearing other than showing that your e-Penis is larger than some guy who can't beat your time).

Enjoy, and keep in mind the mods will patrol this thread pretty heavily...
Diablo III arrives at Angry's house:
Posted Image

#14 Angry_Games

Angry_Games

    Honorary Member

  • Members
  • PipPipPipPipPipPipPipPipPip
  • 27999 posts

Posted 20 August 2006 - 07:46 PM

by the way

I am running some pretty neat tests right now that might be of interest to everyone

I've got two 3800+ dual-core X2 rigs with almost the EXACT same hardware running:

3800+ X2 (939)
DFI Lanparty NF4 Expert
2x1GB OCZ PC4000 3-4-4-8
ATI Radeon X1900XT @ 625/650 (Omega's drivers based on CAT 6.7)
Audigy2 ZS
2x80GB SATA II RAID-0
NEC 3550A DVD-RW
OCZ Powerstream 520w

vs

3800+ X2 (AM2)
DFI NF4 Ultra II-M2 (AM2)
2x1GB OCZ PC2-8000 DDR2-1000Mhz (@ DDR2-800Mhz, 5-6-6-15)
ATI Radeon X1900XT @ 625/650 (Omega's drivers based on CAT 6.7)
Audigy2 ZS
250GB Seagate 7200.10 16MB cache SATA II
NEC 3550A DVD-RW
OCZ Powerstream 600w


right now the results are interesting, but not shocking or anything...the AM2 is scoring slightly higher which tells me that it is purely because of the memory bandwidth...but in the gpu-intensive tests like 3dmark03-06, you don't see much difference at all (single percentage so far it looks like)

I will also probably run a benchie of FEAR and HL2 Lost Coast just to see what they show.

Right now everything is stock speed, but I am going to also re-run them at 2500Mhz with some gpu overclocks (all Mhz speeds will be the same on both platforms however so we can get a true Mhz to Mhz comparison across the two different platforms)

I dig nerdy stuff like this, and you guys dig reading nerdy stuff like this because it is always much better to have some dorko-nerd actually do all the testing instead of just listening to know-it-all blowhards trying to impress their Counter-Strike buddies that they know more than we do in this forum :P
Diablo III arrives at Angry's house:
Posted Image

#15 Kaxx

Kaxx

    Member

  • Members
  • PipPip
  • 316 posts

Posted 24 August 2006 - 06:43 PM

Hey happy, just checked this thread out fairly quickly, but from what I understand there is "basically" no benefit from buying OC'able RAM??

Edit, just reread some of this. I should have done that before posting. You can delete this any time........

#16 Jimmy Question

Jimmy Question

    Member

  • Members
  • PipPip
  • 129 posts

Posted 28 August 2006 - 12:20 PM

Honestly I couldn't actually read the entire rant. Too much emotion and too many technical details deliverd with too much intensity, with too many references to porno thrown in like too much pepper.

Happy_Games, what did you think of Comedy Central's Roast of William Shatner ? I like it for the most part, except for that gross and disgusting little creepy guy that ran around licking people. Gross !! Ruined what would otherwise have been a perfect event. I feel kinda the same way reading this. To much stark science coupled with too much lurid senstationalism. Your wife must have an interesting life.

Okay, back to topic. Yes, I think that's the point. That the differences in this performance vs. that performance is really negligible to most people that want to OC for a reason. That in and of itself is enough for Happy_Games, but I think the kicker for him is when you also consider the COST of the the performance "gain". He's basically drawing a line between extreme performance geeks, and true whack-jobs that miss the forest for the trees.

That's my impression so far. Maybe, if my stomach settles, I'll come back to read some more. But for right now, I've had enough. See, there is also a sense of "proportion" when it comes to reading other people's posts, and this one (as useful as it might be) has provided me with all I need right now.

#17 Ropey

Ropey

    New Member

  • Members
  • 1 posts

Posted 31 August 2006 - 11:11 PM

Thanks for the help on the TCCD DRAM setup. Before reading and testing your method I was at a 180 divider. Now I am running 300MHz FSB 1:1 @ 2.5-3-3-8 TCCD.

Much appreciation,

Ropey

#18 acphydro

acphydro

    Member

  • Members
  • PipPip
  • 76 posts

Posted 02 November 2006 - 06:52 AM

Just about everything you've said there Happy, I can completely agree with. Especially the RAM timings and such. My brother is one of those people. He doesn't want to upgrade to 2GB of memory because the 1GB sticks wont reach the timings of the sticks I gave him at 1:1 [2.5-3-3-6]. I've got his 3000+ Venice running stable on a POS Soltek VIA based motherboard @ 2.25ghz and I'd have to say thats damn nice for being on that POS board. [It was my old board before I went DFI and I'm not looking back in the slightest] He refuses to use a divider and lower timings to take it any higher, I had it running at 2.6ghz completely stable but he didn't like it because the timings of the ram were dropped to 2.5-4-4-7 for it to be stable and said that half life 2 and counterstrike needed tighter timings and that memory timings make the pc perform better. Tried explaining to him that loosening the memory timings and being able to run 350mhz faster will pretty much negate the minor performance loss from doing so and he still resists...lol.

If his memory isnt running at the rated timings above 220mhz he isnt satisfied no matter what. He even tried telling me I should run my memory at 200mhz so that I can use better timings even if I have to lose 250mhz+ cpu speed---I even showed him my own testing and he refuses.

Either way if thats what he wants to do more power to him, but my system impresses the hell out of me compared to what I used before and my real world applications and games run better than they ever did. When I ran 1:1 at 250x10 it was damn fast, but when I set a divider in place to run 305x9 its even faster and thats all that matters to me. I'd rather lose a few performance points in memory bandwidth and have a faster overall system anyday.

Anyways nice work and thanks for providing all the information and details needed to us DFI customers. Just having this forum here is enough to make me want to come back to DFI because I know I can always get the proper information and support needed should I have any problems.

[end brown nosing]

Though right now im enjoying final fantasy 12 on my ps2 too much to care lol

#19 Angry_Games

Angry_Games

    Honorary Member

  • Members
  • PipPipPipPipPipPipPipPipPip
  • 27999 posts

Posted 02 November 2006 - 12:44 PM

lol well....

if after reading through all of this and your brother (or anyone for that matter) still refuses to see the light (ie: cpu mhz is the king, not mem timings when it comes to A64's), then like you say, more power to him.

"you can't argue with a drunk" is sorta how I see it ;)

I been playing waaay too much Guitar Hero on PS2 and NHL 2k7 on Xbox360 to play any pc games lately myself...
Diablo III arrives at Angry's house:
Posted Image

#20 cuoficr

cuoficr

    Certified Geek

  • Members
  • PipPipPipPip
  • 1174 posts

Posted 02 November 2006 - 01:53 PM

by the way

I am running some pretty neat tests right now that might be of interest to everyone

I've got two 3800+ dual-core X2 rigs with almost the EXACT same hardware running:

3800+ X2 (939)
DFI Lanparty NF4 Expert
2x1GB OCZ PC4000 3-4-4-8
ATI Radeon X1900XT @ 625/650 (Omega's drivers based on CAT 6.7)
Audigy2 ZS
2x80GB SATA II RAID-0
NEC 3550A DVD-RW
OCZ Powerstream 520w

vs

3800+ X2 (AM2)
DFI NF4 Ultra II-M2 (AM2)
2x1GB OCZ PC2-8000 DDR2-1000Mhz (@ DDR2-800Mhz, 5-6-6-15)
ATI Radeon X1900XT @ 625/650 (Omega's drivers based on CAT 6.7)
Audigy2 ZS
250GB Seagate 7200.10 16MB cache SATA II
NEC 3550A DVD-RW
OCZ Powerstream 600w


right now the results are interesting, but not shocking or anything...the AM2 is scoring slightly higher which tells me that it is purely because of the memory bandwidth...but in the gpu-intensive tests like 3dmark03-06, you don't see much difference at all (single percentage so far it looks like)

I will also probably run a benchie of FEAR and HL2 Lost Coast just to see what they show.

Right now everything is stock speed, but I am going to also re-run them at 2500Mhz with some gpu overclocks (all Mhz speeds will be the same on both platforms however so we can get a true Mhz to Mhz comparison across the two different platforms)

I dig nerdy stuff like this, and you guys dig reading nerdy stuff like this because it is always much better to have some dorko-nerd actually do all the testing instead of just listening to know-it-all blowhards trying to impress their Counter-Strike buddies that they know more than we do in this forum :P


Or it saves people like me that do all the anal tests the trouble of typing it out :)

#21 Toshigdh

Toshigdh

    New Member

  • Members
  • 4 posts

Posted 30 December 2006 - 10:46 PM

Just wanted to say thank you for all this testing. I just recently got the DFI lanparty nf4 expert. it was a big jump for me. I have always built my own pc's sence 386's, but never did any overclocking. So i decided to jump in with both feet. and have been reading these bords like crazy, and this thread right here has helped me tons :). I only have my board runing at stock speeds right now, except the ram, it is overclocked. but when i finaly absorb enough knowlage from these boards to feal comfortable in doing some minor oc, it will be concentrated on cpu mhz. and possably some GPU as well and ill let the ram fall where it may.

#22 Cardnyl

Cardnyl

    New Member

  • Members
  • 5 posts

Posted 01 March 2007 - 10:44 AM

This post completely changed my understanding of memory when overclocking. Essentially you have saved me from having to buy some god awful expensive ram and I thank you for that.

#23 scruffy1

scruffy1

    New Member

  • Members
  • 18 posts

Posted 01 April 2007 - 02:05 AM

hooray ! thanks for your common sense dissertation that saved me from buying even faster ram (tccd) when i had a pocket full of bh5 that i knew wanted to come with me to the next mobo - along with all my other hardware except the xp-m

as per the sig - not yet burnt in and tested to the max - it sure helped me on the road to hardware abuse :P

my next step after getting it wired will be to see how well two pairs of slightly differing bh5's can handle a similar pace

and then start turning up the softmodded 6800gs agp........

#24 Guest_sutahz_*

Guest_sutahz_*
  • Guests

Posted 06 June 2007 - 06:47 PM

This is exactly what I came to these DFI forums for. I dont know why, but I wanted to know if dividers hurt performance w/ A64's (cause i owned a NF7-S r2.0, so back then, yes CL2 was very desirable). I dont care about benchmarks at all, all I care about is real world results. I bought this board because it was the best nF4 OC'r I could find, and now that I have a GPU to compliment my cpu (had a 7600GT before, hold me over until 8800's came down in price), now I can look into OC'ing it. Im sure im either super n00b or got a poor OC'ing 4400+ because the max OC i could attain was 2.6GHz at 1.64Vcpu..... not good when the author of this post is running 250x11 at stock voltage (i 238x11 was stable enough for me, but didnt pass 24hrs of p95). SO, this post showed me that now tighter timings isn't what its all about. Never took my ram outta 2-3-2-5, think it went up to like 215 w/ that... now im at 250 w/ 3-4-4-8 and running smooth in memtest :) Now i need to find exactly where to post for OC'ing help :)