[coastie65]

Rommel, on 31 October 2010 - 07:13 PM, said:

Nice OCing Snyper.

Here's my test results.

http://valid.canardp....php?id=1458889

http://i18.photobuck...n_/Capture2.png

Nice idea to start this thread compnovo.

Rommel

EDIT:

Nova results,
http://novabench.com/view/27989

I had to use Rommel44 because Rommel was already in use.

Hey Rommel, Good one. Looks like the OCing paid off. I am not a fan of OCing and if I need speed I just buy it in the form of a faster GPU, CPU, and/or Memory. Unfortunately, on this rig, although the MOBO will support up to 1600 MHz ram ( which I am running ), Intel has it locked down to 1066 Mhz in the memory controller ( Integrated into the Core i7 ). Oh well, such is life. At least I can up the CPU speed, but that Core i7 960 is going for $570, which is a bit pricey. Have pretty decent GPU in the HD5770, so that is fine.

[Rommel]

Hey guys,
retested with OC on the graphics card.
Its factory OC is GPU 875MHZ
memory clock 1250 MHZ
In manual it allows on the sliders
GPU 960MHZ
memory 1445 MHZ

It tested that I can run there but my bench testing disagreed.
Setted the values at.
GPU 935
Memory 1400

This gave noticable gains.

3Ds old over-all score was 16797 new 17427
Nova old graphics 421 new 457.

Perhaps this weekend I'll play with the OCing the cpu again?

Thanks for checking my results.

3D OLD
http://i18.photobuck...n_/Capture2.png

3D NEW
http://service.futur...3&resultType=14 <-- link doesn't show results anymore. Because I wasn't registered?
I did register after I got the link. Over-all score is noted above.

NOVA OLD
http://novabench.com/view/27989

NOVA new
http://i18.photobuck...Cedgraphics.jpg

Rommel

This post has been edited by Rommel: 01 November 2010 - 04:13 PM

[SnyperTodd]

waldojim, on 01 November 2010 - 01:34 AM, said:

remember me saying I was going to try something crazy?
Check this out! I am not sure if the FSB will help or not, but it seems perfectly stable!

http://valid.canardp....php?id=1459285

This "crazy" one looks much better, WJ, just because of the much lower Vcore, but a high FSB doesn't do anything for performance in and of itself on an AMD unlike the S775 Intels. Still, if it's stable and allows you to get away with that voltage, that's a good thing. I'm surprised you found an FSB hole on an AMD chipset, that's pretty unusual. It's a trademark on nVidia chipsets, but unusual on an AMD... Now you just need to bring that RAM back up to speed.

Rommel, you've got the best performing HD5770 in the thread I think. You should hang on to that thing, it's a good one. If you need a refresher this weekend, you know how to get ahold of me.

This post has been edited by SnyperTodd: 01 November 2010 - 04:04 PM

[Rommel]

SnyperTodd, on 01 November 2010 - 04:04 PM, said:

If you need a refresher this weekend, you know how to get ahold of me.

Your a good man snyper but I know I may catch you at a bad time with all your going ons rebuilding.

I did find some notes so I'll try a memory recall when I revisit bios.

Thanks man!

Rommel

@ waldojim,

As snyper said, that crazy but you have been down this road a lot.
A curious abservation though, with your FBS at 296 and a 12.5 multiplier you are at 3700MHZ
When you were at 250 FBS you were at 3750. Don't recall your multiplier as I post here.
Unless I missed something which I probably did, not sure why you are as far as MHZ go, 50 less??

Curious George.

This post has been edited by Rommel: 01 November 2010 - 04:25 PM

[coastie65]

Hey Rommel, I have to agree with snyper in that you seem to have the best performing 5770 of the lot.

[SnyperTodd]

Rommel, on 01 November 2010 - 04:23 PM, said:

@ waldojim,

As snyper said, that crazy but you have been down this road a lot.
A curious abservation though, with your FBS at 296 and a 12.5 multiplier you are at 3700MHZ
When you were at 250 FBS you were at 3750. Don't recall your multiplier as I post here.
Unless I missed something which I probably did, not sure why you are as far as MHZ go, 50 less??

Curious George.

I can't speak specifically for Waldojim, but the most likely reason he's running the lower multiplier and therefor lower clock speed is because if he ran a multiplier of 13, he'd be at 3.85GHz, which my not be stable or may require more voltage than he wants to give it or there could be any number of other reasons he has for not running it. (Wow, sorry about the run on sentence...) With the older Intels, increasing the FSB actually did increase performance, even at lower core speeds under many circumstances. In those processors, the FSB was the communication path between the processor and the rest of the system; the faster it ran, the faster the overall system would be. With AMD processors (and Nehalems) what used to be the FSB is now just a reference clock, and the speed of the various communication buses are based on that reference clock and adjustable with individual multipliers. Sometimes it's worth the trade off in core speed to set the reference clock where something else (like the NB or RAM, for example) will run faster than it would at the lower ref clock and higher core multiplier. Know what I mean?

[Rommel]

SnyperTodd, on 01 November 2010 - 05:52 PM, said:

Rommel, on 01 November 2010 - 04:23 PM, said:

@ waldojim,

As snyper said, that crazy but you have been down this road a lot.
A curious abservation though, with your FBS at 296 and a 12.5 multiplier you are at 3700MHZ
When you were at 250 FBS you were at 3750. Don't recall your multiplier as I post here.
Unless I missed something which I probably did, not sure why you are as far as MHZ go, 50 less??

Curious George.

I can't speak specifically for Waldojim, but the most likely reason he's running the lower multiplier and therefor lower clock speed is because if he ran a multiplier of 13, he'd be at 3.85GHz, which my not be stable or may require more voltage than he wants to give it or there could be any number of other reasons he has for not running it. (Wow, sorry about the run on sentence...) With the older Intels, increasing the FSB actually did increase performance, even at lower core speeds under many circumstances. In those processors, the FSB was the communication path between the processor and the rest of the system; the faster it ran, the faster the overall system would be. With AMD processors (and Nehalems) what used to be the FSB is now just a reference clock, and the speed of the various communication buses are based on that reference clock and adjustable with individual multipliers. Sometimes it's worth the trade off in core speed to set the reference clock where something else (like the NB or RAM, for example) will run faster than it would at the lower ref clock and higher core multiplier. Know what I mean?

Thanks, I actually absorbed some of that. Some.
The proof of the slightly lower raw speed vs. over-all bench score. In waldojims case was 20502.

My FSB is 228 with 15.5 multiplier. Perhaps I should raise my FSB and lower my multiplier and retest.
As you said, there are still other factors besides these.

I'll keep that in mind to not just focus on cpu speed.

Rommel

[SnyperTodd]

Rommel, on 01 November 2010 - 06:11 PM, said:

Thanks, I actually absorbed some of that. Some.
The proof of the slightly lower raw speed vs. over-all bench score. In waldojims case was 20502.

My FSB is 228 with 15.5 multiplier. Perhaps I should raise my FSB and lower my multiplier and retest.
As you said, there are still other factors besides these.

I'll keep that in mind to not just focus on cpu speed.

Rommel

As you raise your ref clock, everything will start to get faster. Some of it may be able to handle it, and some of it may not. You will have to lower some multipliers to keep things stable. Just remember the golden rule: only change one thing at a time and you'll never wonder what component is making you unstable.

[waldojim]

Rommel, on 01 November 2010 - 05:19 AM, said:

SnyperTodd, on 31 October 2010 - 08:10 PM, said:

I'm impressed, Rommel! Awesome score! Running 4 sticks even at 1200MHz with timings in the 7s isn't an easy feat on an AMD memory controller. Are you OCing your video card? That's a great score for a 5770, too.

Thank you Snypertodd.

My card is factory OCed. http://www.zipzoomfl...ctCode=10012559

I haven't messed with the OCing tool but perhaps I will just to see what it can do for a bench test than bring it back to its factory speed.

@ waldojim:

You guys bring out new terms I have to google. FSB hole.
Nice adjustment on your rig.

Perhaps thats my block?
When my cpu was X 2 I got it around 3700. After unlocking the other two cores that number was unstable though my cores pasted a harsh CPU test.

Must admit I haven't played with voltages being that seems more destroying than tweeking other values.

I'm going to look through my folder for snypers info and remind myself of what he thought me.

Rommel

There is actually a great deal of play on the voltages, and The MSI bios will go red when you get to voltages you need to stay away from (1.68 I think on mine). But the 1.4-1.5 range should be good for OCing up to the 4Ghz mark, assuming your two extra cores are stable enough.

[waldojim]

Rommel, on 01 November 2010 - 03:53 PM, said:

Hey guys,
retested with OC on the graphics card.
Its factory OC is GPU 875MHZ
memory clock 1250 MHZ
In manual it allows on the sliders
GPU 960MHZ
memory 1445 MHZ

It tested that I can run there but my bench testing disagreed.
Setted the values at.
GPU 935
Memory 1400

This gave noticable gains.

3Ds old over-all score was 16797 new 17427
Nova old graphics 421 new 457.

Perhaps this weekend I'll play with the OCing the cpu again?

Thanks for checking my results.

3D OLD
http://i18.photobuck...n_/Capture2.png

3D NEW
http://service.futur...3&resultType=14 <-- link doesn't show results anymore. Because I wasn't registered?
I did register after I got the link. Over-all score is noted above.

NOVA OLD
http://novabench.com/view/27989

NOVA new
http://i18.photobuck...Cedgraphics.jpg

Rommel

DANG! very nice OC on the video cards - wish mine would get to 1400 on the ram!

[waldojim]

SnyperTodd, on 01 November 2010 - 04:04 PM, said:

waldojim, on 01 November 2010 - 01:34 AM, said:

remember me saying I was going to try something crazy?
Check this out! I am not sure if the FSB will help or not, but it seems perfectly stable!

http://valid.canardp....php?id=1459285

This "crazy" one looks much better, WJ, just because of the much lower Vcore, but a high FSB doesn't do anything for performance in and of itself on an AMD unlike the S775 Intels. Still, if it's stable and allows you to get away with that voltage, that's a good thing. I'm surprised you found an FSB hole on an AMD chipset, that's pretty unusual. It's a trademark on nVidia chipsets, but unusual on an AMD... Now you just need to bring that RAM back up to speed.

Rommel, you've got the best performing HD5770 in the thread I think. You should hang on to that thing, it's a good one. If you need a refresher this weekend, you know how to get ahold of me.

I am working on that memory issue next... but the interesting thing to note here - I found the 296FSB by using the MSI "max front side bus" detect in the BIOS - I turned the multiplier down to 11x to test that, and 30 seconds later, I have a 3+Ghz chip! (don't remember the exact number, but surprised the heck outta me!) I think what I may be dealing with is an odd divider for the PCI and PCI-E buses, and that MAY be why the 296 is more stable. I am going to try later and see what I can get the ram to do, though it is VERY pick about speeds (nothing over 1670 at all!)

[waldojim]

Rommel, on 01 November 2010 - 04:23 PM, said:

SnyperTodd, on 01 November 2010 - 04:04 PM, said:

If you need a refresher this weekend, you know how to get ahold of me.

Your a good man snyper but I know I may catch you at a bad time with all your going ons rebuilding.

I did find some notes so I'll try a memory recall when I revisit bios.

Thanks man!

Rommel

@ waldojim,

As snyper said, that crazy but you have been down this road a lot.
A curious abservation though, with your FBS at 296 and a 12.5 multiplier you are at 3700MHZ
When you were at 250 FBS you were at 3750. Don't recall your multiplier as I post here.
Unless I missed something which I probably did, not sure why you are as far as MHZ go, 50 less??

Curious George.

As Snyper mentioned, the multiplier is lower now. I tried going to 13x, but the chip won't run stable at 3.9Ghz. The one down side to these high clock speeds is that the multipliers have MUCH greater effect. The only truly bright side here, is that it DID open my machine up to a new frequency band, and hopefully I can refine it down a touch further now.

Side note: I re-ran all the benchmarks to check for a performance difference, and there appears to be no impact what-so-ever. Even with the slight loss in raw clock speed.

[waldojim]

SnyperTodd, on 01 November 2010 - 05:52 PM, said:

Rommel, on 01 November 2010 - 04:23 PM, said:

@ waldojim,

As snyper said, that crazy but you have been down this road a lot.
A curious abservation though, with your FBS at 296 and a 12.5 multiplier you are at 3700MHZ
When you were at 250 FBS you were at 3750. Don't recall your multiplier as I post here.
Unless I missed something which I probably did, not sure why you are as far as MHZ go, 50 less??

Curious George.

I can't speak specifically for Waldojim, but the most likely reason he's running the lower multiplier and therefor lower clock speed is because if he ran a multiplier of 13, he'd be at 3.85GHz, which my not be stable or may require more voltage than he wants to give it or there could be any number of other reasons he has for not running it. (Wow, sorry about the run on sentence...) With the older Intels, increasing the FSB actually did increase performance, even at lower core speeds under many circumstances. In those processors, the FSB was the communication path between the processor and the rest of the system; the faster it ran, the faster the overall system would be. With AMD processors (and Nehalems) what used to be the FSB is now just a reference clock, and the speed of the various communication buses are based on that reference clock and adjustable with individual multipliers. Sometimes it's worth the trade off in core speed to set the reference clock where something else (like the NB or RAM, for example) will run faster than it would at the lower ref clock and higher core multiplier. Know what I mean?

And you nailed it BTW - Cores 1 and 4 just aren't stable at 3850 - I am going to try using ACC to reduce their clock speeds 1 or 2% and see if that is enough to stabilize it. I would like to see 3.8Ghz run well.

[SnyperTodd]

waldojim, on 01 November 2010 - 08:28 PM, said:

I am going to try later and see what I can get the ram to do, though it is VERY pick about speeds (nothing over 1670 at all!)

I've noticed Patriot's DDR3 doesn't seem to have the insane overhead that their DDR2 had. I've got a set of Patriot DDR3 1333 that doesn't like anything over about 1600MHz. I have three 2GB sets (the fact that they're 2GB has something to do with it, I'm sure) of Patriot Extreme Performance DDR2-800 that will all run over 1100MHz @ 5-5-5-18 and one of those sets will run 1200MHz (but wants 5-8-8-24 to be stable (hey, it's a 50% OC!)). Even my Redlines don't have the overhead those Patriots do, but they'll do tighter timings and they're happier at slightly lower voltages.

Here's the Patriots at 1125MHz: http://valid.canardp...c.php?id=404776
And here they are at 1200MHz: http://valid.canardp...c.php?id=332111

Both of those were rock solid overclocks. Before you guys go all crazy on me, Sabre is the other name I've used for years prior to Snypertodd. Here's a screenshot of my CPU-Z account to show that I am legit:

This post has been edited by SnyperTodd: 01 November 2010 - 09:11 PM
Reason for edit: it's late. My grammar suffers when I'm tired.

[SnyperTodd]

waldojim, on 01 November 2010 - 08:28 PM, said:

I think what I may be dealing with is an odd divider for the PCI and PCI-E buses, and that MAY be why the 296 is more stable.

Can't you lock those to 33MHz and 100MHz, respectively? On my AMD boards those particular frequencies are set manually, not with a divider. I figured they had their own PLLs because they were far to sensitive to deal with the wide swings a divider off the ref clock would give them. I don't know. It's past my bedtime.

[waldojim]

SnyperTodd, on 01 November 2010 - 09:17 PM, said:

waldojim, on 01 November 2010 - 08:28 PM, said:

I think what I may be dealing with is an odd divider for the PCI and PCI-E buses, and that MAY be why the 296 is more stable.

Can't you lock those to 33MHz and 100MHz, respectively? On my AMD boards those particular frequencies are set manually, not with a divider. I figured they had their own PLLs because they were far to sensitive to deal with the wide swings a divider off the ref clock would give them. I don't know. It's past my bedtime.

They may be... I don't know with this board. I know in the past, my AMD K8N Platinum (MSI) and the K7N2 Delta-L ran the timings from reference... but those were older boards too. I am trying to understand WHY I would have such gaping holes in my OC, and that would sort of make sense.

Right now, I am at 3.76 Ghz, and testing. Seems like the CPU hits the limits here, heat is not an issue - haven't seen anything over 55 yet, but it gets stupidly unstable anything over... Thought it was my HT link there for a bit (2.1Ghz) but I tried cutting that back as well... no dice. After I prove to myself that this is stable, I will start working on ram timings.... that is going to be a PITA.

edit: fyi, just to be clear, there are MANY timings, voltages, etc that I do NOT understand 100% So I am just working with trial and error at the moment, and hoping nothing dies in the meantime. The motherboard does have Failed OC protection though, and I am THANKFUL that I do not have to remember all the "good" bios settings when something goes wrong.

This post has been edited by waldojim: 01 November 2010 - 09:38 PM

[SnyperTodd]

waldojim, on 01 November 2010 - 09:37 PM, said:

They may be... I don't know with this board. I know in the past, my AMD K8N Platinum (MSI) and the K7N2 Delta-L ran the timings from reference... but those were older boards too. I am trying to understand WHY I would have such gaping holes in my OC, and that would sort of make sense.

Hmmm, the K8N definitely should not have run the PCI frequencies from the reference clock, you should have been able to manually set them (it is Socket 939, right?). However, the K7N would have, since it is a Socket A board and those processors had a true FSB more like the pre-Nehalem Intels. Some of them did allow the PCI frequency to be locked or unlinked, though, which obviously helped immensely with overclocking.

edit: Just found a review that confirmed what I suspected- the K8N Platinum allows independent selection of the PCIe frequency: http://www.pcstats.c...eid=1770&page=3 I don't see any option for PCI frequency, maybe it's fixed at 33MHz?

This post has been edited by SnyperTodd: 02 November 2010 - 05:10 PM

[waldojim]

SnyperTodd, on 02 November 2010 - 05:02 PM, said:

Hmmm, the K8N definitely should not have run the PCI frequencies from the reference clock, you should have been able to manually set them (it is Socket 939, right?). However, the K7N would have, since it is a Socket A board and those processors had a true FSB more like the pre-Nehalem Intels. Some of them did allow the PCI frequency to be locked or unlinked, though, which obviously helped immensely with overclocking.

edit: Just found a review that confirmed what I suspected- the K8N Platinum allows independent selection of the PCIe frequency: http://www.pcstats.c...eid=1770&page=3 I don't see any option for PCI frequency, maybe it's fixed at 33MHz?

Yeah that would make some sense... it has been a LONG time since I touched that K8N - that sorry sucker never did want to OC for crap. The only independent clock on the k7n was the AGP clock, and I did have fun running that up to 85Mhz or so. Not sure it really did any good, but it was fun to do and say I did it.

As for the current machine - I did find the PCI-E clock (hiding with something else) but no PCI clock adjustments - or even statistics. I got frustrated trying to adjust the timings though and gave up for now. Running 7-7-7-18 ATM, which I think is freaking terrible.

[SnyperTodd]

waldojim, on 03 November 2010 - 04:05 AM, said:

I got frustrated trying to adjust the timings though and gave up for now. Running 7-7-7-18 ATM, which I think is freaking terrible.

That's not that bad for 1580MHz- a lot of sticks spec 9's across the board at 1600MHz. My Patriots don't mind a little tighter tRP at lower speeds (<1600MHz). Have you tried something like that? It might be all you can do without tweaking some voltages. I don't know what's adjustable on your GD-70, can you change your DRAM pull-up, pull-down, and/or reference voltages? They will have some effect if you can move them. On my X48 board tweaking the NB can give a little more headroom to the RAM, but probably won't do much on an AMD system. At this point it becomes a lot of trial and error for every little gain.

Here is where I ran my Patriots 1333LLKs with the E5200: http://valid.canardp...c.php?id=838065

This post has been edited by SnyperTodd: 03 November 2010 - 07:55 PM

[waldojim]

SnyperTodd, on 03 November 2010 - 07:54 PM, said:

waldojim, on 03 November 2010 - 04:05 AM, said:

I got frustrated trying to adjust the timings though and gave up for now. Running 7-7-7-18 ATM, which I think is freaking terrible.

That's not that bad for 1580MHz- a lot of sticks spec 9's across the board at 1600MHz. My Patriots don't mind a little tighter tRP at lower speeds (<1600MHz). Have you tried something like that? It might be all you can do without tweaking some voltages. I don't know what's adjustable on your GD-70, can you change your DRAM pull-up, pull-down, and/or reference voltages? They will have some effect if you can move them. On my X48 board tweaking the NB can give a little more headroom to the RAM, but probably won't do much on an AMD system. At this point it becomes a lot of trial and error for every little gain.

Here is where I ran my Patriots 1333LLKs with the E5200: http://valid.canardp...c.php?id=838065

Well, they run 1600 8-8-8-24 stock at 1.65V bumping up to 1.8 doesn't seem to help anything.7-7-7-18 is as tight as they will go. The system refuses to even BOOT if I try 6-7-7-18 (put the 6 anywhere - it doesn't matter) until I hit 1.9V - and I am not about to leave the system running that way. I might try to lower the clock speed some on the ram - but I am not sure I want to run ~1250Mhz.

With the X48, the memory controller is still on the motherboard, giving you a bit more flex there depending on the board itself... me - I think I have a processor holding me back.

edit: BTW - for the sake of a random bit of coolness - I played with my Wifes PC a bit - with the extra 2 cores unlocked it will OC to 3.5Ghz, but the impressive part was with only two cores - I was at 3.7 with no effort at all - just set the multiplier, and go. Didn't even change the voltage. That collermaster kept that machine humming along at room temp even! I will toy around more with it later... curious what the dual core can do, and how it fares in game.

This post has been edited by waldojim: 03 November 2010 - 11:05 PM