- Nov 9, 2004
- 11,465
- 1
- 0
Dado, you need to isolate(refer to zebos o/c guide)
It looks like all your testing has been with ram running 1:1, and I haven't seen you mention any memtest results? Without isolating you don't know whats giving out on you, CPU, Ram, or memory controller.
Test for your max HTT by lowering CPU multi to like 8 and HT multi to 3x and use a low mem divder like 133 or 100, and see how high you can get HTT. This is your HTT max
Test ram by setting with cpu multi at 8, and a ram setting of 200 1:1 and max DDR volts that your comfortable with, and loose timings, and see how high you can get stable up to your HTT max. (Ram will always be limited by your max HTT). This is your max Ram
Then test CPU by setting max vcore 11x multi, setting mem divider to 133 or 100 and raising HTT until you loose stability. This is your max CPU
Take this approach and report back with your max HTT, Ram, and CPU and I'm sure the guys here especially Duvie since he has your board, can help you "Consolidate" your results for your best overclock.
My 2.8ghz is acheived with a memory divider, noway I can run 2.8ghz 1:1. A64's are divider friendly
, its been proven time and again that the very small performance hit from running your ram on a divder is easily overcome by the higher mhz you can get.
My settings with CPU at 2.8 and ram at 200mhz will beat 2.7 with ram at 270mhz every day
*edit*
I,m seeing a trend here. It looks like to me a large portion or people having trouble overclocking Venice and Sandiegos here and on other forums are hung up on running ram 1:1 and aren't "Isolating and Consolidating". As much as people don't want to believe it, the A64's ondie memory controller usually means running 1:1 is not a good idea
*edit again*
Try these settings
HT 3x
10 x 280 vcore max (on my board its 1.58-1.60 actual)
mem div 133 stock ddr volts and timings
It looks like all your testing has been with ram running 1:1, and I haven't seen you mention any memtest results? Without isolating you don't know whats giving out on you, CPU, Ram, or memory controller.
Test for your max HTT by lowering CPU multi to like 8 and HT multi to 3x and use a low mem divder like 133 or 100, and see how high you can get HTT. This is your HTT max
Test ram by setting with cpu multi at 8, and a ram setting of 200 1:1 and max DDR volts that your comfortable with, and loose timings, and see how high you can get stable up to your HTT max. (Ram will always be limited by your max HTT). This is your max Ram
Then test CPU by setting max vcore 11x multi, setting mem divider to 133 or 100 and raising HTT until you loose stability. This is your max CPU
Take this approach and report back with your max HTT, Ram, and CPU and I'm sure the guys here especially Duvie since he has your board, can help you "Consolidate" your results for your best overclock.
My 2.8ghz is acheived with a memory divider, noway I can run 2.8ghz 1:1. A64's are divider friendly
, its been proven time and again that the very small performance hit from running your ram on a divder is easily overcome by the higher mhz you can get.My settings with CPU at 2.8 and ram at 200mhz will beat 2.7 with ram at 270mhz every day
*edit*
I,m seeing a trend here. It looks like to me a large portion or people having trouble overclocking Venice and Sandiegos here and on other forums are hung up on running ram 1:1 and aren't "Isolating and Consolidating". As much as people don't want to believe it, the A64's ondie memory controller usually means running 1:1 is not a good idea
*edit again*
Try these settings
HT 3x
10 x 280 vcore max (on my board its 1.58-1.60 actual)
mem div 133 stock ddr volts and timings
I can see why running 1:1 might not be as necessary as with P4s and AXPs, but why would it "not be a good idea" as opposed to running with a divider?
- Nov 9, 2004
- 11,465
- 1
- 0
Because with the ondie mem controller it reduces your max CPU overclock when you run 1:1, and sacrificing CPU speed for ram speed is never good on A64's.
Personally I think its a vcore thing. Both the CPU and memcontroller are in competition for the available voltage, thats why the guys on extreme cooling and volt moded boards are more successful running 1:1 because they can pump the necessary voltage to the CPU, plus stability is not a requirement in extreme overclocking. The popularity of Extreme overclocking probabally accounts for so many new overclockers desiring to run high ram speeds when its really meaningless to real world performance.
Personally I think its a vcore thing. Both the CPU and memcontroller are in competition for the available voltage, thats why the guys on extreme cooling and volt moded boards are more successful running 1:1 because they can pump the necessary voltage to the CPU, plus stability is not a requirement in extreme overclocking. The popularity of Extreme overclocking probabally accounts for so many new overclockers desiring to run high ram speeds when its really meaningless to real world performance.
Wow... so, uhh... thats the first time I've heard that one.
So, lets say hypothetically your ram can do 260FSB and your CPU can handle 250HTT (at any given multipler) before it maxes out. Now, are you saying that if you tried to run your ram 1:1 with your HTT, that your CPU would no longer be able to hit its maximum HTT? Or, put another way, that trying to run your RAM 1:1 with your HTT actually reduces the maximum potential speed of your CPU? If so, wouldnt an increase in Vcore fix this?
People seem to be having great overclocks running 1:1 with high speed OCZ and Corsair?
Edit: There is a guy in the OCZ help thread that is running a Venice at 2.8Ghz with 4 sticks of OCZ Gold 1:1 at CL2, 267mhz. I would imagine thats putting a hell of a lot of stress on that memory controller.
So, lets say hypothetically your ram can do 260FSB and your CPU can handle 250HTT (at any given multipler) before it maxes out. Now, are you saying that if you tried to run your ram 1:1 with your HTT, that your CPU would no longer be able to hit its maximum HTT? Or, put another way, that trying to run your RAM 1:1 with your HTT actually reduces the maximum potential speed of your CPU? If so, wouldnt an increase in Vcore fix this?
People seem to be having great overclocks running 1:1 with high speed OCZ and Corsair?
Edit: There is a guy in the OCZ help thread that is running a Venice at 2.8Ghz with 4 sticks of OCZ Gold 1:1 at CL2, 267mhz. I would imagine thats putting a hell of a lot of stress on that memory controller.
- Nov 9, 2004
- 11,465
- 1
- 0
On my setup I have to give up 40-50mhz CPU speed to run 1:1 even if using 2T command rate. And don't believe everything you see on the forums, its easy to get CPU-Z screenies and short SuperPI runs with a totally unstable systems. And its also easy to show stability tests run at slower speeds with CPU-Z showing much higher numbers.
The bottom line is, current A64 archeticture can't use all the memory bandwidth available at stock speeds. When you overclock ram you are providing extra resources that the system can't utilize. So test for yourself, and if you can truly run your max CPU overclock at 1:1 then go with it, but if you have to give up CPU mhz to get it stable, don't do it! Drop your memory instead
The bottom line is, current A64 archeticture can't use all the memory bandwidth available at stock speeds. When you overclock ram you are providing extra resources that the system can't utilize. So test for yourself, and if you can truly run your max CPU overclock at 1:1 then go with it, but if you have to give up CPU mhz to get it stable, don't do it! Drop your memory instead
- Nov 9, 2004
- 11,465
- 1
- 0
- Nov 9, 2004
- 11,465
- 1
- 0
Dadofamunky
Platinum Member
- Jan 4, 2005
- 2,184
- 0
- 0
Originally posted by: GuitarDaddy
Dado, you need to isolate(refer to zebos o/c guide)
Take this approach and report back with your max HTT, Ram, and CPU and I'm sure the guys here especially Duvie since he has your board, can help you "Consolidate" your results for your best overclock.
My 2.8ghz is acheived with a memory divider, noway I can run 2.8ghz 1:1. A64's are divider friendly
My settings with CPU at 2.8 and ram at 200mhz will beat 2.7 with ram at 270mhz every day
*edit*
I,m seeing a trend here. It looks like to me a large portion or people having trouble overclocking Venice and Sandiegos here and on other forums are hung up on running ram 1:1 and aren't "Isolating and Consolidating". As much as people don't want to believe it, the A64's ondie memory controller usually means running 1:1 is not a good idea
*edit again*
Try these settings
HT 3x
10 x 280 vcore max (on my board its 1.58-1.60 actual)
mem div 133 stock ddr volts and timings
Agree, in fact I'm working on that stuff right now.... No longer running 1:1 - I don't think OC'ers will benefit from the updated memory controller if they still wanna run 4 stix of RAM. I do know from previous work that my RAM was rock stable at 267 9x267 w- Winnie 3000+) so I believe I have a baseline there.
Although I'm running into another irritating problem - on Windows 2K, the MSI CoreCenter application will not let me run my FSB beyond 250MHz. It locks the FSB value. I have changed the multiplier and discovered that - I set the BIOS to 10x275, and when Windows came up, the system was still running at 2.5 GHz! (Verified with other utilities.) Rebooted and reset multi to 11x temporarily.
***EDIT*** Oh hell, I know why. I need to change the HT setting. DOH! Never mind.
Thus far at 2.75 GHz (11x275) I have reached personal bests in the following: :thumbsup:
3DMark03 13327 (Shattered 13K barrier for first time)
3DMark05 5618
PCMark 5418
Aquamark 3 70,707 (Broke 70K for first time)
Aquamark CPU 11,040
This CPU is some SERIOUS artillery.
BUT! Prime still doesn't run at 11x250, 166 divider. It's crashing after a few minutes instead of immediately
so it IS getting better, but I'm obviously not satisfied. Call me crazy but I don't like getting BSODs unless it's due to the software jocks. I'm now going to step through more of the items you've mentioned. Also now running Duvie's 1.62b2 BIOS which he kindly sent me.
RichUK
Lifer
- Feb 14, 2005
- 10,341
- 678
- 126
Originally posted by: GuitarDaddy
I guess the insane overclocks that were seen by some of the eary Winchester 3000+'s got everyones expectations up too high
All I've seen for the last couple days is tons of Wha Wha Wha, my venice sucks it will only do 2.7, and so many people expecting 3gz on air from 1.8-2.2 chips. Personally I,m more than happy with 2.8ghz
MAN O MAN ... i am SOOOOOOOOOOOOO glad that i did not rush out and get a venice, i was so so close to doing so, but ended up saving for a new 19" tft
, i have now lost total interest in them now .. slightly interested in the SD, but not overly, might have a silly moment tho and purchase an FX57 and get 3.2/3.4 on air .. the way i see it, i have a winnie that does 2.53 rock stable, and i am not repeat not !! spending money on an another chip to get like 170Mhz more .. NOOO WAAAYYY...I think these dual cores (X2) are looking sexier and sexier over each poor OC i see on the venice .. i would have bought the chip if i could get full potential from my RAM (TCCD) and scale upto like 3Ghz, but that seems a no go so far..
GOD DAMN .. i am so glad that i didnt go and buy one of these chips, i am so dissapointed, i think the reviews were big time cherry picked chips ... :|
EDIT: BUY is spelt "buy" not "by"
Dadofamunky
Platinum Member
- Jan 4, 2005
- 2,184
- 0
- 0
Yeah, you have no real reason to switch at this point, what you have is solid, what's a couple hundred MHz between friends?
I swear, OCing is like heroin, only more expensive.
blckgrffn
Diamond Member
- Nov 9, 2004
- 11,465
- 1
- 0
I believe that the 1mb L2 cache doesn't provide a big advantage currently, but once Window64 and 64bit apps become common it will become more important. Thats the main reason I got the San Diego vs. Venice
Dadofamunky
Platinum Member
- Jan 4, 2005
- 2,184
- 0
- 0
Well, I have some big success to report. I am now Prime Torture Test stable for 5 hrs. and counting at the following settings:
HTT 3T
FSB 10x267, CPU at 2.670 GHz
2 GB Memory at 3-4-4-8 2T 166 divider, 222 MHz
1.62 BIOS
Voltage in CPU-Z 1.539V Set to 1.45V +5% in BIOS
Chip info: ADA3700DAA5BN CABGE0515VPMW
I have a lot more to do, but I think I've found my baseline stable OC. An extra 4.66 GHz on the 3700+ San Diego chip. A 20% OC - not spectacular, but not bad either. I think I have some more headroom. Now it's time to retire and let this thing burn in for awhile.
HTT 3T
FSB 10x267, CPU at 2.670 GHz
2 GB Memory at 3-4-4-8 2T 166 divider, 222 MHz
1.62 BIOS
Voltage in CPU-Z 1.539V Set to 1.45V +5% in BIOS
Chip info: ADA3700DAA5BN CABGE0515VPMW
I have a lot more to do, but I think I've found my baseline stable OC. An extra 4.66 GHz on the 3700+ San Diego chip. A 20% OC - not spectacular, but not bad either. I think I have some more headroom. Now it's time to retire and let this thing burn in for awhile.
Dadofamunky
Platinum Member
- Jan 4, 2005
- 2,184
- 0
- 0
Originally posted by: Elfear
Wow! Show me which settings you used to get that!!! Lol. Sorry, couldn't resist
***Blush***
Sorry, it was gettin' late....
- Nov 9, 2004
- 11,465
- 1
- 0
:thumbsup: Good news! and your still @ 1.45 +5%, got some room to work with. What are your temps when priming?
I have found that in my setup the SD really doesn't like overclocking the memory. It maybe a bios issue with my board? If I run a moderate mem o/c like you 220-230mhz, I get stuck right around 2.7ghz, but if I drop the memory to around 200mhz I can run at 2.8ghz (it seems to like stock FX57 speed). And 100mhz of CPU will clobber 20-30mhz of ram speed. Using A64 tweakers custom dividers I was able to run a 32m SuperPI at 2.8ghz with ram at 215mhz but I didn't prime test at those settings.
Dadofamunky
Platinum Member
- Jan 4, 2005
- 2,184
- 0
- 0
:thumbsup: More good news. I am experiencing exactly the issue you mention - I'm now at 10x270, 166 divider and Prime stable for 4 Hrs. and counting. Also tried the following combos:
10x280 - Prime fails immediately
10.5x260 - Prime fails immediately
If it is a BIOS issue with your board, then yours is not the only board.
I can't measure temps accurately because of Duvie's mentioned BIOS bug with measuring temps. It shows as 27C idle and 32C load regardless of the utility. It is running like a tank at 10x270. My best-educated guess (after using a Winne 3000+ for a stable 33% OC in the same board with 1.41 BIOS - see sig) is that the San Diego CPU is actually running 10-13C above the reported levels on both ends respectively.
Yeah, I'm just dispensing with all the toys and using Prime right now. (I think SuperPI is a really good tool, too.) I don't want to spend weeks more getting this thing in shape (I have a family to raise and music to play, you know). Thanks for the tip. I'm going to tune my divider again.
After all this, I'm starting to get a little tired of the tendency that Taiwan board makers have with a certain lack of attention to detail. I mean c'mon, not reporting TEMPS accurately after previously fixing the bug? Don't these guys use ClearCase or ClearQuest to track bugs and make sure they STAY resolved? Hell, I know how to use that stuff and I'm not even an engineer! :roll:
It almost (but not quite) makes me miss the days when a 386 mobo cost $750 (as did the CPU) but they were made in the US and they were almost always rock solid. I used to roll my own when the 286-10 was hot sh*t and 4 MB of RAM was sick. Then again, this board only cost me $110 US. And at least they're not too shy about letting beta BIOes out the door. Most of them are probably freak tweakers like we are.
Mrvile
Lifer
- Oct 16, 2004
- 14,066
- 1
- 0
I just read through all 4 pages and I'm drooling. Too bad I just purchased this 3500+ winny a few months ago, My next upgrade won't be for another year maybe. But seriously guys, 2.8ghz is unheard of on any 3x00+ on air, I would be happy with anything over 2.6. Good luck with both your OC's guys, and tell me, where did you get your CPUs, just out of curiosity?
- Nov 9, 2004
- 11,465
- 1
- 0
Dadofamunky
Platinum Member
- Jan 4, 2005
- 2,184
- 0
- 0
I know, huh? The 3700+ was absolutely my limit. I couldn't see spending more.
ANOTHER REPORT: Well, edging closer to 2.8 but not there yet. Using a 133 divider at 2.78 GHz and Prime fails after 4-9 minutes. CPU voltage is now set to 1.550 with +10% - the max my mobo will go (see below). CPUZ reports core voltage at 1.616V. I've started twiddling with A64 Tweaker, which is an interesting utility. Adjustment of 133 divider to 7.8us refresh rate has not helped results - utility won't let me use a 180 or 150 divider, which I think is a limitation of my mobo. I'm ready to start dialing down my FSB and CPU voltage again before I retire for the night.... hey, I can't complain about 2.7 GHz on air....
Duvie's 3000+ is one sick chip....
ANOTHER REPORT: Well, edging closer to 2.8 but not there yet. Using a 133 divider at 2.78 GHz and Prime fails after 4-9 minutes. CPU voltage is now set to 1.550 with +10% - the max my mobo will go (see below). CPUZ reports core voltage at 1.616V. I've started twiddling with A64 Tweaker, which is an interesting utility. Adjustment of 133 divider to 7.8us refresh rate has not helped results - utility won't let me use a 180 or 150 divider, which I think is a limitation of my mobo. I'm ready to start dialing down my FSB and CPU voltage again before I retire for the night.... hey, I can't complain about 2.7 GHz on air....
Duvie's 3000+ is one sick chip....
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter