Little bit confused...here is what I have done...set memory to ddr333 (166) change clock to 236...236+166=402...I was under the impression that with that setting my memory would be at 200mhz or 202mhx but cpu-z says it is at 185....what I am I not getting here?
Here is what I posted in another place. I believe it explains these calculations, although it was written in response to a question about the relative overclocking potential of the 3200+ versus the 3000+. The example calculations were written with the official Bioses for Asus nforce4 boards in mind; other boards or Bioses might offer more fine-grained control over memory frequency by providing, for example, an FSB:HTT ratio of 366/400 (11/12).
Suppose you are targeting a CPU frequency of 2500 MHz. Meanwhile, you have determined that your memory is only capable of running at, say, 220 MHz (DDR440) at timings that you regard as acceptable. If you have a 3200+, you could run at 10x 250 (=2500 MHz), with a memory setting of 166 (DDR333). This would give you a real memory frequency of 208 (DDR 416) * (see below). On the other hand, if you have only a 3000+, you would have to run at 9x 278 (=2500 MHz) with memory at DDR266, giving you a real memory frequency of only 179 MHz (DDR358). So, even though, as I have said, the Athlon64 is not really sensitive to memory bandwidth, it is possible you could benchmark a small difference in performance between these two set-ups.
* Calculations: an example
1. Take you CPU multiplier, for example 9.
2. Find your memory ratio (aka FSB:HTT ratio): @ DDR266, it is 266/400 (=2/3)
3. Obtain your memory divider using the numbers from steps 1 and 2:
Divide the CPU multiplier by the memory ratio: 9 / (2/3) = 13.5
Non-integer values are rounded up. So 13.5 gives you an actual memory divider of 14.
Even if the calculated value had been 13.01, the actual divider used would still be 14.
4. Discover your real memory frequency:
Divide the CPU frequency by the memory divider: 2500 / 14 = 178.57 MHz
If you had tried the memory at DDR333:
333/400 = 5/6, so 9 / (5/6) = 10.8. Round this to 11.
2500/11 = 227.27 MHz
But this exceeds your empirically established limit of 220 MHz so this configuration would not work.
Now, with a 3200+ you could use a 10x multiplier: 10 / (5/6) = 12 exactly.
2500 / 12 = 208.33
So you would be within your memory's capabilities and at a higher frequency than possible with a 3000+.
Note you should always run the memory at 1T command rate (sometimes known as command-per-clock), regardless of your other memory timings (Cas latency etc.) or your memory frequency. However, be aware that the memory controller on the Athlon64 is not capable of running at 1T if you are using four DIMMs. The forthcoming core revision known as Venice can operate with four *single-sided* DDR400 SDRAM modules at 1T.
Suppose you are targeting a CPU frequency of 2500 MHz. Meanwhile, you have determined that your memory is only capable of running at, say, 220 MHz (DDR440) at timings that you regard as acceptable. If you have a 3200+, you could run at 10x 250 (=2500 MHz), with a memory setting of 166 (DDR333). This would give you a real memory frequency of 208 (DDR 416) * (see below). On the other hand, if you have only a 3000+, you would have to run at 9x 278 (=2500 MHz) with memory at DDR266, giving you a real memory frequency of only 179 MHz (DDR358). So, even though, as I have said, the Athlon64 is not really sensitive to memory bandwidth, it is possible you could benchmark a small difference in performance between these two set-ups.
* Calculations: an example
1. Take you CPU multiplier, for example 9.
2. Find your memory ratio (aka FSB:HTT ratio): @ DDR266, it is 266/400 (=2/3)
3. Obtain your memory divider using the numbers from steps 1 and 2:
Divide the CPU multiplier by the memory ratio: 9 / (2/3) = 13.5
Non-integer values are rounded up. So 13.5 gives you an actual memory divider of 14.
Even if the calculated value had been 13.01, the actual divider used would still be 14.
4. Discover your real memory frequency:
Divide the CPU frequency by the memory divider: 2500 / 14 = 178.57 MHz
If you had tried the memory at DDR333:
333/400 = 5/6, so 9 / (5/6) = 10.8. Round this to 11.
2500/11 = 227.27 MHz
But this exceeds your empirically established limit of 220 MHz so this configuration would not work.
Now, with a 3200+ you could use a 10x multiplier: 10 / (5/6) = 12 exactly.
2500 / 12 = 208.33
So you would be within your memory's capabilities and at a higher frequency than possible with a 3000+.
Note you should always run the memory at 1T command rate (sometimes known as command-per-clock), regardless of your other memory timings (Cas latency etc.) or your memory frequency. However, be aware that the memory controller on the Athlon64 is not capable of running at 1T if you are using four DIMMs. The forthcoming core revision known as Venice can operate with four *single-sided* DDR400 SDRAM modules at 1T.
Originally posted by: GuitarDaddy
Originally posted by: Tiorapatea
Look, I know some people are not satisfied with the 1T limitations of the Asus but since most of the performance on this platform comes from CPU clock speed, some additional performance comes from low memory latency, and almost none comes from bandwidth, why not just sell off your high-bandwidth memory and buy some low-latency memory? Supposing you have a decent chip, you could run 9x289 @ DDR333 with timings as low as you can go. This would give you 2.6 GHz with memory at a real frequency of 236 (DDR472) at 1T. Be happy!
If for some reason you *must* have higher numbers because of an obsession that has little to do with real-World performance, then by all means sell the board and go DFI but at least acknowledge why you are doing that.
I only make this point because I feel that a lot of people may be put off the Asus by all this talk of disastrous overclocking potential when the reality is that it provides ample scope for overclocking.
I am crossing my fingers at this point because I have just ordered one of these boards. I reserve the right to hate this board in future and yes, I will come to you guys for help
I couldn't agree more. I've had this board since December and it is a fantastic board and I wouldn't trade it for any of the other NF4 boards. I have been running at 2.6ghz 24/7 since the first week I built it with my 3200+ using exactly the settings you described.
I would like to see the 250 1T issue fixed as much as anybody, if for nothing else just so this board gets the recognition it deserves as an excellent overclocker. But as you stated the real performance of increased bandwidth on A64's is zero.
And if you want/need additional vcore (extreme O/C ing with phase change/highend water cooling) do a vmod on the board (OP Painter at Extreme gives a good description of how to do this on the A8N-SLI) or if your into programing, get the pheonix bios editor and create your own modded bios. If you do of course you must share it with me
And Tiorapatea feel free to PM me any time, I will gladly share any tricks I've learned about this board
I agree 100% with your assertions about the memory bandwidth and that it's best to run the memory with a divider so it run with lower latencies. For that reason, the 1T/250 mhz limitation is a non-issue. But the Vcore is an issue, as I understand it the board is capable of supplying higher voltage to the CPU, but the BIOS seems reluctant to allow it, even on the versions which give the option to set higher voltage. I feel that the voltage issue is a limiting factor, as I can only get 1.56 Vcore, and I'm sure my CPU can reach 100-200 mhz higher with 1.65 V.
Fair enough, this is a bigger issue. Personally, I would prefer not to go above 1.55V because of heat, noise and component life but for some people this is an annoying restriction.
GuitarDaddy
Lifer
- Nov 9, 2004
- 11,465
- 1
- 0
I agree the voltage is an issue for serious overclocking. But I should qualify my earlier statements that the 1.56-1.58 limit seems to be only with Winchesters. I beleive Newcastles, Clawhammers, and Sledgehammers can successfully run over 1.6 with this board. And with the Winchester default vcore off 1.4 this board provides an adequate but not optimal vcore range for air cooling.
Just upgrade to the DFI. I did.
Best swap I could have made. This board is rock solid and has all the overclocking options I'll ever need.
I agree, I should have waited.. but I made up for it in the end!
I still have my Asus sitting here, I coudl go back if i wanted.. but guess what one I'm sticking with?
The one that forces your primary graphics card to be positioned directly above the nforce4 chipset? No thanks.
How about The One that forces you to not use over 245fsb@1T? No thanks.
The One that has no vcore over 1.55v with anything but shipping bios? No thanks.
The One that AT said was blown away by the DFI? No thanks.
garikfox
Senior member
- Sep 1, 2004
- 508
- 0
- 71
Lock up PEEPS I might have a clue for U !
"My DFI just Bombed after 1.46 hrs of EQII i was putting her to the final test, BUT she failed, After 2hrs or so of EQII it gave me a The famous little window tellin me this module etc failed... SO I am back to my ASUS which worked on 10% OC this DFI wouldnt even load 10% on a game at all it failed. So my Conclusion is Its The BIOS that doesnt excatly accpet my mem."
When I had My DFI I Flashed a p BIOS means Performance. When i did that and came back into windows seemed i had some weird video stuff going on. EG.
Once XP is Loaded under the "current BIOS" then U load up SLI drivers etc..
NOW heres the catch Freezing guyz, Cuz I just freezed also, lol
Once U have a Loaded XP etc. with a BIOS and IF u FLASH OR CHANGE the PEG etc. then U WILL have a LOCK UP ... I just did like my main poster here.
So the conclusion is Did u Flash or change BIOS setting WHILE yer vid drivers were loaded ?
YES u have to do a CLEAN intall to fix the FREEZING, Seems the Newer BIOS's have major PCIE code upgrades. That Conflict with just like APIC or whatever.
So If u Have Lock-UP's in SLI once a game LOADS then this is the KEY.
BTW:1007-004 will NOT load a clean isnstall of XP with RAID0 as main. nice bug it has.
"My DFI just Bombed after 1.46 hrs of EQII i was putting her to the final test, BUT she failed, After 2hrs or so of EQII it gave me a The famous little window tellin me this module etc failed... SO I am back to my ASUS which worked on 10% OC this DFI wouldnt even load 10% on a game at all it failed. So my Conclusion is Its The BIOS that doesnt excatly accpet my mem."
When I had My DFI I Flashed a p BIOS means Performance. When i did that and came back into windows seemed i had some weird video stuff going on. EG.
Once XP is Loaded under the "current BIOS" then U load up SLI drivers etc..
NOW heres the catch Freezing guyz, Cuz I just freezed also, lol
Once U have a Loaded XP etc. with a BIOS and IF u FLASH OR CHANGE the PEG etc. then U WILL have a LOCK UP ... I just did like my main poster here.
So the conclusion is Did u Flash or change BIOS setting WHILE yer vid drivers were loaded ?
YES u have to do a CLEAN intall to fix the FREEZING, Seems the Newer BIOS's have major PCIE code upgrades. That Conflict with just like APIC or whatever.
So If u Have Lock-UP's in SLI once a game LOADS then this is the KEY.
BTW:1007-004 will NOT load a clean isnstall of XP with RAID0 as main. nice bug it has.
Originally posted by: garikfox
Lock up PEEPS I might have a clue for U !
"My DFI just Bombed after 1.46 hrs of EQII i was putting her to the final test, BUT she failed, After 2hrs or so of EQII it gave me a The famous little window tellin me this module etc failed... SO I am back to my ASUS which worked on 10% OC this DFI wouldnt even load 10% on a game at all it failed. So my Conclusion is Its The BIOS that doesnt excatly accpet my mem."
When I had My DFI I Flashed a p BIOS means Performance. When i did that and came back into windows seemed i had some weird video stuff going on. EG.
Once XP is Loaded under the "current BIOS" then U load up SLI drivers etc..
NOW heres the catch Freezing guyz, Cuz I just freezed also, lol
Once U have a Loaded XP etc. with a BIOS and IF u FLASH OR CHANGE the PEG etc. then U WILL have a LOCK UP ... I just did like my main poster here.
So the conclusion is Did u Flash or change BIOS setting WHILE yer vid drivers were loaded ?
YES u have to do a CLEAN intall to fix the FREEZING, Seems the Newer BIOS's have major PCIE code upgrades. That Conflict with just like APIC or whatever.
So If u Have Lock-UP's in SLI once a game LOADS then this is the KEY.
BTW:1007-004 will NOT load a clean isnstall of XP with RAID0 as main. nice bug it has.
BTW:1007-004 will NOT load a clean isnstall of XP with RAID0 as main. nice bug it has.
[/quote]what do you mean I am running raid0 on my raptors just fine...with clean install
These are BETA BIOS, no? What's with the drama?
Please the this MB vs. that MB arguments to your own rant threads please, some of us are looking for information here.
Please the this MB vs. that MB arguments to your own rant threads please, some of us are looking for information here.
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 Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter