I keep reading that a64's don't like half multiplier's but my system doesn't seem to have a problem with it and I need it to balance out my max stable overclock. What exactly is the problem they have when using half multi's?
AMD64's and half multi's
- Thread starter Koyanisquatsi
- Start date
the problem isn't with running 1/2 mults... it's what happens with your mem with 1/2 mults...
seems as though your ram runs at a lower speed than what is actually set at... kind of like using a divider
seems as though your ram runs at a lower speed than what is actually set at... kind of like using a divider
I also noticed thast my sandra scores seem to jive for the speed I roughly calculated...so I am not sure if they were off or not....some programs like my corecenter did not even recognize 1/2 multiplpier and always said I was runing the round up to the next whole number...cpu_z got it right though...
I was at a loss and asked about this awhile back...
i'm getting 6700 on sandra bandwith tests. i'd say that's about right. both cpuz and uguru are showing correct frequencies.
is he saying that since im running a half multiplier that my ram isn't really running at the speed cpuz states it is?
SunSamurai
Diamond Member
- Jan 16, 2005
- 3,914
- 0
- 0
This is a silly myth people. Its akin to saying anything with a 6X multiplier will summon the devil.
i understand what you're saying... sandra... cpu-z will tell you that you're not running a divider but the 1/2 mults don't make your ram run at lets say 1:1
let's say you use the 8.5 mult and 1:1 (200 mhz or htt)
so you're speed is 1700 mhz but your mem is only running at 188.89 htt, not 200
please check this thread out at xtreme (1st post)... i'm not on crack... i promise... and oskar wu is a very very very very credible source
here
well i've run sandra bandwith benchmark with my ram at same speed using a whole multi and a half multi and the scores are near identical so..........
Also according to that link you posted the memory controller rounds up to the higher number when using a half multi so it would be increasing the memory speed not decreasing it.
Correct
The "problem" is that some A64's will not run stable when using a .5 increments ...like 10.5, 9.5, etc. when tryingto oc. That is the only "problem".
take this example...
13 * 200 / 166 (divider) = 15.67 --> rounded up is 16
2600/16 = 162.5 mhz
so when you think your ram is running at 166 htt, it's actually running at 162.5
so it's slower than what you think... make sense?
i had to think about it for a second as well
GuitarDaddy
Lifer
- Nov 9, 2004
- 11,465
- 1
- 0
On NF$ board like my A8N-SLI, this is very easy to see using ClockgenNF4. Clockgen shows that you are actually using the next higher whole multi. And it shows your HTT and ram speed correctly. CPUZ shows the .5 multi and shows HTT and ram at the higher fiqure, CPUZ just shows what your bios is reporting.
Toms Hardware did an extensive article on this a couple of months ago, and even quoted AMD sources confirming that there are no 1/2 multi's on A64 chips, its all smoke and mirrors implemented by the mobo makers.
And by the way, you have to use an Intel chip to summon the devil, satan likes HT
Toms Hardware did an extensive article on this a couple of months ago, and even quoted AMD sources confirming that there are no 1/2 multi's on A64 chips, its all smoke and mirrors implemented by the mobo makers.
And by the way, you have to use an Intel chip to summon the devil, satan likes HT
Also, from what I understand, there are no 1/2 multipliers on A64s per se. Motherboard manufactures can choose to make it an option (many do) to be used at your own risk.
See LINK. (Funny what you can find with a google search)
Now this may just apply to MSI boards, as other manufacturers may have different implementations. But you should at least be aware that:
a) YMMV and you use 1/2 multipliers at your own risk,
b) in at least some cases, while perhaps not qualitatively (i.e. in benchmarks) noticable, a 1/2 multiplier can decrease your memory performance and stability, and
c)1/2 multipliers are achieved not through an actual A64 multiplier setting, but through a motherboard's adjusting of dividers, frequencies, etc.
See LINK. (Funny what you can find with a google search)
Now this may just apply to MSI boards, as other manufacturers may have different implementations. But you should at least be aware that:
a) YMMV and you use 1/2 multipliers at your own risk,
b) in at least some cases, while perhaps not qualitatively (i.e. in benchmarks) noticable, a 1/2 multiplier can decrease your memory performance and stability, and
c)1/2 multipliers are achieved not through an actual A64 multiplier setting, but through a motherboard's adjusting of dividers, frequencies, etc.
Okay, I'm having trouble getting my head around this one. I've read what Oskar wrote and all that's been talked about half multi's. However, I need someone to explain this to me...
I cannot even POST at:
310 FSB
8x multi
3x HT
120 Mem divider
@1.65v
BUT I can Prime for hours with:
333 FSB
7.5x multi
3x HT
120 Mem divider
@1.55v
I did not change any other options. I'm sure there's an explanation. Am I not truly running at 2500mhz right now? CPU-Z says so, but that may not mean squat.
I cannot even POST at:
310 FSB
8x multi
3x HT
120 Mem divider
@1.65v
BUT I can Prime for hours with:
333 FSB
7.5x multi
3x HT
120 Mem divider
@1.55v
I did not change any other options. I'm sure there's an explanation. Am I not truly running at 2500mhz right now? CPU-Z says so, but that may not mean squat.
What total clockspeed is ClockgenNF4 reporting your system running at? That will be the first hint as to what out of those settings is not as it should be.
It's gotta be the memory, although I'm not familiar with A64 enough to know why. My guess would be that the divider is wrong somehow and you're ending up much higher on your memory speeds than you think you are.
I really don't know how actual mem speed works with half multis. But normally, I would say that at 333 FSB with a 120 divider, my actual mem is 200mhz (333 * 3 / 5 ).
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: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter