Printable View
I'm reasonably happy with the results anyway. I'm going to keep pushing, but if you look at the ballistix, people are averaging about 270-280 at 3-3-3-8. This ram is doing 245 at 2.5-3-2-6, so thats 25-35mhz slower, but with better timings, so its probably what, 10-15 mhz slower? And it costs almost half as much.
I just plucked in my 2x 1GB Patriot PC3200LLK 2-3-2-5 kit into my rig, seems for first 12hrs i've played with this ram i've gotten it to the about same limit as AndrewK at i4memory.com ~245mhz 2.5-3-2-5 at 2.69v vdimm seemed disappointing compared to crucial ballistixs at first but then the memtest bandwidth is very nice for san diego core cpu due to tighter timings since san diego cpus clock for clock show 350-450MB/s less bandwidth in memtest than venice/winchester/clawhammer cpus
http://www.fileshosts.com/memory/pat...5325/test5.jpg
http://www.fileshosts.com/memory/pat...5325/test8.jpg
250mhz is no go right now but will see heh... Gaspar's ram is a rare gem :D
not bad
http://www.fileshosts.com/DFI/NF4_SL...rpi-32m_tn.jpg
http://www.fileshosts.com/DFI/NF4_SL...15_6768_tn.jpg
There's a BGA-version of both BE- and CE- in production. I wonder where do they go?
Part number's are identical except for the letter between die revision (B/C) and speed grade (-5/-6). The BGA chips end in BF- and CF-.
Theoretically BGAs should be a bit better than TSOPs.
Anyone seen Infineon DDR1-BGA modules floating around anywhere?
Im also running these modules but my clawhammers mem controller craps out at 234X11 anything more than 2.5 volts on the chips causes many errors. Nice gains in sandra at tref 2048 and 3684 tho.
probably DFI NF3 250Gb limiting you
some more benchmarks
2x 1GB Patriot PC3200LLK with Brainpower B6815 PCB
@12x 245mhz 2.5-3-2-5 7-18-2222 at 2.69v
VID = 1.45v
VID Special = AUTO
chipset = 1.6v
LDT = 1.3v
LDT 4x
CPC = Enabled
CAS Latency(CL) = 2.5
RAS to CAS(Trcd) = 3
Min RAS# Active time(Tras) = 5
Row Precharge Time(Trp) = 2
Row Cycle Time(Trc) = 7
Row Refresh Cycle Time(Trfc) = 18
Row to Row Delay(Trrd) = 2
Write Recovery Time(Twr) = 2
Write to Read Delay(Twtr) = 2
Read to Write Delay(Trwt) = 2
Refresh Period (Tref) = 4708
DRAM Bank Interleave = Enabled
Skew = AUTO
Skew Value = 0
DRAM Drive Strength = AUTO
DRAM Data Drive Strength = AUTO
Max Async Latency = 7ns
DDRAM Response Time = Fast
Read Preamble Time = 5.5ns
Idle Cycle Limit = 16
Dynamic Counter = Enabled
R/W Queue Bypass = 16x
Bypass Max = 7x
32 byte Granularity = Disable(4burst)
Super Pi
1M = 28.312s
8M = 5min 12.828s
32M = 25m 02.203s
Pcmark04
Score = 5,815
Memory = 6,768
Pcmark05
Score = 5,776
Memory = 5,213
Everest Home 2.20.405
Read = 7,033
Write = 2836
Latency = 38.5ns
Sandra 2005 Lite SR2a
Buffered = 7,260 / 7,191
Unbuffered = 3,418 / 3,586
i was under the impression pretty much no 2gb kit could do 2-2-2-X because of the strain on the memcontroller?
well it ain't 2-2-2- it's 2.5-3-2-5 :)
High memory density and low timings do not exclude each other.Quote:
Originally Posted by Vincentvega18
The problem is that atm 512MBit chips simply can't do 2-2-2-x at speeds high enough. Remember that not all 256MBit chip do 2-2-2-x, even at stock.
A memory controller doesn't really "care" wheather they are 2x1GB or 2x512MB dimms that are slapped to the dimm slots. All it cares is the amount of ranks a dimm has. "All" double sided 1GB ddr-dimms are 2 ranked just like all double sided 512MB ddr-dimms are 2 ranked. That's why you can even run 4x512MB - for example Mushkin Redline - at DDR500 2-2-2-x 2T.
After all, bigger memory space is only more options for the controller to place a single bit. ;)
Im still testing my kit for the max..
So far I have whats in my sig 10 hours prime stable..
Have any of you tried a 4x cpu multi? Mine do 280@2.7 volts2.5-3-3-6 rest on auto. Using a higher multi is when things get tricky :p:
Some nice results guys. I am considering the patriot for my dothan build. The lowest price and cl 2.5 performance. I wonder what they would do on Intel?
I got some patriot 3200llk 2gig kits yesterday, its got be-6 ic's
so far im stable at 500mhz but like andrew said, antmore and its a no go.
http://home.comcast.net/~thinkblaze/2gig_500_stable.jpg
Could be CE-6 you've got there (if you haven't already popped the hood to peek)....anyway, those are some good results for not-uber expensive 2GB kits, good job :D
I check under the heat spreader already, there deffinitly be-6's
VERY good results for later BE-6 then :toast:
weid thing is they top out at 250, any higher and its bsod. I just primed for 6 hours with no errors but at 251 its bsod. its got to be a certain timing stopping it.
after a little burn in I was able to get a little more out of them
http://home.comcast.net/~thinkblaze/...ig_3.05ghz.JPG
Babyelf bro....
i came across HYPERX PC3200AK2 which uses the CE-5 chips....i'm kinda dissapointed which is want was CH or BH series chips....
Now that i found this thread its like being lost at sea and when it just seems to fade away...a mountain from an island nearby shows its peak...
could someone post some benches for 2.5-3-2-x vs. 3-3-3-x, cos i'd hardly expect any differences. or did i mis something on this topic?
(i'd test it myself but i'm still waiting for the psu :( )