He's testing a new, unrealeased board, so I'm sure he removed that information to keep us from seeing it. He's got l33t photoch0p skillz.Quote:
Originally Posted by nugzo
LOL
Printable View
He's testing a new, unrealeased board, so I'm sure he removed that information to keep us from seeing it. He's got l33t photoch0p skillz.Quote:
Originally Posted by nugzo
LOL
ROFL... I checked the pic, unlikely it's ch0p'd ;) But no need to... Memtest doesn't recognize the new ATI chipset on a Halibut :p:Quote:
Originally Posted by dnottis
Board is a Halibut reference board as EMC2 mentioned. The fact that the memory can only give the bandwidth the on die memory controller lets it kind of eliminates the board as the cause of the high bandwidth.All A64 boards will give very close to the same bandwidth at the same timings etc as the memory is tied directly to the CPU.
Next up I have never used Photoshop on a bench screenie, i work for OCZ yes, but I also persue only the absolute truth when it comes to memory benches, i have never BS's anyone on a forum with memory benches and never will. Please check my posts, you will see OCZ thru myself and thru Ryan have always been completely open about our products and lay the info on the line.
I resent the fact that you would feel i photoshopped a memtest screenie :mad:
hehehe i think they were saying you "photochopped" the screenie so no one would know what mobo you're using. :)
Tony I don't think they were refering to the memory bandwidth, only the mb info and as pointed out by EMC2 it wouldn't be displayed anyway if Memtest didn't recongize the bios info.Quote:
Originally Posted by bigtoe
I can personally vouch for Tony, I've never seen him post anything that was less than truthfull or that was misleading. Just the oppisite I find that he goes out of his way to shed some insight into what is to come from OCZ and resolve issues with OCZ products.
Quote:
Originally Posted by situman
Tony - thats what I meant. That you photoshopped it to remove any board specs and chipset info, nothing more. I would NEVER accuse you of anything like that to show more bandwidth. Sorry for the confusion.
Yea, Tony is a standup guy, he's given me assistance with memory issues I've had, he's the best memory guru I will ever know - if you are reading this thread please understand I wasnt indicating he was photoshopping the results, I thought he removed the chipset and any motherboard information due to NDAs that we sometimes have to deal with.
Dean
Dean, my issue was not really with you but best to let it drop now.
The memtest issue is that X86 don't have the chipset info as yet so memtest does not record what the registers are setting, it does however record bandwidth correctly as i have seen the same with the NF4 also.
I actually feel interleaving may be working more efficiently which shows much better bandwidth as a result, i actually had over 3k with a Sandiego with 2gigs and on the same exact settings with a 1gig kit 2800+ was all i got so there looks to be something with running 2gigs over 2 dimms.
Ofcourse it may be just some OCZ magic in the SPD, i will have to ask later ;)
For those who wanted a windows sandra bench here ya go. 254fsb on the mem 3-4-4-5, weird thing is 3-4-3-5 gave the same results but was just not quite as stable...
Over 7k buffered at 254 is pretty efficient, around 3600 unbuffered is also not bad either.http://www.xtremesystems.org/forums/...id=36070&stc=1
Quote:
Originally Posted by bigtoe
So i guess your issue was with me. I NEVER once said anything about photoshop or cheating. Nor did i mean to imply it. I did Imply that maybe something was wrong with memtest b/c the info was missing. Check my other posts ive never once said anything bad about OCZ, I've always given nothing but praise! You know 2342563x more than me when it comes to memory. Its your job to know. If you or anybody else took it the wrong way i apologize, It wasn't meant that way.
I think this proves Tonys point a little better....Not saying it's stable yet, but 260 passed with flying colours. As Tony said, 3500+ bw is pretty nice with 2GB kits.
http://www.danghardware.com/40002GBDFIBW.jpg
This is my sisoft with ballistix tracers at 254fsb same mem timings and 704-2BT bios. memtest86 in bios shows 2815 bandwidth. I am thoroughly :confused: :confused: confused :confused: :confused: now. How on earth am i beating that 3500+ bandwidth on your memtest86 screenie.Quote:
Originally Posted by bigtoe
http://img371.imageshack.us/img371/7082/untitled6vc.jpg
Your sandra score is buffered, its dependent on L2 cache and processor speed. The memtest86 is unbuffered, just raw memory bandwidthQuote:
Originally Posted by nugzo
nugzo
I was running in memtest86, i posted a sandra glory benchmark for you and we are real close but you are running 11x254 and I was running 311x9 in 166 mode, try your board the same way and see what you get in sandra. Please run an unbuffered also as this is closer to real bandwidth than the bloated bus dependant buffered bench.
One point a few have failed to notice is that the 2gig kits seem to give more bandwidth than the 1gig kits....
Tony, have you tried running 4x1gig OCZ PC4000 on DFI NF4 ? I'd like to see some results with that as I'm concidering going with a second set for my work rig. I think it would benifit from as much memory as I can throw at it.
I have actually noticed this....care to explain why it is bigtoe?Quote:
Originally Posted by bigtoe
I don't trust Memtest bandwidth anymore, with my 3200+ Venice I would get 3300MB+ in Memtest at 10x250, the same exact settings with my FX-55 and 3700+ don't even get me 3000MB in Memtest, even at 11x250 its under 2900MB.Quote:
Originally Posted by nugzo
I haven't but Rgone has over at the DFI street forums, i will try and get him to post over here with some results.Quote:
Originally Posted by Major
I can tell you he really liked it.
Memtest is still cache dependant like most bandwidth benches, all you can do is run the same cpu and run benches back to back to make a comparison.Quote:
Originally Posted by HiJon89
I can tell you I have seen over3K easy with a sandiego using a 2gig kit, pull the 2gig out, put a 1gig in and 2700 was reported. There are gains to be made with 2gigs over 2dimms over a 2dimm 1gig kit thats for sure.
Its either due to the density or interleaving is working more efficiently, im still researching and will post when i get some answers.Quote:
Originally Posted by Vapor
Can this OCZ 500 gold do 2.5-3-3-6@230fsb?
Quote:
Originally Posted by Major
I had a feeling you would be back to DFI, I've been down that ABIT road before and it does suck :rolleyes: I think the new Grouper board will be a MUCH better option to play with.
pretty much yes, some may need 3-3-3, some are getting lucky and seeing 3-3-3 at 250+ alsoQuote:
Originally Posted by rodman
Well 3500+ isnt that much bandwidth in memtest. But the bigger the banks on the memory the more bandwidth you get, I couldnt break 3500MB/s with 2x256BH-5 evven at 270Mhz, but with 2x512MB UTT BH-5 I can get 3500 with just 3.43v, and 261Mhz. And this is at 2-2-2-6-7-12 as well. 1M stable too. :D
Anyway, my point is that bank interlaving gets more efficiant the more ram you have. So 2x1GB would be faster than 2x512 and that is mush quicker than 2x256. This may also be why 2x1GB cannot clock very well, and are extremly memcontroller dependant.
Yeah I agree with ya man.. it's good to see that some 2x1gb kits are ok OCers.Quote:
Originally Posted by mr_knowitall15
It can be taken up to 2.9 volts... I thought it was 2.8 like a lot of other OCZ RAM, but I had a look on their site, and its 2.9...
http://www.ocztechnology.com/product...l_channel-2gb_
Later,
Yan