Announcement

**Tradesman** · 06-04-2009, 08:11 PM

Good post, makes for good referneces when people come looking for help, I'll be filing it away for future use.

**genetix** · 06-07-2009, 02:15 AM

Yeah, some of the values above still needs final stabilizers. Like 425-437FSB -> 1020-1041Mhz. The 2.1v probably is enough and the setup above passes memtest clean up several times.

however, for example running clean-up test something like full RAM Linpack might arise instability not too sure is it RAM of the board the CPU is stable that I am sure of.

Other style of RAM testing:
Anyway, my advice to everyone who tests RAM would actually be to use compression softwares to do so. Something like 7-zip latest versions in over 4GB cases like this x64 (and looking for their forums for latest versions v9.04 would be ideal because of new LMZA2 compressions multi-threads). 7-Zip style of compressions with dictionaly sizes around 256MB uses around 8GB of RAM if that RAM is not clean you will surely see system drop. (Watch for crossing over 8GB Windows OSes has since begin on time have serious bugs when RAM runs out they freeze even while there's paging).

**genetix** · 06-30-2009, 02:33 PM

Well, found something really curious.

It's not exactly 6400 or even 8800 PI-B chips, but this is actually pretty damn unreal.
Here's the URL: http://www.xtremesystems.org/forums/...d.php?t=194916
The dude is pushing the F2-9600CL5D-4GBPI absolute insanity level. The voltages he has in his screen are like 2.200v on lowest and ~2.30-2.32v actual volts on box. The the facinating part is this fellow was able to push the chip under 4-4-3-15 to 1008Mhz...

I mean seriously I have to give applauds to that one! Although, meant to ask G.Skill on Max 24/7 voltages on these chips since the voltages used on those are straight around 2.30-2.32v where he/she is running while the actual performance jump as I explained above on first post is actually at 2.28v, but considering he is doing -1t CLs on everything plus it's quite something. O and btw, this same fellow, if you scroll the bottom just beat my record (however on only 2 sticks). He pumped the Absolute Maximum speed at 5-5-5-15 under the performance level 6 to 1333Mhz. Lol, only prob is he is using 2,46 volts on that, hehehe, which is like crazy. Plus his NB is like 1.6v FSB is 1.6v which is for one chip killer. and for Northbridge consideration that system absolutely has to have like water cooling or something (even while yes I know that board on OC + Auto will pump the NB to 1.65v and still stay below 47-55 degrees).

Absolutely amazing result.

**GSKILL TECH** · 07-02-2009, 03:22 PM

Definitely good testing and write up. Hopefully this will help those that are interested in advanced memory timing.

GSKILL SUPPORT

**genetix** · 07-07-2009, 12:01 PM

Wanted to ask, on G.Skill many things above and you give me 'good testing'. Why don't you guys there help me testing while you got much more resources than I do. I think, if you guys actually took latency calculations in hands and actually provided some support based of actual low limits to see errors on these forums people have & solve errors by them would help me and the others to actually get functional systems.

I know calculations are not always correct and sometimes or they are lower than expected while still functional however even your EPP specs on DDR2 are faulty in calculations which is why there's so much people asking why they don't work. You could also tell people the truth on speeds of memory while at it. Not speaking of XMP spec speeds, but actual realistic results of same blocks and how they act on low latencies against the higher speed so people doesn't think faulty ideas that faster the memory would by anyway be faster on their boards.

**GSKILL TECH** · 07-07-2009, 01:43 PM

If I could sit all day testing limits, I would, but that is not my job. Advanced memory timings are exactly that. The reason why we do not offer Advanced timings is because the average customer simply wants to plug and play. As you know, different hardware combinations also generate different values as well, which is why we can only offer basic memory timings. I have had cases where we tested 5 of the same exact motherboards, and one would be plug and play, three would need slight adjustments, and one would simply not run speed at all. But point is that Advanced settings are left to those that are capable of configuring those settings, while the vast majority simply don't care. Although we could provide technical data, it does not automatically solve issues.

That's why I stated it was good testing, because most people wouldn't even know what they are reading, but your testing gives good insight to your combo.

GSKILL SUPPORT

**genetix** · 07-07-2009, 02:36 PM

Originally posted by GSKILL TECH View Post

If I could sit all day testing limits, I would, but that is not my job. Advanced memory timings are exactly that. The reason why we do not offer Advanced timings is because the average customer simply wants to plug and play. As you know, different hardware combinations also generate different values as well, which is why we can only offer basic memory timings. I have had cases where we tested 5 of the same exact motherboards, and one would be plug and play, three would need slight adjustments, and one would simply not run speed at all. But point is that Advanced settings are left to those that are capable of configuring those settings, while the vast majority simply don't care. Although we could provide technical data, it does not automatically solve issues.

That's why I stated it was good testing, because most people wouldn't even know what they are reading, but your testing gives good insight to your combo.

GSKILL SUPPORT

This is incorrect information and you know it.

When issue arise you should be the person who is capable of tell the person exact advanced timing calculation of specified memory this does not need testing nor any kinda experience. You do not need to test anything all are logical calculations of what the latencies are possible to be launched inside the block. All needed spec is the minimum spec of the technology & that gives an sight over what is possible and what is not. So, all you really need is a single set of blocks to determ how it can function in full controller & as an single. Any optimized profiles like EPP are overclocked specs can also be calculated by speed accordingly & base latency from speed factor. So when people say they cannot run something this means she or he has some value below the spec. Advanced timing chapter above is pretty perfect example of this.

Providing technical data should be exactly what should be done in situation. Person not understanding it is another aspect people are Plug-n-Play culture these days, but that doesn't say even I had to dig from the memories themselfs they are not on forum and I doubt I even get answer from support for the actual correct spec of the memory. Also should be mentioned that your EPP/XMP is below actual minimal specs of each timing at least somewhere so people could calculate their higher speeds correctly from begin with or see where's the problem when EPP/XMP doesn't work.

(Reference to an Random factor for some board works some doesn't. I do not believe this. I admit I've believe since begin of PCs that boxes has their own soul, but seriously everything has explainatory, why it's not working there is no random X38 which doesn't work on same manufacturer X38.

I also have seen exact same setups acting differently, but I have never seen an exact clone of system which would be acting differently and I wouldn't of find what is causing the difference.)

**GSKILL TECH** · 07-07-2009, 06:02 PM

I completely agree but there are too many factors involved when troubleshooting. What I mean by that is even if we have minimum timing specifications for each memory, it still would not solve all issues. Sure, it would be much easier to have some sort of base for advanced users, but like I said, it's not my job, so I can't test each combo for minimum values. This is where forums come in handy, where customers can test things out and share with everybody. So in the future, if someone wants to push it to the limit with the same combo, they can refer to the forum and see what the settings were for someone that has completed the same goal before. That's why I said your testing was good information because the PI is a common gaming and overclocking memory and someone else will definitely benefit from it.

Parts are all mass produced, which explains the differences in results. I have many i7 cases recently, and I used the same exact settings in BIOS and some worked, some didn't. Same with AM2+/AM3, some work immediately, some just don't want to work at all. Even when using the lowest spec, it will have issues, when the other one could handle standard easily.

I'm not here to argue or ignore anyone, I'm simply here to help with what I can and give input to what I see. And if I'm not here I'm responding to emails or phone calls. Otherwise if I had a choice I would just do testing all day and overclock everything for fun so I can post on the forum. =)

GSKILL SUPPORT

**Tradesman** · 07-07-2009, 08:03 PM

I agree with Genetix, when issues come up, the purpose of SUPPORT is find and provide the answers, if you can't solve the problem then I would expect that a refund or a trade up exchange would be in order, rather than just blowing people off or 'SUGGESTING' they buy a higher priced model (with no guarantee of it working).

Further, in another post today you stated the forums are just a place for users to talk and chat - since it's a SUPPORT forum (listed under support in your new WEB design), that's what it should be. Also, in a typical technical business plan, these forums are considered an ideal place to publish the results of any and all testing you perform....after all.........IT'S TO GSkill's BENEFIT!!!

As a single example, if you were to publish all the results of testing on the P6T board that you have done, just think of how many email requests for help that might be eliminated, I can't remember how many posts I've seen about that mobo and a wide range of your DDR3 RAM....on the other hand it might hurt sales since it hasn't appeared that you all have had a lot of luck w/ your higher end RAM on the P6T, but I've had some success as had Genetix.....and, the more you publish, the less duplicate requests you will receive (hopefully), then the more time you'll have to provide more in-depth support to others. In short you need a plan..

I've written/designed support policies/programs for a number of entities from small business through Federal agencies over the past 30 years and GSkill (no offense) as a company doesn't seem to have any plans, i.e. you released the MLC SSDs and after buying two of them I couldn't find anyone in your company who had any idea of how they work, there positives/negatives, tuning them, ...anything (and since I've now been accused of bashing GSkill, I'd be happy to offer full documentation and to what I just stated.

Tman

**genetix** · 07-07-2009, 08:43 PM

Originally posted by GSKILL TECH

if we have minimum timing specifications for each memory, it still would not solve all issues.

but you do have minimum and tweaked calculations inside the first 256KB of memory bank stored in chip as an EPP/XMP & minimal capabilities by RAM. Viewable by todays hardware monitoring softwares. This information can be used to calculate any spec the memory should have to be able to POST or to error free operation, if this fails in any scenario memory should be automatically RMA in.

Originally posted by GSKILL TECH

"Sure, it would be much easier to have some sort of base for advanced users"
"but like I said, it's not my job, so I can't test each combo for minimum values."

This should be standard procedure inside an documented manual to memory not something what either forum nor support should should solve or add online separately/user uniquely.

Actually, as far I am concerned you should be provided the information by the people who does test the memory by the last inc. there is and this information should be publicly available also through your pages instead of memory spec card.

Originally posted by GSKILL TECH

I have many i7 cases recently, and I used the same exact settings in BIOS and some worked, some didn't.

I didn't say 'same settings in BIOS' wouldn't have issues. I said identical systems or systems which are functioning in same spec from same manufacturer.

Originally posted by GSKILL TECH

I'm not here to argue or ignore anyone, I'm simply here to help with what I can and give input to what I see. And if I'm not here I'm responding to emails or phone calls. Otherwise if I had a choice I would just do testing all day and overclock everything for fun so I can post on the forum. =)

I don't think you are ignoring anything and main argument was raised by me since support for to get spec has been like 'not-there' for a long time and now all the suddenly there's answers even on posts which doesn't need answer, but no real answers other than 'replications', but that's still hell a lot better than no answers at all.

Originally posted by Tradesman

if you were to publish all the results of testing on the P6T board that you have done, just think of how many email requests for help that might be eliminated

Probably pretty minor affect, but you are correct on in a sense that on 'popular' board adding functional specifications raises sales & helps many with same board.

Better however would be to know different board manufacturer standard ideas of voltages & defaults(POST Automatic settings) by chipsets which G.skill does have. Base the chipset + specs / manufacturers and work away to get that to sticky. This strategy would IF developed all the time help even more I think.

Originally posted by Tradesman

but I've had some success as had Genetix.

I never said I had anykind of success with 'PI-B' memories. I just stated on above what this memory is capable and what it's not. Where it will fail & what it needs to actually get to spec, but as of yet I have not got the spec working as reads on package without voltage increase to unbearable levels.

Wouldn't exactly call that success. However, I generally like the style how these blocks works and how high they can go. So ain't really an problem for me.

**genetix** · 07-10-2009, 08:25 PM

Well, updated the first message. Corrected timings to correct topics and included new chapter 'DRAM Static Read Control' at ASUS boards.

Edit #01:
I am having insane issues on READ speeds on this thing. I mean Write and Copy speeds are sure faster always than read speeds. I assume this is G.Skill fault because I read the JEDEC standart profile & the G.Skill profiles out from memory 256K profile bank and it seems only XMP/EPP profile keeps the drive strenght at 1.00x while all other setups drops the drive strenght to 0.75x.

It's still funny that the memory XMP profile active the Read and write are the same while any manual settings (even the same damn 4-4-4-12 on PC2-6400) and read speed is around 800MB/s from write speeds. (this might be board specific but I tested this on 2 manufacturer boards & on 3 different chipsets.)

Edit #02:
I added CL5-5-4-17 Profile which seems to kick out any other profile straight up with DRAM Static Read Control enabled to get native tRD values. Seems this one is gonna be the best of them all. However, this one will need a bit more kick on voltages on full 8GB controller than I was planning to not too much, but more than spec.

..

If someone has some other specific FSB to do good timings against you can ask for them. I will need to know your board, CPU, your memory model(and amount of memory modules), FSB you are aiming at & System you are gonna run in it.

**genetix** · 07-17-2009, 10:20 PM

I'll post two timing specs here, which both are really questionable in voltage and in functionality terms. Also Voltage functional spec for get proper current. Windows OS systems will be largest issues. As when memory voltages are stable at below system this does not mean the memory voltage will be stable at Windows Operating system using all components. This means +0.04-0.06v increase in voltage to get them stable, however, also north bridge voltages in this case might come in question and this will mix entire set, if so.

Voltages:
1.882v-1.972v, 2.112v-2.172v, 2.248v-2.288v, 2.352v, 2.448v
NOTE: All Voltages mentioned on this page are ACTUAL voltages from memory controller NOT something you add in BIOS. You can see these through OS as above.

This chain of voltages each has levels of bandwidth to function. Voltages missed between was neither for any use nor stable in any condition unless less voltages needed for original. The last on is specific for ASUS overclockers this value actually is their voltage for 'Auto' board voltages while RAM is at *1200Mhz*.

1:1 Memory array (333Mhz Strap):
CL: 4
RCD: 4
RP: 4
Ras: 14
RRD: 2
RFC: 42-45
WR: 4
RPT: 2

RTW: 7
WTR: 2,4
RTR: 4,5
WTW: 4,5

WTP: 9
RTP: 2
PTP: 1
PRE to Active/REF: 3

Static Read: Disabled
'Clock Twister': Strongest you can add it.
Performance Level: 7

Voltage: 1.882v-1.972v
Minimum North Bridge Voltage: 1.488v

This combination speeds at 909Mhz (454FSB):
Read: 9192 MB/s
Write: 9610 MB/s
Copy: 9312 MB/s
Latency: 52.1ns
Theoretical limits: 6070MB/s from 6400MB/s (theoretical)

5:4 Memory array (266Mhz Strap):
CL: 6
RCD: 6
RP: 5
Ras: 17
RRD: 4
RFC: 80
WR: 7
RPT: 4

RTW: 8
WTR: 3,5
RTR: 4,5
WTW: 4,5

WTP: 18 (or 'Auto' 16T)
RTP: 5 (or 'Auto' 6T)
PTP: 1 (or 'Auto' 1T)
PRE to Active/REF: 6 (or 'Auto' 6T)

Voltage: 2.248v
Minimum North Bridge Voltage: 1.520v

Static Read: Disabled
'Clock Twister': Moderated/Lightest (Normal, Lighter depending, This in these speeds doesn't have high adjustment of speed)
Performance Level: 7

This combination speeds at 1125Mhz (450FSB):
Read: 9102 MB/s
Write: 9490 MB/s
Copy: 9722 MB/s
Latency: 50.4ns
Theoretical limits: 6282MB/s from 6400MB/s (theoretical)

-----

Both of above combination's are highly questionable as what the motherboard North bridge and PSU can do. Your Northbridge at 1125Mhz will need Minimum 1.520v this will shake the PSU Rails quite badly and any such shake will render your board unbalance.

Voltages are for North Bridge are calculated on P35/X38 and X48 boards. Speeds above are more than any benchmarks I have seen on network. Also still adjustable by tREF for read. 1:1 will be rock solid in any scenario you can boot it as long Northbridge and CPU are stable. 5:4 will need very high knowledge of board to get stable on some cases which doesn't have BIOS to adjust all values.

-----

Troubleshooting *Special Scenarios*:

My USB controller totally stopped functioning on POST, what can I do?

You can get your USB controller back by 2 means.
1. You do full power cycle(from your PSU), reboot & set values again on BIOS with better voltages.
2. You do force reset to 'Bad overclocking': 1. Press reset 2. straight after hit power button down for 5 seconds 3. Wait for couple seconds & power up. > You should now be able to add all values again.

My IDE Controller vanished, what can I do?

exact same procedure as above. However, in this case you will need Full Power Cycle and it will work just fine with current settings.

My USB started lagging/slugginess on system, what can I do?

1. This error is directly an memory/North Bridge voltage issue try raising them. However, both of above specs are calculated to not to do this. The specs are on very edge and so this might happen. This might also be only temporal effect of first setuped values and your north bridge has not stabilize yet to new voltage in this case try rebooting it 1-2 times see does it simply stabilize.

2. You can try rising South bridge to 1.100v or 1.125v. Seems to have effect of stabilizing OR general guideline +0.025-0.050v from what ever the board does in stock settings at OS.

My Network Controller vanished, Jammed, Stopped working in Windows, what can I do?

1. Go to BIOS and set Network controller 'Boot ROM' Enabled.
2. Reboot the machine and try booting to Network controller.
3. Reboot again and your network controller will start working as normally.

On case of network controller totally vanishing. You need to do following, what can I do?

1. Reboot the machine to BIOS Defaults.
2. Do Full Power Cycle from your PSU.
3. Restart machine and your Network Controller should be back, if error continues as none functional repeat 'Boot ROM' procedure above.

Why does this happen?

This would be because all of these devices depends on voltages we change 'too low' in a sense of bandwidth increase pretty aka. shakes the deck a bit, but pretty normal behavior. Of course for anyone with one computer and not familiar (or panic on error :P) none of above settings are recommended anyway.

I didn't find answer here what should I do?

Well, simply dump an guestion to screen I've done more than 17 years of support and work with thousands of machines. I'd say there's no problem I wouldn't have some idea what's going on.

**dima_s** · 07-24-2009, 01:39 AM

Hi!
I have a q9550 chip, striker ii formula motherboard (n780i) and 2x2gb gskill pi pc8500. I want to run fsb 1800 and memory 900 so the ratio will be 1:1. which timings and advanced timings should i change and what voltage to use for northbridge?
Should I mess with GTL Ref voltages? I have positive and negative values for gtl. what is the difference?

thank you!

**dima_s** · 07-24-2009, 09:43 AM

really need help with this. the settings given at first page didnt work for me, i always get errors like page fault in non paged area bsod or rounding was 0.5 expecting 0.4 and etc. it wont even post at stock. the hardwere is ok, maybe need another bios? now with 1901 bios (striker ii formula)

Announcement

6400/8800PI-Black some timing testing

6400/8800PI-Black some timing testing

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment