Who would be interested in a new Winchester revision?

[conrad.maranan]

I plan on drafting a petition to be sent out to AMD later on this evening. This petition will propose a push for AMD to revise the current Winchester by researching and developing a fix for the memory controller-related issues that many of us are faced with.

There is absolutely no doubt in my mind that these new 90nm chips are flawed. Bottom line: who's interested?

[Aphex_Tom_9]

...and fixing the sub-zero problem (if it's really a winchester issue)

[conrad.maranan]

I can also add that to the petition. Please feel free to mention any problems that you feel are common to the current 90nm processor so that I may include this in my draft. At this point, I will include issues regarding the memory controller, Prime95 stability, and sub-zero bug.

I'll PM one of the admins to suggest that the petition be made into a sticky. From there, I'll contact AMD and refer them to the appropriate thread.

[SkyGod]

put me down too! I cant prime 95 @ 250X10 on my winnie but as it turns out I can do anything else I can play hl2 for hours and hours, run 3d mark05 over and over no problem there! I think prime 95 fails Because of the memory controller issues, but that just kinda what I've gathered or atleast learned to accept ;)

PS it will prime 220x10 but basically it wont go past 2.2ghz :(

[Joe Camel]

you can put me down with over 2 weeks worth of total frustration and bla bla bla...

GOOD LUCK!

[Big Deel]

put me down too! I cant prime 95 @ 250X10 on my winnie but as it turns out I can do anything else I can play hl2 for hours and hours, run 3d mark05 over and over no problem there! I think prime 95 fails Because of the memory controller issues, but that just kinda what I've gathered or atleast learned to accept ;)

PS it will prime 220x10 but basically it wont go past 2.2ghz :(

Not to sound like the forum police,but what are your winchester rated at ?
I bet it's not 10x250 :stick:
Also AMD did not make it to run frozen.
The bottom line is it is runs at what they rated it.

Just for the record my winchester does 240x11 at 2-2-2-5 :D

[conrad.maranan]

The bottom line is it is runs at what they rated it.
Unfortunately, for a lot of us, it can't even run Prime95 stable at stock speeds, regardless of cooling measures.

[=[PULSAR]=]

Hmmm put my name on a petition about how there winchesters having issues overclocking. Isnt that why all cpu manufacturers suggest to run them at spec. Fear of having my name red flagged when I try to RMA anything...kind of risky :shrug:

[conrad.maranan]

CPU manufacturers can care less about how we do in terms of overclocking their chips. Our complaints and gripes have to be related to issues that affect our CPUs at stock speeds.

For example: I have a 3000+ rated by AMD to run at 9x200MHz (1800MHz). My Jedi Knight Mega Deluxe PC-3200 memory runs 320MHz 2.5-3-3-10 in dual channel on an FX-55/K8N Neo2 Platinum combo. Let's replace the FX chip and swap in the 3000+ Winchester and push the memory at a lower 300MHz. In order to run at the speed AMD suggests, we'll drop the CPU multiplier to 6 for 1:1 operation. Guess what? Prime95 fails after 1 minute. Houston...we have a problem.

[trans am]

I thought AMD is going to release a new stepping revision in 1st quarter of '05. Maybe they are aware of the issues and they have already found a fix. I would post a thread in the AMD support forums, and attach the link here and we can all go there and complain directly on their forum. ACtually, I'm going to do that right now.

[RaptorRaider]

But there is going to be a new revision soon. :confused:
There's a topic right here (http://xtremesystems.org/forums/showthread.php?t=47268).

[trans am]

Please post your issues here:
http://forums.amd.com/index.php?act=ST&f=7&t=32222&st=0#entry280162

[jjcom]

Sure you can put my name down on that thing. I'll be going 90nm dual core later on, but I want to know that I can have a high FSB since dual cores will be wanting memory bandwidth.

jjcom

[Joe Camel]

Not to sound like the forum police,but what are your winchester rated at ?
I bet it's not 10x250 :stick:
Also AMD did not make it to run frozen.
The bottom line is it is runs at what they rated it.

Just for the record my winchester does 240x11 at 2-2-2-5 :D

HERE (http://www.xtremesystems.org/forums/showpost.php?p=624340&postcount=42) is some stock P95 testing...:bsod:

[phobix]

CPU manufacturers can care less about how we do in terms of overclocking their chips. Our complaints and gripes have to be related to issues that affect our CPUs at stock speeds.

For example: I have a 3000+ rated by AMD to run at 9x200MHz (1800MHz). My Jedi Knight Mega Deluxe PC-3200 memory runs 320MHz 2.5-3-3-10 in dual channel on an FX-55/K8N Neo2 Platinum combo. Let's replace the FX chip and swap in the 3000+ Winchester and push the memory at a lower 300MHz. In order to run at the speed AMD suggests, we'll drop the CPU multiplier to 6 for 1:1 operation. Guess what? Prime95 fails after 1 minute. Houston...we have a problem.

I agree with you that there is a problem.

I don't know what it is exactly but I have experienced alot of the issues you are descrbing, my brother's newcastle runs like a bat out of hell at high HTT with agressive timings and my winnies seem to complain so I have to relax the timings to get prime passible high HTT.

There are a lot of possibilities, I think the problems may be caused by higher current draw with the new AMD 90nm process and the power regulation design in the current crop of NF3-250 motherboards. It is almost if the motherboard manufacturers designed the boards based on specs on 130nm processors. The lack of an official AMD chipset leaves us at the whim of the interpretation of AMD specifications by VIA and Nvidia. I can say one thing, the NF3-250 does have it's fair share of issues but seems pretty much rock solid when using a 130nm CPU. I dunno, I hope we at least get to the root cause of these issues. It will be interesting to see if the stability with the current crop of 90nm SOI's increases once the NF4s are released.

[conrad.maranan]

Okay, folks. There will be no petition to draft as there is no reason to at this point.

Memory controller issues
I had a somewhat lengthy conversation with an AMD representative earlier this afternoon. He assured me that AMD engineers are very well aware of issues revolving around the on-die memory controller of 90nm process chips. Although he was unable to delve into detail where the problem lies, he continued on stating that a fix has been slated for integration into the next revision. So, when can we expect the new revision? He thinks first quarter of 2005. Whatever. Anyway, he instructed me to read-up on some documentation on their site (http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/31411.pdf) related to memory controllers.

So. After reading the aforementioned documentation, I'm stuck on sections 2.4.2.4 to 2.4.2.4.3. I have absolutely no understanding of scrubbing so forgive me if I ask the following: Are the issues we are faced with regarding memory errors related in any way to scrubbing? Is this more of a BIOS issue as opposed to a hardware problem? Are the memory-related errors that we see in Prime95 related to scrubbing?

Super mega tech nerds: please step up and further my understanding of all of this mess.

Cold boot issues
On an unrelated note, there are pin designations on the above link. If you're feeling brave, pins AG10, AJ1, and AJ2 have something to do with temperatures. I'm going to e-mail Onepagebook and see if it would be wise to disable these in a BETA BIOS.

Again, I know nothing about this and can only assume what the average Joe can.

[DevilsRejection]

what problems are there with memory controller? anything specific? so far i only heard the ones with the people trying to hit above 250mhz fsb.

[conrad.maranan]

Here's what I've noticed:

- Hitting a memory speed of x on a 130nm CPU, but not even coming close on a 90nm counterpart.
- Hitting a memory speed of x on a 90nm CPU, but not being able to hit that same speed several weeks down the line.
- Having to drop the CPU multiplier in order to compensate for the difficulty encountered by the memory controller when pushing higher frequencies.

I had to drop down a few hundred MHz on my 3000+ just to push my VX modules to 260MHz at 2-2-2-10. 6x260MHz (1560MHz) on a processor rated at 1800MHz is bad news. I started out with a CPU that clocked all the way up to 2610MHz on the retail HSF, and ended up with a square piece of silicone that's packed up and waiting for an RMA number. All of this in less than one month's time. Go Winchester! You rock! :upset:

[Geforce4ti4200]

are winchesters ondie controllers so bad? exactly how high do they max out at 2.5-3-3-7 with tccd? 250fsb? good thing I didnt get tccd, I would just end up reselling it for a loss, total waste of money. some of you guys have suggested it to me, I have told you over and over winchesters ondie controllers cant handle 290fsb(2% might be able to) and that if I dont get at least 283fsb 1:1 ill be better off 290 with "166" divider. I am hoping the winchesters ondie controller is good enough for at least 236fsb for when I get my corsair value which ill want to run it at this MHz. as for the sub zero issue, I have seen some people run that just fine so maybe amd is taking mesures even though they shouldnt care since .1% use prometeias. as for prime95 failling, its either the ram, mobo or the buggy program itself. It is known that prime95 can fail with stock settings but every other program in the world passes, for this reason I refuse to play the p95 game and if games are rock stable, whos p95 to tell me how high I can clock? :stick: I think amd is doing a good job with those winchesters, all they intended for them was to run at 200fsb and stock, any overclock is a bonus and I got 800MHz oc, that is just insane!


"I had to drop down a few hundred MHz on my 3000+ just to push my VX modules to 260MHz at 2-2-2-10. 6x260MHz (1560MHz) on a processor rated at 1800MHz is bad news."


and at 2.6GHz what is the highest fsb you can do? can you relax timings to 2.5-3-3-7 and see if you get higher? amd will tell you overclocking isnt guaranteed and to run 200x9=1.8GHz but for us overclockers, its annoying :( you might want a newcastle till amd comes out with the next winchester steppings or maybe just run a ram divider for now

[STEvil]

1T/2T and multiplier bug (9x seems to be an evil multiplier for some?).

[phobix]

you might want a newcastle till amd comes out with the next winchester steppings or maybe just run a ram divider for now

But is the ram divider method the tried and true method for eliminating the instability @ high HTT? Can anyone else here confirm that?

[charlie]

dunno.....

mine ran like a timex at 291HTT, 2 x 256mB TCCD @ 2.5-3-3-7.......

[Vince2005]

Add me I'll help

[bachus_anonym]

actually, maybe there's something more to it than just a CPU...
there's probably quite a few people that have had their Winnie up high in the clouds as far as HTT and total clocks go... and that's with TCCD as well as BH-5 mems.
like it was said before - maybe up to certain week of production there were problems but newer Winnies have no probs doing it...
i can hit 265x10 with my 3200+ and 315HTT with 166 divider - and that's Prime95 stable (screenies upon request)...
to me the only issue that should be addresed is the "below 0C" problem... although some folks had no problem at all :confused:

[conrad.maranan]

and at 2.6GHz what is the highest fsb you can do? can you relax timings to 2.5-3-3-7 and see if you get higher? amd will tell you overclocking isnt guaranteed and to run 200x9=1.8GHz but for us overclockers, its annoying :( you might want a newcastle till amd comes out with the next winchester steppings or maybe just run a ram divider for now
Highest memory frequency is 193.3MHz at 2-2-2-10 in dual channel using a divider. Winchester will not overclock my OCZ Platinum Revision 2, OCZ VX, or G.Skill TCCD memory at 2610MHz in dual channel. I can do whatever I want with my timings. I put a lot of time and effort into my testing, so whatever I say is the Holy Gospel as far as my components are concerned. My wife comes second to overclocking, which I basically practice as my religion.

We all know there are no guarantees with overclocking. Read my signature and it says it all.

chilly1 will have something in the works for me after the Holidays. I'm buying an FX-55 and making love to it when all is said and done. :slobber:

EDIT:
bachus_anonym - I've got a Week 39 CPU, so it's fairly new.

[corrupt]

My wife comes second to overclocking, which I basically practice as my religion.
Not something I would brag about, but I'm not you.

I would just wait until a newer revision comes out and give them a shot, I just got done fighting with my older 3200+1m CO and it wouldn't do 250 1:1 but this new 3400+NC does 275-280 all day with the same ram. Give it time newer revisions will come into play and might have some memory controller bugs worked out.

[conrad.maranan]

Not something I would brag about, but I'm not you.
It wasn't my intent to brag. It's more of a sad fact. Anyway, I'm buying an FX so I'll just hold off on a CPU until late January.

[corrupt]

I'm buying an FX so I'll just hold off on a CPU until late January.
Probably a good idea, good luck with it.

[Geforce4ti4200]

guys, for those unable to get high fsb, have you tried running command rate in t2?

[MUCHO]

EDIT:
bachus_anonym - I've got a Week 39 CPU, so it's fairly new.

UH oh... I got a week 39 too! Yikes!

:stick:

[conrad.maranan]

I would not run 2T even if it was my only option. 2T kills performance - big time.

Here's what I used to run with my OCZ EL PC-3200 Platinum Revision 2 modules in dual channel. Just days prior to my memory controller conking out, I couldn't run at this speed without Super PI 1M failing at the initial calculation.

http://webpages.charter.net/conrad-digital/WinchesterNeo2/winchester_at_290_on_air.png

[MUCHO]

Were you running high Vdimm?

I'm going to be at the AMD thingie on Wednesday. Maybe I'll ask them about this issue...

Conrad - you should go. Its in the City of Industry

[conrad.maranan]

2.80V or 2.85V for that shot, if my memory serves me correctly.

I signed up for that. If I can finish all of my work by tomorrow, I'll take the day off and see if I can learn something at the convention. I just really have a ton of questions that I need answered.

EDIT:
I wonder if I can save the hassle of having to ship my CPU back to AMD and just bring it with me to the event. Maybe they'll be nice enough to swap a 3000+ for an FX-55...hahaha!

[saaya]

There is absolutely no doubt in my mind that these new 90nm chips are flawed.

they are? :confused:

well, rest assured as the E0 stepping will be released very very shortly :D
hope that fixes the sub zero problem wich seems to be related to the on die thermal probe and the emergency temp shut down.

what other problems are there with those chips? :confused:

[Geforce4ti4200]

will the E revisions have improved ondie controllers? cause if they are capable of like 300fsb 1:1 with tccd at 2.5-3-3-7 I may consider tccd, else I am keeping corsair value because of winchesters ondie controller limits :( also do you or anyone know if the E revisions will be Venice or retain the winchester name and will they be binned below 4000+ like I hear? If not itll be 2 years before that 4000+ dips to $200

[saaya]

corsair value? :confused: thats not a memory type, just a memory brand... they uses like 300 different memory chips for all their different ddr400 sticks over time, so you cant talk about "the" corsair value memory.

current chips can do 300 1:1 so why do you care about an improved on die memory controller?
there will be 4000+ 90nm chips sooner or later, but they wont be a lot cheaper than the current 4000+ chips i guess.

you seem a little confused about the code names, check out amds roadmap.

[Murmlos]

saaya, when is the new "e" revision due? Is it first quarter 05 ? or am i confusing it with something else :O ?

[Geforce4ti4200]

saaya:

"fix for the memory controller-related issues that many of us are faced with."

" I will include issues regarding the memory controller, Prime95 stability, and sub-zero bug."


"I cant prime 95 @ 250X10 on my winnie but as it turns out I can do anything else I can play hl2 for hours and hours, run 3d mark05 over and over no problem there! I think prime 95 fails Because of the memory controller issues"


"For example: I have a 3000+ rated by AMD to run at 9x200MHz (1800MHz). My Jedi Knight Mega Deluxe PC-3200 memory runs 320MHz 2.5-3-3-10 in dual channel on an FX-55/K8N Neo2 Platinum combo. Let's replace the FX chip and swap in the 3000+ Winchester and push the memory at a lower 300MHz. In order to run at the speed AMD suggests, we'll drop the CPU multiplier to 6 for 1:1 operation. Guess what? Prime95 fails after 1 minute. Houston...we have a problem."


howcome everyones having problems running 300 at 1:1 on winchesters if you say its so easy? are they doing something wrong? He got 320 on the fx55 but on the winchester, 300fsb fails in 1 min, even with the cpu at stock!

[bldegle2]

speakin' of memory how is that new corsair value you got???

we haven't heard anything about it yet, my guess it didn't do what you wanted.

baldy :D

[Badge56]

I am now using Corsair XMS TwinX PC4000 with my 3200+ week 37 winnie and the darn stuff will not go over 220mhz. Its rated 250 !! 3-4-4-10 1T
It will run 2.5-3-3-8 1T at lower DDR333 settings.

I ordered some Mushkin LV2 black and when I got it it would not even run with the winnie when both where installed... Dual or single chanel.
The Mushkin guy says its Tccd and should run fine. He his sending me replacement sticks.

Did I waist my money buying the mushkins? From what I just read they wont go 1:1 with the CPU either. Is that correct?
Maybe I can get them 2-2-2-5 1T at DDR333 ?

What do you guys think? :banana4:

[Geforce4ti4200]

I have emailed the seller why I havent recieved yet. Maybe the mail is slow or it got lost. I hope he insured it. did you see alot of people having problems with winchesters ondie controller? good thing I didnt go tccd, I wouldnt do tccd justice anyway! if the venice's ondie controller is much better then ill also get tccd. question: will venice work 100% on nf3 mobos? I hear its too new and you need a nf4 mobo for it!

[saaya]

saaya:

"fix for the memory controller-related issues that many of us are faced with."

" I will include issues regarding the memory controller, Prime95 stability, and sub-zero bug."


"I cant prime 95 @ 250X10 on my winnie but as it turns out I can do anything else I can play hl2 for hours and hours, run 3d mark05 over and over no problem there! I think prime 95 fails Because of the memory controller issues"


"For example: I have a 3000+ rated by AMD to run at 9x200MHz (1800MHz). My Jedi Knight Mega Deluxe PC-3200 memory runs 320MHz 2.5-3-3-10 in dual channel on an FX-55/K8N Neo2 Platinum combo. Let's replace the FX chip and swap in the 3000+ Winchester and push the memory at a lower 300MHz. In order to run at the speed AMD suggests, we'll drop the CPU multiplier to 6 for 1:1 operation. Guess what? Prime95 fails after 1 minute. Houston...we have a problem."


howcome everyones having problems running 300 at 1:1 on winchesters if you say its so easy? are they doing something wrong? He got 320 on the fx55 but on the winchester, 300fsb fails in 1 min, even with the cpu at stock!

it sounds more like a bios issue to me, but im not sure, and when did i say winchesters dont have a problem running at 300fsb?
whoever has problems with prime stability, have you used memtest? results?

[conrad.maranan]

I think mdzcpa has summed up everything in this thread (http://xtremesystems.org/forums/showthread.php?t=47268&page=1&pp=25). I'm tired of whining and I'm 100% sure everyone is fed up with me complaining about memory controller this and memory controller that since early November. I'll shut up now and let AMD do what they need to do. Everything else is simply beyond us.

Cheers! :toast:

[saaya]

there are no infos about 90nm prime issues in that thread :confused:
or what do you mean?

[conrad.maranan]

Prime95 issues are not in that thread. That link basically sums up what the known issues with Winchesters are. All in all, I think the failed memory controller, Prime95 failures, and what have you are all somehow related to this.