agree :DQuote:
Originally Posted by Nanobot
Printable View
agree :DQuote:
Originally Posted by Nanobot
:D Hey dude - nice show here, think You're a lucky man now..... :p:
Maybe You turned it into a FX-57 by breakin' the pins or You got the last bless from the pope for that little thingie....... :ROTF:
But do me a fav and contact Your local doctor - I wonder about Your healthy, it NEVER happened to me I forgot which CPU plugs in the socket.. :p:
Maybe we should open a new section here at XS: OCing with damaged hardware...... :rofl:
But good to hear the FX is still alive at all........... :toast:
I mapped out all the pins that are broken if anyone cares still. There are some that I cant tell if they are broken or just didnt show up. I only mapped the ones I could tell for sure. If I had a better pic, with boxes over the indevidual pins marked I could give a complete break down. I think you might have a couple more pins broken that comtroll the VTT_SENCE pin since it soundn't alow you to boot with a fulty VDD pin. You are lucky you lost one gound and one power of the same linkage.
vss=Ground. OMG!!! YOU HAVE NO GROUND...let the sparks fly!!!! Not really though. Since you lost one VDD losing one gound almost stabilises the situation...luck bastard. If you wouldn't of lost this ground and the VDD, one of each, you didnt have a core melt down.
nc_al12
nc_al11
nc_10
nc_al9 hmmmm. Having a hard time finding the functinaly of there pins. My guess is that they are not connected to anything (NC). I think this is AMD of integrating featues like SSE3 after prossesor has been relesed. The a64 chip has LOTS of NC connections. My guess is that
AMD has a LOT more up its sleeve for the a64 chips. SMT could even be in order, but that
will take some time given the nature of the very short pipe that AMD has desinged.
nc_al8
nc_al7
nc_al6
vdd=core power supply (yikes!!!!) this is where the power comes in. You lost your main power pin!!! how does this thing run??? Answer. There are over 50 VDD pins. Just one getting broken means that your volt core will not be stable. It will be a value of X read by the VTT, since the VTT reads it against the VDD. Not a real issue though for the 939. Also, since there is a pin that has no power, VTT_SENSE will have a value of z. I am very suprized to see that AMD lets you boot even though you might have a power short, very very bad if that were to happen.
man u have a 932 pin FX55 now :D
Wow.. :toast: Good explanation Gsus......I wonder from all 939 pins, how many AMD reserved for future use/features?
ROFL :party:Quote:
Originally Posted by $0m#0n#
You can then ebay it, quoting:
That would be a good one ;)Quote:
Originally Posted by EBAY SALES THREAD
hahahaha, one more motherboard revision from amd...nothing unusual there
You could've looked in the picture a couple of posts above you, you would see a piece of the socket 939 datasheet with the pins missing on this CPU.Quote:
Originally Posted by Gsus_Freak
You can prolly break off about 10% of all VSS pins without noticing (which is a lot of pins seeing as you have a lot of VSS pins), the A64 doens't require a lot of power so the pins themselves are thick enough to let at least double the power in. Also they're all connected via the processor package so there is no way to tell if there are pins missing. The only reason they spread the vss pins over the CPU is to prevent EMI.
The NC pins are just that: Not connected, these are reserved pins or pins with a debug function. If you don't connect these to the rest of the system there will be no change in functionality. With AMD most NC pins are really NC but with Intel most have a secret function (for shorting out the MP's for example).
The same story that goes for the VSS pins also goes for the VCC pins, they are spread to prevent EMI and get high currents to parts that need it. If you break off about 10% this would prolly not have any noticable effect. A VCC pin doesn't exist on it's own, its connected to all the other VCC pins.
If you read the datasheet you will see that VTT is measured against VSS and not VCC, also if you break of 1 or 2 pins there will be no change in either VSS or VCC so VTT doesn't change at all.
VTT_Sense therefor will not prevent you from running the CPU, the CPU works just fine. So there is no "power short" as you call it (you prolly mean shortage and not short-out?). Also there is no way for the CPU to know when it can't get enough power from the rest of the system, the vcore will simply get unstable and the cpu will crash.
If you break more VSS pins then VCC pins or the other way around doesn't make any difference. AMD makes great CPU's, don't say they do something wrong just because you don't understand.
@OPB: Nice drop that was, just the right side. Now you know: Don't fool around with frozen CPU's :D
The increase in OC is prolly a fluke, just a lucky day. It hasn't anything to do with the missing pins. If anything they will decrease OC speeds, not increase them.
... Nothing unusual? The socket-A platform has had only 1 revision in all the 5 years of it's lifetime.Quote:
Originally Posted by Gsus_Freak
You are confused with Intel introducing new sockets every 6 months I think.
to answer your question there are 36 edit*** (errr I recounted and it is 56 give err take....I cound count ballets for florida it seems) ***edit or so pins that have a NC designation. So, got room for DDR2 or DDR3 mem controler. I think DDR 2 will be skiped and move strate to DDR3. Problem with DDR2 is the ammont of Latecy inherent with DDR2 makes the onboard memory controler speed worth nothing. DDR3 has all the benifits of DDR (low latency) and DDR2 (high dencity, low power) with none of the drawbacks. Thank the Vid card people for the invention of DDR3. By the looks of it, SMT could be in order as well. Takes intel about 15-20 pins by the way I see it. Hard thing with SMT with the AMD chip is the built in memory controler can't keep up with the HTT setup in a SMT situation. You would need a whole need step up of SMT memory control switches. So AMD great streghth is their greatest weakness on the SMT front. Though not imposible to pull of, it would max the chips features out, so a new board revision wouldn't be too far behind.
uhhh thory. I didnt say that AMD doesn't make great chips, I think they do. I was mearly commenting that the boot proceduce doesn't check the VDD_SENSE proble like the intel chips do. I consider that a slight problem for cases like the above where someone has damages some of the VDD chips. The boot sequnce, IHO needs to be revised to prevent an over volt or under volt. Then again, I am sure that AMD knows more than I do about the ability of their chips, so I hardly think it is an over sight.
The CPU does, however, know if it can boot or not boot baised on power. I could be wrong, but if you read section 7.8 and past that to the boot sequence, it requires the VTT to hte VDDIO, and the VDDIO is set relitive to the VSS and related back the the VTT_SENCE. I could be reading the power diagram wrong, so set me straite if I am...it's been awhile since I had to read power circuit diagrams. But I am farily confident.
O and the MB comment I was refeing to is the whole state of affairs where AMD said the 940 chip was for high perfmance server parts and 754 was for everyone else. Then they came out with the 939 for servers and enthusits and the 754 will be used as the economy board then phased out. Your right though....socket a was the best. Intel has so many board types out it makes my head spin hahahaha.
Please get your fact's straight: There are 55 NC pins and another 25 to 39 (depending on the situation) that are left NC on the motherboard because the functionality isn't required.Quote:
Originally Posted by Gsus_Freak
However that a pin must be left NC to run the CPU doens't mean these pins are up for grabs, they can be used for debug purposes or reserved for new functions.
Implementing DDR2 or DDR3 on a A64 wouldn't be a big problem, you can multiplex pins so you'll have plenty of pins available. Also it isn't just a cost of pins you must have available, these pins must be thought over.
However, there is currently no need to increase the bus speed on the A64 (most out of the box systems can run @ 230 no problem) and therefor you will not see a DDR2 or DDR3 implementation anytime soon. If they move from a 200 bus to a 250 bus you'll still be OK with DDR technology.
Also because of the way the K8 architecture is setup (it's designed for SMP and SMT systems) AMD will have no problem with the memory on these setups. You don't seem to understand the meaning of a bus? (No higher tech education?) You can connect as many memory modules and memory controllers on the memory bus as you want without much trouble.
You'll prolly won't want to because it will cost you bandwidth but 2 memory controllers for 4 memory slots works just as well as 1 memory controller for 2 memory slots or 1 memory controller for 4 memory slots for all that matters.
So... AMD has designed the K8 architecture with a lot of thought, you can't compare it to Intel's old FSB/Net-burst architecture. And with Intel you'll need a chipset upgrade for every cpu change (like SMT). With AMD this isn't required. Also the bandwidth on the HTT bus is enourmous, it can easily support a lot of CPU's no problem.
wow, talk about rude. Just trying to be help full and you get all pissy, jeesss. O well cant please some people. I wont dignify your insults of no higher education. Just trying to provide some help and blast me it every turn...so nice to see you elivating yourself by taking other points and giving them no credit at all.
Plus there is a huge difference when impbeding an SMT thread vs a stadard bus thread cycle...but anyway, I digress. Good night
OPB...any word on stablilty. You getting 100% rock solid gaming...or are you getting strange system hangs an reboots? And with the missing pins...could you identify in more clarity which exsact ones you broke. Read over in the venice OC thread...when you upgrade your chip and have no use for it anymore, I want to try and exparenent with the thermal probe, the venice and the winny have a thermal bug problem, I want to see if giving the probe some impedence would stop the problem. Would like your thoughts as well Thorry.
Read section 7.1 you'll see it requires a lot to take down the A64 (and we know these are safe estimations, you can go to 25% above those). So AMD doesn't not boot because of some missing VDD or VSS pins, neither wil Intel.Quote:
Originally Posted by Gsus_Freak
The boot sequence (not really called boot sequence but PrePOST) doens't need to be revised, if the motherboard can't handle the power the system will crash because of stability. The CPU itself can handle a lot so you won't easily damage that.
Undervolt and overvolt situations are 99.9% of all the times checked at the power provider and not at the power user (this applies to everything on earth not just the mobo/cpu combo). And us tweakers remove the OVP if we see the need to. If this were to be integrated into the CPU core you would be very sad.
About the sockets:
They are still phasing out socket-A, it'll be another 5 years before socket 754 is going to phase out (budget systems can use it for many years to come). Also the difference in performance between 754 and 939 is minimal so the Socket 754 platform is just great for those of us that just want a PC.
Socket 940 is now and has always been a server-socket with some early FX'es to provide the early-adopters group. Socket 940 is here to stay, as is socket 939 and socket 754 (which will prolly last the longest of all).
Please keep in mind that the ONLY difference between 939 and 754 is the dual-channel memory interface. Therefor the market splits in 2 sections: the normal users who want good value for money and the power users who just want a great performance. This is followed through into all products (CPU's, mobo's, etc)
Sorry didn't mean to come over as rude, but I would like to see the facts right on this forum.Quote:
Originally Posted by Gsus_Freak
I didn't mean any insult when I asked you if you have a higher tech education. It was a question not a statement. Some of the best overclockers I know don't have any high tech education. They learn from experience.
I'm not blasting you, simply correcting some (IMO) mistakes.
I am pretty sure that socket A was EOL like 3 months ago... As for the socket 940, they eliminated it from future products so I consider it EOL as well. Unless you see 940 on a road map that I missed? (cause we have a 940 246 Opteron that we are pissed about since no more chips are heading our way)
hahaha yeah...i must apologys too. I got angry for no real reason at all. My bad...I better go to sleep, it is like 4:00am where I am...I guess I am getting emotinal hahahah
From the way I understand the way AMD does there APU and FPU, you would need an memoy latch for SMT to work correctly...you are right you wouldnt need a separate memory controler...need to get my lingo right sorry.
No problem :toast:
AFAIK Opteron will use Socket 940 at least until 2006 and prolly longer than that.
The life cycle of server cpu's is a lot longer then that of desktop cpu's and there are rumours that there will be Italy and Denmark cored cpu's in 940 pin packaging. Just because there aren't any new cpu's for a while on a platform doesn't mean the platform is written off.
AMD has a lot of different server CPU's in the market at this time, throwing even more in at this time will not serve a lot of purpose. If there is demand they will release the dual-cored 940 opterons. Imagine a 8-way dual cored opteron system :slobber:
Socket-A has had it's EOL, but there are still vast reserves that are now being sold so it'll be here for a while. And maybe it will be adopted by Via since they are having problems with the 370 sockets they use.
Socket 754 is also used in mobile systems so that's prolly here to stay.
we just have a measly 246, cant afford the 8way, it is 1700 just for the rack. So 2006 you say. I guess that is good. But I was looking to last till 2008...wishful thinking I know. But anything is better than the k6-m's that we were on...dear god they were phased out fast. Had 12 racks with no new processors in sight, still runnig strong though...guess that is to AMD credit.
Would like to see a dual core 24x (well they are on 250 now I think). We have so many threads running at ones it is crazy (run a server for a local college). Sometimes the students have fun with starting their own threads (getting unsed to unix and solarice, with assambaly languae and pearl scripts). Have more threads running that a quilt at that time of year. hahaha
those 370's were sweet though for low power cunsumtion. I thought VIA was going to get in bed with AMD with the GEOAD(err something like that, err was it gecco cant remeber). Didn't think they would pick up on the socket a stuff. Little hotter than the stuff they were working with before, lot faster though. I find it finny that intel is having to jump ship with the netburst and go back to a pIII type archtechre (like the centreno) for the new yonah chip (or that is the rumor at least). Finaly they know that cycles per second aren't as important as caculations does per cycle. AMD knew all along :) way to go.
Anyway. I think I identifed the wrong pins. I used the chart provided by the other guy...problem is that is the left side of the cpu....not the right side...o crap. I will re find the pins according to the amd pin read out (later...im tired now)
I checked the pins, they are correct. Check in the photo where pin 1 is located and then check with the datasheet. Please take notice that the datasheet has the pins shown from above so left and right are reversed.
AMD's Geode cpu is a semi-competitor to VIA's line of chips. AMD leans more to the embedded side, VIA more to the desktop side.
However for the VIA desktop chips they could take up Socket-A. The socket-A doesn't use that much power, the minimal vcore is 1.1 or 0.925 with a mobile. With not much effort VIA could even bring this voltage down.
Because they don't offer a lot of performance they want to keep the price down, making a cpu for a already existing platform is a nice way to do this.
actualy thorry i think the one that is VSS/gound is ak8 which is another NC..double check that if you would (it is right inbetween two Vdd)
i get the bad pins as al: 7-13 and ak-8
You can't be sure on this photo, but safe to say he didn't break anything that causes the CPU not te run because it is running.
He broke some NC pins and maybe some VDD and VSS pins, no biggy, very lucky indeed.