P55 overclocking destroying chips

Venom

http://anandtech.com/mb/showdoc.aspx?i=3661&cp=1#comments

This looks to be a major problem for motherboards using foxconn sockets which lots of the bigger brands like Asus, Gigabyte and DFI etc use. DFI and Gigabyte have stated they are moving over to Lotes sockets to avoid this issue. Asus has stated they don't condon overclocking of their motherboards with is utter bollocks imho.

Posting here and in Tweek/modding as this could be costly for some folks.

Leman_Russ

Venom wrote: »

http://anandtech.com/mb/showdoc.aspx...&cp=1#comments

This looks to be a major problem for motherboards using foxconn sockets which lots of the bigger brands like Asus, Gigabyte and DFI etc use. DFI and Gigabyte have stated they are moving over to Lotes sockets to avoid this issue. Asus has stated they don't condon overclocking of their motherboards with is utter bollocks imho.

Posting here and in Tweek/modding as this could be costly for some folks.

Yawn.

Old news is old.

Instead of trying to scare people, post the full facts. Unless you are using 79 million volts and using Liquid Nitrogen or Dry Ice, this is not a problem.

Also, your link doesn't work.

Venom

Links working now. Reason I posted is alot of people are seeing chip damage on non overclocked systems.

Solitaire

Lots of ASRock and ECS P55s turning up with blown transistors too. This is what you get when you don't threaten to release 130W CPUs on a socket; OEMs take the piss and put in ludicrously weak VRMs. No point having 8 CPU power phases if there's no heatsinks and the whole enchilada is designed to "do a Raidmax" and detonate once CPU power draw heads north of 100W! :rolleyes:

Leman_Russ

Solitaire wrote: »

Lots of ASRock and ECS P55s turning up with blown transistors too. This is what you get when you don't threaten to release 130W CPUs on a socket; OEMs take the piss and put in ludicrously weak VRMs. No point having 8 CPU power phases if there's no heatsinks and the whole enchilada is designed to "do a Raidmax" and detonate once CPU power draw heads north of 100W! :rolleyes:

True, but the ASRock mobos that are assploding are the really low end crappy ones.

Also note that Gigabyte Mobos don't explode

Solitaire

Don't get me started on Gigabyte mobos...

leninbenjamin

Considering our previous discussions on the subject have derailed about 3 different threads at this stage i thought it best practice to revive this one. (Solitare maybe you should move them?)

Anyway, Swirlser on the graphics forum links to the article i'd been looking for. Basically it's not just extreme overclockers that are seeing this.

As some of you might know by now there has been some issues regarding the LGA-1156 Foxconn socket, but it looks like it's getting worse. Foxconn said they corrected the cause of the problems back in May, however even with the new revision of the socket the number of damaged motherboards and processors keeps growing. At first the issue was only reported by users with extreme overclocking and after demanding benchmarking runs, but now it's starting to appear with average overclock settings and on a higher number of motherboard models. There are already full stocks of motherboards on the market that use the socket behind this problem, with the highest number of them being from Gigabyte and ASUS (a recent AnandTech review suggests both may be considering switching socket providers on some boards at least). Some brands, including EVGA and MSI have suspended the use of Foxconn and are now using TYCO AMP and LOTES parts for their P55 motherboards. The problem is apparently caused by bad contact between some of the CPU pins and the socket eventually leading to the death of both of them. Let's hope Foxconn gets over this issue soon.

and here's the original article on the issue from anandtech

Also, in a subsequent article on anandtech regards overclocking p55 I saw that there has been at least one report of this happening with non-overclocked systems (although it's unconfirmed). I also saw someone claiming the same on extreme systems forum, although it could be the same person. But perhaps more worryingly a confirmed case of it happening with a LOTES socket is referred to (but not linked to) in the article.

All in all, I'd be wary of recommending a p55 right now...

also Leman, you might want to consider putting that i5 back to stock.

Leman_Russ

leninbenjamin wrote: »

also Leman, you might want to consider putting that i5 back to stock.

I checked my Socket immediately after placing the CPU in while I was building it. I have full contact.

Venom

Leman_Russ wrote: »

I checked my Socket immediately after placing the CPU in while I was building it. I have full contact.

Have you checked it since tho?

Leman_Russ

Venom wrote: »

Have you checked it since tho?

I'm checking it on monday. I need to replace some bits and bobs so I am gonna strip the rig down and rebuild it from the ground up.

I haven't had any BSOD, and I have had it run Prime95 for 24 hours stable, so I am not worried at all.

TBH, I am more worried for the people buying the P55A boards. I hear they are even worse and the PCIe lanes are completely ****ed when you use USB3 and Sata3

leninbenjamin

Leman_Russ wrote: »

TBH, I am more worried for the people buying the P55A boards. I hear they are even worse and the PCIe lanes are completely ****ed when you use USB3 and Sata3

Is that a limitation of the on-die northbridge? or is it a cost cutting decision by Gigabyte.

the other thing about the p55 though is that it's not clear what's causing the issue. we know it's connectivity, but there are reports that Gigabte have said it's as a result of poor grade plastic. That would seem to imply that it melts under heat and the pins fall away... so it's worth keeping an eye on it

youcancallmeal

Is there any clear way to establish if your socket is Foxconn or Lotes without going back to the manufacturer and asking them?

leninbenjamin

youcancallme>al wrote: »

Is there any clear way to establish if your socket is Foxconn or Lotes without going back to the manufacturer and asking them?

If it's a LOTES socket it will say it.


This is a Foxconn afaik:


sidenote: I found the guy claiming to have the issue on it at stock speeds:
http://forums.anandtech.com/showpost.php?p=28780238&postcount=18 credible?

youcancallmeal

Dang looks like I've got the Foxconn socket

Leman_Russ

youcancallme>al wrote: »

Dang looks like I've got the Foxconn socket

Easiest way to check, is look at the back of the motherboard. Foxconn Sockets have "Foxconn" on the Socket Backplate.

Doesn't get any easier than that!

And before you go RMAing the board, Put the CPU in, clamp it down, put the heatsink on (without the TIM), take off the heatsink and check the CPU. You should see "scuff" marks where it has connected. If it's connecting fine, you don;t have to worry, if it isn't, then send it back.

Leman_Russ

leninbenjamin wrote: »

Is that a limitation of the on-die northbridge? or is it a cost cutting decision by Gigabyte.

To repost what I read:

Sure, Gb new P55A line up of New Mb looks good!
until your read the fine prints!!!

Only x8 for your graphic-card, Yes only x8!!! Huh Sad
Or x16 using pcie rev 1.
and rev 1 is half speed of rev 2 = 250 for rev 1 and 500 for rev 2
So only way to get full speed on the x16 (rev 2)for your graphic card is not to use usb3 or sata .
As soon as using usb3 or sata3 then you cripple the graphic card bus.
For everyone to understand what I am saying I will give example

ex1
not enabled usb3 or sata 3
x16 = 16 lanes of pcie rev 2 =16 x 500 = 8 000 speed x16 port for your Gf card

ex 2
enabled usb3 or sata 3 using rev 1 (this set up will criple the sata 3 and usb3)
x16 = x16but in rev 1 = 16 x 250 = 4 000 max speed x16 port for your Gf card


ex 3
enabled usb3 or sata 3 using rev 2
x16 will be x8 but in rev 2 = 8 x 500 = 4 000 max speed x16 port for your Gf card


Sorry this Mb do not look as good as it´s cover!!!!
Sad Sad Sad

http://www.ocforums.com/showthread.php?t=623288

Looks like a limit of the bridge. -Shrug- I am not fussed to be honest. x8 > x16 for single GPU is less than 1% Difference, and even x4/x4 to x8/x8 for duals is only a 2% Difference, 5% difference for x4/x4 > x16/x16

Solitaire

Intel were criticised about not leaving any margin in Bloomfield's PCIe controller... Looking more and more like P55 is going to come to a very early and mediocre end. You can fix the sockets, but you can't magically find more PCIe lanes on a platform with a crippled PCIe controller lurking in the guts of the CPU itself!

Leman_Russ

Solitaire wrote: »

Intel were criticised about not leaving any margin in Bloomfield's PCIe controller... Looking more and more like P55 is going to come to a very early and mediocre end. You can fix the sockets, but you can't magically find more PCIe lanes on a platform with a crippled PCIe controller lurking in the guts of the CPU itself!

It's not an issue really.

Run a single GPU at 8x mode and the Sata+USB. A MASSIVE HUGE 0.5% DIFFERENCE!!!!!!!!!!!!!!!!!

Or run two GPUs at 4x/4x and the Sata+USB. A nominal 5% difference. Boo freaking hoo. :pac::pac::pac::pac:

leninbenjamin

Leman_Russ wrote: »

It's not an issue really.

Run a single GPU at 8x mode and the Sata+USB. A MASSIVE HUGE 0.5% DIFFERENCE!!!!!!!!!!!!!!!!!

Or run two GPUs at 4x/4x and the Sata+USB. A nominal 5% difference. Boo freaking hoo. :pac::pac::pac::pac:

Correct me if I'm wrong, but i think that's only with the HD5xxx man. I don't think other GPU's scale as well.

Edit: Btw lads, unseen posting on the tweaking and modding forum has said this issue happened to him with a LOTES socket.

Leman_Russ

leninbenjamin wrote: »

Correct me if I'm wrong, but i think that's only with the HD5xxx man. I don't think other GPU's scale as well.

Edit: Btw lads, unseen posting on the tweaking and modding forum has said this issue happened to him with a LOTES socket.

Ha! See, so all this foxconn malarky is a load of rubbish! :pac::pac::pac::pac:

And people use anything but 58xx?
What kind of mad world hav I woken up to!!!!!!!!!

leninbenjamin

Leman_Russ wrote: »

Ha! See, so all this foxconn malarky is a load of rubbish! :pac::pac::pac::pac:

Nah, it means it's an issue with the CPU pads themselves and not the socket.

Leman_Russ wrote: »

And people use anything but 58xx?
What kind of mad world hav I woken up to!!!!!!!!!

some of us like CUDA

Solitaire

Jebus Christo! I knew Intel wanted to force all enthusiaists over to X58 and was probably going to do something rather underhanded to do it, but... I don't think this was how they intended it to happen! Karma-slapped ftl!

Venom

Solitaire wrote: »

Jebus Christo! I knew Intel wanted to force all enthusiaists over to X58 and was probably going to do something rather underhanded to do it, but... I don't think this was how they intended it to happen! Karma-slapped ftl!

Im surprised AMD aint jumping on this more tho tbh as it totally killed any P55 upgrade plans I had.

Leman_Russ

Venom wrote: »

Im surprised AMD aint jumping on this more tho tbh as it totally killed any P55 upgrade plans I had.

Why? Because some CPUs are exploding when you pump massive voltages through them (ya know, just like every other CPU)?

c-note

Leman_Russ wrote: »

To repost what I read:

Quote:
Sure, Gb new P55A line up of New Mb looks good!
until your read the fine prints!!!

Only x8 for your graphic-card, Yes only x8!!! Huh Sad
Or x16 using pcie rev 1.
and rev 1 is half speed of rev 2 = 250 for rev 1 and 500 for rev 2
So only way to get full speed on the x16 (rev 2)for your graphic card is not to use usb3 or sata .
As soon as using usb3 or sata3 then you cripple the graphic card bus.
For everyone to understand what I am saying I will give example

ex1
not enabled usb3 or sata 3
x16 = 16 lanes of pcie rev 2 =16 x 500 = 8 000 speed x16 port for your Gf card

ex 2
enabled usb3 or sata 3 using rev 1 (this set up will criple the sata 3 and usb3)
x16 = x16but in rev 1 = 16 x 250 = 4 000 max speed x16 port for your Gf card


ex 3
enabled usb3 or sata 3 using rev 2
x16 will be x8 but in rev 2 = 8 x 500 = 4 000 max speed x16 port for your Gf card


Sorry this Mb do not look as good as it´s cover!!!!
Sad Sad Sad

Quick question lads,
I'm configuring my first ever build,

its for a core i5 system for general use + some gaming (should run gta iv cod mw2 + new games hopefully) budget < 850€

i had the gigabyte p55 ud4 provisionally selected as the mobo
the main difference i see between that an the p55A ud4 is that the "A" version has usb3 and 2*sata @ 6gbps

So would the issue above affect the p55 ud4 board, seeing as it hasn;t got the usb3 sata3 marvell chip?
this note appears for the p55A version of the ud4

* When dual graphics cards are used in 1st and 2nd PCIex16 slots, SATA3 / USB 3.0 (Marvell 9128 /NEC USB 3.0 Controllers) will work at normal mode.

OR should i steer clear of gigabytes p55 boards, (i was under the impression that gigabyte has a good rep?) (i've never bought a mobo before), my alternative board is the intel dp55wg.

Leman_Russ

Get the P55A board.

The "issue" is a non issue, as for a Single Graphics Card Solution, it will result in a balls shrinking 1% difference in performance :P

Solitaire

Leman_Russ wrote: »

Why? Because some CPUs are exploding when you pump massive voltages through them (ya know, just like every other CPU)?

Nah, because there's good odds you'd only have to wait a little longer before a mid-OCd one detonates, and if you wait just a bit more your stock-clocked chip might melt for fun too! :P

I'm not certain but it may indeed be epic tolerances fail on Intel's part, with the sloppier pad tolerances on the Foxconn sockets only exacerbating the issue. No idea about Tyco but there has been at least one LOTES horror story which shocked EVGA - they thought they'd dodged this bullet already!

Venom

Leman_Russ wrote: »

Why? Because some CPUs are exploding when you pump massive voltages through them (ya know, just like every other CPU)?

No because far to many CPUs are exploding at stock speeds for my liking.