Gigabyte GA-P67A-UD5 + 2600K - DEFECTIVEx2

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
N

nickdouglas

Distinguished
Jan 15, 2011
9
0
18,510
This is a story of mega-fail by Gigabyte...

My friend, Chris, and I ordered parts from Newegg last Sunday (9th) to build new Intel Sandy Bridge LGA 1155 systems.

We ordered the same parts except for memory...

-Intel Core i7-2600K Sandy Bridge 3.4GHz (3.8GHz Turbo Boost) LGA 1155 95W Quad-Core Desktop Processor BX80623I72600K
-GIGABYTE GA-P67A-UD5 LGA 1155 Intel P67 SATA 6Gb/s USB 3.0 ATX Intel Motherboard
-G.SKILL Sniper 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory Model F3-12800CL9D-8GBSR (Nick)
-G.SKILL Ripjaws X Series 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory Model F3-12800CL9D-8GBXL (Chris)
-CORSAIR HX Series CMPSU-850HX 850W ATX12V 2.3 / EPS12V 2.91 80 PLUS SILVER Certified Modular Active PFC Power Supply
-Thermalright Venomous X - RT 120mm CPU Cooler
-Arctic Silver 5 Thermal Compound - OEM

...and we both of us received DEFECTIVE Gigabyte GA-P67A-UD5 boards. 🙁

By defective I mean, both boards boot, but there is NO DISPLAY if you plug a video card into the 1st PCIe (x16) slot. Both work if you plug the video card into the 2nd PCIe (x8) slot or even the 3rd PCIe (x4) slot. Additionally, on Chris's board, only the 4th DIMM slot works. Plugging RAM into any of the other three DIMM slots on Chris's board makes it unbootable.

The trouble LEDs (awesome feature) on the GA-P67A-UD5 board confirm these problems. When a video card is plugged into the 1st PCIe (x16) slot, the PCIe trouble LED light comes on. When RAM is plugged into the 1st, 2nd, or 3rd DIMM slot of Chris's mobo, the DIMM trouble LED comes on.

Answers to questions you probably have:

Yes, we called Gigabyte Tech Support (626-854-9338) and went through hours of troubleshooting. Gigabyte confirmed that these boards are both defective and should be exchanged. The Gigabyte tech suspects that there is a bad controller chip on this batch of boards. How big is a "batch" I wonder?

Yes, we tried booting them outside of any case. They are resting on the anti-static bags they came in, which are resting on firm antistatic foam sleeves. In or out of a case makes no difference, both boards boot with the same PCIe and DIMM slot defects listed above.

Yes, we tried multiple video cards. We tried two different XFX Radeon 5870 cards, an EVGA GeForce 8800 GT, and an EVGA GeForce 8800 GTX. All these video cards work in these systems (and in our old Core2Quad systems), BUT NOT in the 1st PCIe (x16) slot on our GA-P67A-UD5 boards.

Yes, we tried the DDR3 RAM individually in both boards and in every slot. All the RAM (Sniper and Ripjaws) boots in my GA-P67A-UD5 in any DIMM slot, and all the RAM boots in Chris's 4th DIMM slot. None of the RAM will boot if placed in Chris's 1st, 2nd, or 3rd DIMM slot.

Yes, we tried new firmware. They came with F2, we upgraded to F3, F5 and beta F6a. No change.

No, I don't think we are newbs. :kaola: We both work in IT as system admins, building and maintaining servers and desktop systems and we have been building our own systems for 10+ years.

No, I'm not here to deride Gigabyte. We like Gigabyte. We both have Intel Core2Quad systems with Gigabyte GA-P35-DS3R boards that we built 3.5 years ago. Those systems have been overclocked to 3.6GHz and have been rock solid for 3.5 years. No problems at all.

We have ordered new Gigabyte GA-P67A-UD5 boards from Newegg and opened return RMA's for these defective boards. I'll update after the new boards are installed on Monday night.

Has anyone else had good or bad experiences with a Gigabyte LGA 1155 board?

Nick
 
Sort by date Sort by votes
I assume that you are NOT using XMP and are maunally entering the RAM values??!!

F3-12800CL7D-8GBXH http://www.gskill.com/products.php?index=357

M/B Chipset
Intel P67
CAS Latency 7-8-7-24-2N
Capacity
8GB (4GB x2)
Speed DDR3-1600 (PC3 12800)
Test Voltage 1.6 Volts

CPU: | Slot 1 | Slot 2 | Slot 3 | Slot 4 |
Slots: 1 & 3 ; otherwise 2 & 4 - either is recommended

Extreme Memory Profile (X.M.P.) -> Disabled
System Memory Multiplier -> {16 or 1600MHz} ; BCLK 100 MHz
DRAM Timing Selectable (SPD) - > Quick
CAS Latency Time - > 7
tRCD - > 8
tRP -> 7
tRAS -> 24
Command Rate (CMD) -> 2

The next big issues are the Voltages; begin low {spec} and next try the higher voltages in the range. None of them are 'high.'
DRAM Voltage -> 1.60v~1.65v
QPI/Vtt Voltage -> Auto, or 1.1v~1.22v
 
Upvote 0 Downvote


"QPI/Vtt voltage, formally known as Processor Power for I/O, is the voltage for the integrated memory controller as well as the PCI-E controller. Intel’s Default is 1.050v and Intel's Maximum is (1.05 +/- 3%) 1.080v."


Don,

After manually setting my "System Memory Multiplier (SPD)" to 16.00 and manually setting my timings to 9-9-9-24-1T, I then had to raise my "QPI/Vtt voltage" to 1.100v to remain stable and not crash in Windows. I have read that others have had to go as high as 1.200v to remain stable.

I think you should try this too.

Nick
 
Upvote 0 Downvote
It's also known as "Uncore" on the previous-generation processors because it's the voltage for parts that aren't in the "Core", but it's now called "System Agent" voltage and includes the integrated GPU.

The reason I mention this is because the new processors don't seem to respond well to high System Agent voltage. I haven't been able to do much with settings over 1.25V, but you should check with someone who's done additional experimentation to find out "when, why, and how" to use higher voltage levels.

This of course is important because of the relatively-high memory controller voltage set in some XMP profiles.
 
Upvote 0 Downvote


In EasyTune6 I see "QPI/Vtt Voltage" and "System Agent Voltage" both listed. I wonder how they differ?

Nick
 
Upvote 0 Downvote
I'm not certain. I do know that CPU VTT is something else, and that various motherboard manufacturers use different labels. That's why I refer to "Uncore" and "System Agent" rather than the labels manufacturers use.
 
Upvote 0 Downvote
Well, you guys prodded me to get to a task that's been nagging at me - digging through the documentation for the new gen! Vtt is very interesting - if you do a search through the 1156's "Intel® Core™ i7-800 and i5-700 Desktop Processor Series Datasheet - Volume 1" for Vtt, you will find around eighty or a hundred citations; in the 1155's "2nd Generation Intel® Core™ Processor Family Desktop Datasheet - Volume 1" - exactly zero!

For the 1156, the Vtt literal definition is: "L3 shared cache, memory controller, and processor I/O power rail"

Limits, per doc are:

0304yr.jpg


...but, as usual, we never get a clue as to whether these limits are 'advised', or 'enforced' - a lot of the spec limits can be exceeded without problems, and sometimes with stability improvement... Another interesting aside is the infinitely fussy measurement spec - must mean something?! [:isamuelson:8]

I think, however, the most telling piece in the doc is this:

0305p.jpg


...although this is spec'd specifically for the PECI (dammit, I'm already completely spoiled by the bbcode on my forum - if I use an acronymn, I can add a 'roll-over' definition!) (Platform Environmental Control Interface), which is a comm setup to xfer thermal info, I believe it's pretty much applicable to all the off-die I/O. I've spoken to this discussing grounds, and their importance: nearly every single thing a piece of modern digital circuitry does is dependent on comparators - circuitry to determine whether a signal is high or low. The primary 'object of comparison' is the ground level, which is why proper, effective grounding is so important. The second is the termination voltage, which is the level for the high determination. For most circuitry, the term voltage is simply the supply voltage; thing is that CPUs have gotten so ungodly fast that to 'swing' the signals fast enough, and minimize leakage losses, different chunks run with different supplies. I think this is the reason raising Vtt sometimes gives more stability, especially when other voltages are 'pushed' - it gives you a bigger gap between hi and lo (histeresis - just a fancy way of saying the 'here be dragons' area where a signal is undefined) An interesting point is that this is not written in stone - in fiddling around with digital gizmos, sometimes you get improvements in speed and 'solidity' of signal detection by increasing the 'swing', sometimes by decreasing it [:fixitbil:9]

Anyhow - so far, I'm stymied in finding out what's happened to Vtt on the 1155's - but there's a crapperload to cover:
Datasheet V1 signal def's, technologies, elecspec's 110 pages
Datasheet V2 config registers defs, address mapping 290 pages
Thermal/MechSpec's 134 pages
Spec Updates 41 pages (so far!)
6 Series Chipset Datasheet 936 pages [:fixitbil:1]

In addition, Intel have released a full new set of IA32/64 instruction/system notes, which they are now kind enough to supply on a CD, along with a couple interesting whitepapers, for the asking!
 
Upvote 0 Downvote

I agree, my original numbers were for the X58 and not the P67. I did include the 'Auto' but until I posted -- Where were your numbers?? The IMC on the P67 apparently is extremely touchy. The DRAM Voltage - 0.5v doesn't pertain to P67.

DRAM Voltage -> 1.60v~1.65v
QPI/Vtt Voltage -> Auto, or 1.1v~1.22v

Safety is in talking people to death.
processork.png


Great OC Link for P67 -> http://www.overclock.net/intel-general/910467-ultimate-sandy-bridge-oc-guide-p67a.html
 
Upvote 0 Downvote
Isn't that nice? I still don't know exactly why the memory controller voltage is more sensitive, but I still believe it's some other components on the same voltage controller that are causing that sensitivity.
 
Upvote 0 Downvote
The P67 is all about low voltage, and clearly there are good things and bad things. I get Voltage = Heat.

P67 look at the IMC plus various RAM {Speed + CAS}, it makes for 90%++ of the people very little sense to buy faster than 1333 MHz RAM -> http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/1 For a 0.7~2 FPS gain from 1333 ($43 2X2GB) -> 2133 MHz ($160 2X2GB) you'd have to b an idiot to spend 400% more for 2 FPS; buy an SSD or ANYTHING but ultra fast RAM for the P67. I get the IMC is 1333 MHz but jeez!

My major fear for the X68 is it's going to be the SOS as the P67: no BCLK OC, #1 fear -> limited PCIe meaning no gains on lanes or lane speed vs X58, IMC that sucks, and enough 'caveats' to write a book. Slapping in 8-cores with no way to utilize them is also a waste. What are you going to do with x8/x8/x8 is saturated with a 30" or triple HD monitors. The GTX 580 SLI on 2650 X 1600+ is already showing signs of saturation on x8/x8; I get the X58 is x16/x16 but 3/4-WAY that isn't the case.

BTW - I've heard the rumor mill of X68 offering PCIe 3.0, but my bet is that it won't and you'll b SOL with some PCIe 2.x variant. Also, the SATA3 limit is almost broken, excluding PCIe RevoDrives that already break SATA3 limits, so what's going to happen there -- some SATA(4) card {similar to some of the Enterprise SSDs now}??
 
Upvote 0 Downvote
If Intel follows their own historical path, the x68 will be based on one of the lower SKUs of the Xeon Patsburg, whose details have already been leaked. Like the 1156 CPU's, the 2011 moves the PCIe onto the processor for better graphics integration - will have 5 x 8 lanes, combinable/bifurcatable anyway you want - though the most common board setup will likely be the already loveable four primary slots, where you can either populate two for 2x16, or all four, for 4x8, with the remainder either on another x8 slot, or bifurcated across two fours... The lower Patsburg SKU has 4xSATA2+2XSATA3, 14xUSB2, at least eight (likely more) PCIe Gen2 on the PCH for random stuff like enet PHYs, WLAN, BlueTooth via an ExpressCard hookup, loses the QPI link to the CPU in favor of a second gen x4 DMI [the higher SKU Patsburgs will sport a x4 PCIe Gen3 link here...], and, most intersting, has a hard port for a new standard's pieces - ONFi (Open NAND Flash Interface) - integrated hybrid SATA3 HDDs, perhaps?!
 
Upvote 0 Downvote
It's X78, to avoid confusion with the Z68...oops, you didn't see that...these are not the chipsets you're looking for... 😛
 
Upvote 0 Downvote
[:bilbat:9]

I dunno if I'm just getting confused easier in my advanced state of decrepitude [:bilbat:6] , or if Intel's working a lot harder at obsfucation!?!

I like Charlie Demerjian's name for the new stuff: "iCoreDSN" [Designation Signifies Nothing!]
 
Upvote 0 Downvote
Trust me, I can read the wiki versions as easy as anyone. AND hope it's wrong. AND assume {hope} that something similar to the NF200 will be added.

Sadly, yes I realize per spec both the X58 & X68 both offer 40 PCIe lanes of PCIe 2 crap again. The PROBLEM is GPUs & SSDs are leap frogging the chipsets, and unlike before it's creating a major problem. I'm not aware of 'X78' other than rumor mill but I am aware of Z68 {OC version of H67}. http://www.tweaktown.com/news/18396/lga2011_consumer_board_pics_ahoy/index.html

X58 http://en.wikipedia.org/wiki/Intel_X58
X68 http://en.wikipedia.org/wiki/Intel_X68

 
Upvote 0 Downvote


Well then, it's just a matter of waiting to see if the Wikipedia entry will "need to be corrected" :)
 
Upvote 0 Downvote
I dunno about that WIKI article - the Intel systems diagram shows no QPI except the chip to chip, for NUMA and cache-snooping, which disappointed me, as I was hoping for another four-way loop, like the Tyan 7025. Hoped that Sapphire, having kidnapped EVGA's MOBO design team, would be bringing us some striking stuff - an SR-2 withouth the NF200's buggery...
 
Upvote 0 Downvote
It's really all about latency, where 1600 CAS 8 has the same response time as 800 CAS 4 or 1200 CAS 6. So, I'd find a good price on 1600 CAS 8 or 1333 CAS 7 and go that rout.
 
Upvote 0 Downvote

On an X58 or anything that I am use to seeing before the P67 I agree, but look again at these tests. They are not what I expected. The synthetic are to me of little consequence.

http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/4
http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/6

 
Upvote 0 Downvote



Hi Nick,

Thanks for the tip, but at the moment I've given the motherboard back to be checked to see if it's working properly. I felt like changing my 1600MHz RAM to 1333MHz. So I can rule out the RAM for sure. This is getting too fiddly and annoying for me. I just want a PC that works without having to tweak it just to find what is the right settings.

So now, I've got to wait for 2 or 3 weeks.. yes.. weeks for the distributor to get back to the shop that I bought the motherboard from.

I put a request for them to test it and to also provide me with the the best memory and config setting (which I don't think will ever happen)

Also, the place where I bought it says they've sold the same motherboard to a few other people and haven't got a single one returned. Which I think is a silly statement from their part.

I'll keep you guys updated when I get the motherboard back.

Still no new PC and it's been almost 2 weeks :-(

Don.
 
Upvote 0 Downvote

I've seen you link that 'tool' before, the link is broken [url=http://www.mediafire.com/?zgt3ujgjg4n]http://www.mediafire.com/?zgt3ujgjg4n 'fixing' it I saw DDR3MemCompareTool.xls

My above observations, are the very poor gains in DDR3 speeds over 1333 MHz. I've seen similar benches before on different chipsets and CPUs and typically there's some gain in FPS 1333 vs 1600 vs 1800 vs 2000+ MHz; the P67 is like hitting a wall above 1333. Strange?!
 
Upvote 0 Downvote

Assuming P67, then you really need to read this article - http://www.bit-tech.net/hardware/memory/2011/01/11/the-best-memory-for-sandy-bridge/1

As far as 'tweaking', the BIOS is not going to 'magically' set your RAM to its optimum settings. The best case is that your RAM supports XMP and your MOBO properly recognizes it w/o becoming unstable or BSOD, but you still need to set the XMP in the BIOS. Building a system requires some level of going into the BIOS. Just create a post and people will walk you through the steps. Otherwise by default the BIOS assumes you installed poor quality RAM and sets the speed slow and CAS high.

Maybe 'someday' the BIOS will read your components pull settings from the cloud and ideally set the optimum values...someday.
 
Upvote 0 Downvote

It actually doesn't assume anything. The SPD programming on memory tells the motherboard what speed and timings are supposed to be used at default voltage. This is where OCZ got in trouble with its "Gold" series, because the SPD programing was often wrong. SPD should always work without setting anything manually, but those "stable 1.5V settigs" result in the reduced speed and high CAS you mentioned.

XMP will always be "extended", so you'll always need to pick an XMP profile manually.

 
Upvote 0 Downvote

My 'real' point being if you slap-in say 1600 MHz 8-8-8-24 @ 1.65v RAM -> I have NEVER seen poof -> 1600 MHz 8-8-8-24 @ 1.65v in the BIOS -- ever - ever never. Instead more like 1066 MHz 9-9-9-2X @ 1.50v even using QVL + XMP RAM. Meaning the Speed, CAS, Voltage are at basic JEDEC settings. Heck I've seen XMP RAM + BIOS -> XMP = BSOD a rig plenty of times on Intel. Namely GA.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom