Thursday, February 12th 2009

Phenom II AM3 Plagued with DDR3-1333 Issue

Barely a week into the introduction of the DDR3-supportive AM3 socket CPUs, the processors seem to be having design flaws. This, as circulated by AMD in its revision guide document for the 10h family of processors (found here, page 80). The issue, as described by AMD, centers around the DDR3 memory sub-system. On machines with more than one PC3-10600 (1333 MHz) memory module populating a memory channel, the users may experience unreliable operation. The company does not get into the specifics of the symptoms. This issue however, does not affect systems with a module per channel (one or two modules installed in the motherboard), and only those with three to four modules installed.

The AMD K10 memory controllers on AM3-socket processor provide a 128-bit wide memory interface (with DRAM Ganged mode enabled), which amount to two 64-bit wide memory channels. On most motherboards, four DIMM slots with two slots sharing a memory channel are present. With this issue, one is not recommended to use more than one DDR3-1333 memory module per channel. AMD recommends a quick fix for the issue for systems using more than one DDR3-1333 module per memory channel: to manually specify the memory to run at 533 MHz (1066 MHz DDR), and accordingly set DRAM timings. As a little compensation, one can tighten DRAM timings with the drop in frequency. AMD will fix this issue in the next stepping (sub-version) of the CPUs. The "x-factor" with this erratum revolves around DRAM voltage, a significant factor. One might note AMD saying "the processor memory subsystem may exhibit unreliable operation over the allowable VDDIO voltage range", which leads us to think if there is a potential workaround with adjusting the DRAM voltage beyond the allowable range (read: over-volting the memory). We hope to hear more from AMD on this.

UPDATE (02/13): AMD replied to the report, saying that work is in progress toward fixing the issue. While not getting into the specifics, AMD indicated to us that apart from addressing the issue, the company is also working toward something "which will make you ‘feel cozy’ about the DDR3 support". The statement is ambiguous, and is best left uninterpreted at this point in time.
Add your own comment

105 Comments on Phenom II AM3 Plagued with DDR3-1333 Issue

#1
ShadowFold
If I am reading it correctly it's gonna get fixed. Calm the hell down. And it only affects two DIMMs right now, I would never get 4 dimms anyway.
Lrn2read
Posted on Reply
#2
Castiel
ShadowFold said:
If I am reading it correctly it's gonna get fixed. Calm the hell down. And it only affects two DIMMs right now, I would never get 4 dimms anyway.
Lrn2read
Well I am getting tired of waiting. But I rather have performance than lag.:rockout:

It will get fixed. I'm just making jokes. Except the part about switching to Intel. I really am going intel.
Posted on Reply
#3
jbunch07
ShadowFold said:
If I am reading it correctly it's gonna get fixed. Calm the hell down. And it only affects two DIMMs right now, I would never get 4 dimms anyway.
Lrn2read
LOL,Right.

It's not like ddr3 wont work at all, just an issue when using certain amount. It's not the end of the world people.

But I'll stick with what I said, untill ddr3 is more affordable and less problematic ill stick with ddr2.
Posted on Reply
#4
r9
DDR3 why spend a ton of money for 5% speed boost. If you want the top performance and money are not the problem you know where to go. Whole DDR3 Phenom II is illogical.
Posted on Reply
#5
johnnyfiive
Eh, long live DDR2, for now. DDR3 cost too damn much.
Posted on Reply
#6
Mega-Japan
I love AMD but... WTH is this?! Sorry AMD, I'll be abandoning you for the upcoming Core i7 975 coming in the next couple of months...
Posted on Reply
#7
newtekie1
Semi-Retired Folder
Has any of you claiming DDR3 is expensive actually looked at DDR3 prices recently. You can get 4GB of decent quality DDR3 for under $80, that isn't expensive at all.
Posted on Reply
#8
devguy
newtekie1 said:
Has any of you claiming DDR3 is expensive actually looked at DDR3 prices recently. You can get 4GB of decent quality DDR3 for under $80, that isn't expensive at all.
You are absolutely correct. It is not THAT bad right now. But look at the price for 8GB ddr3 RAM. Now, would you see yourself paying that much? If no, then why is this issue even relevant? And yes, I expect the prices of 4x2GB ddr3 to go down eventually, but I also expect the price of 2x4GB ddr3 to also go down as well, thus re-evading the problem whilst not having to occupy 4 DIMMs.
Posted on Reply
#9
TheMailMan78
Big Member
newtekie1 said:
Has any of you claiming DDR3 is expensive actually looked at DDR3 prices recently. You can get 4GB of decent quality DDR3 for under $80, that isn't expensive at all.
Currently it doesn't seem like it gives you much of an advantage except in benchmarks. And even then its minimal.
Posted on Reply
#10
newtekie1
Semi-Retired Folder
devguy said:
You are absolutely correct. It is not THAT bad right now. But look at the price for 8GB ddr3 RAM. Now, would you see yourself paying that much? If no, then why is this issue even relevant? And yes, I expect the prices of 4x2GB ddr3 to go down eventually, but I also expect the price of 2x4GB ddr3 to also go down as well, thus re-evading the problem whilst not having to occupy 4 DIMMs.
$160 for 8GB, yeah I could see myself paying that much if I felt I needed 8GB of RAM. Nothing I do really needs that much RAM, but there are plenty of other applications that do. And the truth is that the people that do need that much RAM will probably see far more benefit from having 8GB running at the slower speed than 4GB running at the faster.

TheMailMan78 said:
Currently it doesn't seem like it gives you much of an advantage except in benchmarks. And even then its minimal.
This is true, the gains are minimal, but I've heard the real gains are in overclocking headroom.
Posted on Reply
#11
Binge
Overclocking Surrealism
DDR3 can always go down in value, but since I know that you can get high end micron sticks that don't have XMP profiles for $30 per 2gb then I have to say all complains about cost are bogus.

AMD has a minor issue and it may be setting them back, but they still have a market even if it's the one some people might not want to be a part of. I still say phenom II is the ultimate lanbox cpu and if anyone contests that fine. DDR2-3 whatever, these processors have a purpose, and bugs will be fixed/put up for errata prior to a large scale launch.

Now if you all won't mind... if we're going to riot can we manage to swing by a Dunkin Donuts? I'm craving everything bagels.
Posted on Reply
#12
devguy
newtekie1 said:
And the truth is that the people that do need that much RAM will probably see far more benefit from having 8GB running at the slower speed than 4GB running at the faster.
So then they downclock their 2x4GB RAM modules to PC3-8500 speeds and run their 8GB like AMD advises. Problem solved.

And if I wanted more overclocking headroom, I'd probably buy at least ddr3-1600mhz RAM. And for 2x2GB of decent RAM at that speed (I don't mean the crappy 8/9 CAS modules), I'd be spending a little more (although MIR helps to make it only 20-30 more expensive than ddr3-1333, but I hate MIR). Hopefully the prices on those will go down to. It is nice to see that both Gigabyte AM3 boards natively support DDR3-1600, so maybe this issue is motherboard dependent? I wanna see someone use either of Gigabytes new AM3 boards and fill in all DIMM slots and see what happens.
Posted on Reply
#13
TheMailMan78
Big Member
Binge said:
Now if you all won't mind... if we're going to riot can we manage to swing by a Dunkin Donuts? I'm craving everything bagels.
:rockout: Ill drive! I already got Pearl Jam "Do the evolution" playing :laugh:
Posted on Reply
#14
adrianx
so this is the same bug like the frist phenom.... "you can not user 4 memory module in 1066mhz the memory controller will set the speed in 800mhz" and now if you want to use 4 dims... will be set in 1066mhz not in 1333mhz like normal :D (but this time ... is not speed it's the voltage or the phenom memory controller :)

from that I have 2 ideea :)

one is that will a cpu revision soon (aka stepping)

two it's possible to be only the amd marketing idea:), like the new stepping/version of cpu to give as a bug resolved and possible a higher speed for memory (1600/1800 or 2000 (if santa wants...) :)


also I want a hybrid mobo with ddr2 and ddr3 (evident optional) :)
Posted on Reply
#15
EastCoasthandle
I am not sure there is really enough information to formulate an opinion at this time. What is the allowed VDDIO voltage range for DDR3 1333 or DDR3 1600?
Posted on Reply
#16
btarunr
Editor & Senior Moderator
EastCoasthandle said:
I am not sure there is really enough information to formulate an opinion at this time.
It was an inquiry, not an opinion. We contacted AMD with it, and are expecting a reply.
Posted on Reply
#17
newtekie1
Semi-Retired Folder
devguy said:
So then they downclock their 2x4GB RAM modules to PC3-8500 speeds and run their 8GB like AMD advises. Problem solved.

And if I wanted more overclocking headroom, I'd probably buy at least ddr3-1600mhz RAM. And for 2x2GB of decent RAM at that speed (I don't mean the crappy 8/9 CAS modules), I'd be spending a little more (although MIR helps to make it only 20-30 more expensive than ddr3-1333, but I hate MIR). Hopefully the prices on those will go down to. It is nice to see that both Gigabyte AM3 boards natively support DDR3-1600, so maybe this issue is motherboard dependent? I wanna see someone use either of Gigabytes new AM3 boards and fill in all DIMM slots and see what happens.
Correct, I don't know why you are getting so defensive, I'm not saying there is a issue here.
Posted on Reply
#18
EastCoasthandle
btarunr said:
It was an inquiry, not an opinion. We contacted AMD with it, and are expecting a reply.
When the title reads "Phenom II AM3 Plagued with DDR3-1333 Issue" it reads like an opinion. That's why I asked about the VDDIO voltage range. In order for me (us, whoever) to have an understanding "what's being plagued" I inquired about the VDDIO voltage range. That way I understand what plague you are talking about. Once we have this understanding would we know why you have to downclock the ram, use one memory channel, etc.
Posted on Reply
#19
Haven
Well at least amd has the 920 and the 940 goin for them.
Posted on Reply
#20
btarunr
Editor & Senior Moderator
EastCoasthandle said:
When the title reads "Phenom II AM3 Plagued with DDR3-1333 Issue" it reads like an opinion. That's why I asked about the VDDIO voltage range.
That needn't be an opinion. It is an adequately established fact. The VDDIO range afaik is 1.5~1.8V. The influence of voltage on the issue is the area of uncertainty and we're waiting to hear from AMD on that.
Posted on Reply
#21
Darren
btarunr said:
one is not recommended to use more than one DDR3-1333 (or beyond) memory module per channel. This issue however, does not affect systems with a module per channel (two modules in all). AMD recommends a quick fix for the issue for systems using more than one DDR3-1333 module per memory channel: to manually specify the memory to run at 533 MHz (1066 MHz DDR), and accordingly set DRAM timings. As a little compensation, one can tighten DRAM timings with the drop in frequency. AMD will fix this issue in the next stepping (sub-version) of the CPUs.
Seems like a lot of people are talking bad about AMD already without reading the text thoroughly. DDR3 will work. However if you have ultra fast DDR3 such as DDR3-1333 or faster you have to manually configure it (opposed to automatic configuration) and reduce the bus slightly to 1066 MHz The user will benefit from faster latency timings.

However not all ram stick will cause incompatibility because it says "the users may experience unreliable operation" and hence you may have no problems with DDR3-1333 or faster memory, and therefore this issue is 100% hyperthetical and it might materalise that users having no problems whatso-ever! in essence it will vary user to user.
Posted on Reply
#22
EastCoasthandle
btarunr said:
That needn't be an opinion. It is an adequately established fact. The VDDIO range afaik is 1.5~1.8V. The influence of voltage on the issue is the area of uncertainty and we're waiting to hear from AMD on that.
btarunr, to me your title reads as an opinion. Re-read my previous post, I've edited it. Until we hear from AMD I really don't see what's being plagued here.
Posted on Reply
#23
btarunr
Editor & Senior Moderator
EastCoasthandle said:
In order for me (us, whoever) to have an understanding "what's being plagued" I inquired about the VDDIO voltage range. That way I understand what plague you are talking about. Once we have this understanding would we know why you have to downclock the ram, use one memory channel, etc.
It is pretty understandable from the title and the subsequent post itself that product is not functioning as intended and hence "plagued" with an issue, description of which from whatever we learn from the AMD document, is provided below (the title).
Posted on Reply
#24
EastCoasthandle
btarunr said:
It is pretty understandable from the title and the subsequent post itself that product is not functioning as intended and hence "plagued" with an issue, description of which from whatever we learn from the AMD document, is provided below.
I see, ok let's do a quick example then because I really don't understand. For example, a Asus M4A79T Deluxe, PII 925, twin kit of DDR3 1600/1333. Without knowing the VDDIO voltage range when should a user expect a problem?
Posted on Reply
#25
btarunr
Editor & Senior Moderator
EastCoasthandle said:
I see, so under what circumstances will this plague happen? For example, a Asus M4A79T Deluxe, PII 925, twin kit of DDR3 1600. Without knowing the VDDIO voltage range when should a user expect a problem?
"the allowable VDDIO voltage range". Since AMD didn't specify that range and yet went on to issue it as an erratum, AMD makes you believe it is the common range in which memory modules operate in.
Posted on Reply
Add your own comment