Why doesn't your memory default to the advertised speed?

[notdrugged]

It's one giant snowball of variables. You are the DIYer afterall. But let's begin from the basics.

In studying the DIY market and PC component manufacturing since the 1990s, I've seen electrical tests done on motherboards. I've watched discussions take place on trace timings of motherboards and BIOS programming. I've seen specifications come and go on chipsets, CPUs, and memory modules along the way. Alot of us have. All of this, and more, have everything to do with what you're experiencing in utilizing high speed memory.

Your BIOS is capable of telling a controller chip what voltage to run your motherboard chipset, memory, CPU, PCI bus, etc. This is the first variable that requires consideration. What your bios states is not precise, and usually, by the time that voltage reaches the component, it is either higher, or lower. This is because all motherboards are based on a reference design but most of them are engineered from the ground up by different manufacturing companies. I want to think this voltage is usually lower, but I can't guarantee that.

Next, the memory controller of your system is not universal. Currently, two companies manufacture their own memory controller for the PC market. AMD and Intel. At one time, this was done by the motherboard chipset manufacturer, and these motherboards still exist, and can even be bought today. Of course, Intel has been making their own chipset based controllers almost continually. The point is, they are all very different between chipsets, and the variable gap becomes very wide when you're comparing between vastly different CPU platforms.

AMD builds on a history of utilizing memory technology very agressively. Probably because they've been designing their own cache architecture since the mid-1990s. It is almost impossible to detail how this was done, but based on what I have witnessed, this has alot to do with why AMD competes in the gaming market still with slower CPUs. I can remember back when VIA was using 4-bank interleve on their SDRAM chipsets. This is just one example of a memory tweak that can be done which doesn't involve adjusting timings. You won't see this in today's motherboards probably, but it could be due to a refining of the technology that no longer requires user interaction. What you see in the BIOS, therefore, is not everything your system is capable of or designed to do. The BIOS is merely a flat meaningless switchboard of levers and pulleys. And for this non-universal chipset using a non-universal CPU and other non-universal components, there's a couple competing BIOS programming outfits who through contract write their own controller software in collaboration with the whole of the industry.

All of this is very relevant in the fact that any specific memory module is unlikely to share the exact same specification when used in different platforms, most notably between AMD and Intel. For instance, it has been historically the case that because an AMD typically puts more stress on the memory than Intel platform, you might be more likely to require relaxed timings or a slight bump in voltage to achieve stability. However, this does not explain the whole phenomenon by itself. Not by a long shot. This is merely an example of more variables. Therefore, it cannot be universally assumed between all Intel and AMD systems in order to ignore all other variables. That won't work. Especially since it is likely that Intel have become considerably more agressive with memory efficiency lately.

So, what would the memory manufactures say? Well, I've read quite a bit of JEDEC documentation in the past. Maybe not as much recently, but here's the gist of the story: The advancement in memory speed relies entirely on materials science and manufacturing techniques. Furthermore, this is a very rapidly moving industry so much so that, JEDEC has always been very far behind in publishing standards. If I say what the fastest standardized speed is today, it may not be the same a month from now.

Like BIOS programmers, the standardization process you can imagine is a collaboration with the whole of the industry. But as I pointed out earlier, the memory manufacturers have already tested, marketed, and been selling modules capable of faster speeds months before the latest specification gets published. Maybe, this is where it gets hard to understand. But it shouldn't be when you truly understand the variables involved. All they know is that it works reliable at that speed with what they've tested, and they feel comfortable enough to warranty the memory even if you "overclock" it. The industry has to catch up now, and you're caught in the middle.

So, in a sense, you would be right to feel negative towards the module manufacturer because at some point [with technology advancement] the memory speed will be so universal that every motherboard will be capable of defaulting to the proper speed, timings, and voltage automatically. But then, isn't research more your responsibility?

So, your memory defaulted to a lower speed so that your system would boot. When it was manufactured, they didn't know what all systems it was capable of working in, or even what exact settings they required because of all the variables we discussed above. Furthermore, the additional voltage could not be specified because it had not been standardized yet. Therfore, it defaulted to a slower speed so that you could manually make it work. This was done purposefully. Any memory manufacturer will tell you that.

But you are the DIYer afterall. The reality is that you bought memory modules that when first manufactured were on the bleeding edge of technology. So it is really nobody's responsibility but yours. There has almost always been a market for people that want to run their memory as fast as possible, and reliably. If you didn't know that you got caught up in that, then you weren't the first. It's noob initiation. I've complained and literally raised all hell about the same exact thing. We were all noobs once.

One of the first things you learn is how to buy both quality and reliable memory. It's simple: Stick to the name-brand ones that have the lower default voltage, and stop worrying about raising voltage. JEDEC actually specifies a maximum voltage before sustaining damage to be allowed to call their memory DDR3 at the specified bandwidth capability. The manufacturer should not recommend a voltage higher than that. You can hit up wikipedia on DDR3 if you don't want to go through becoming a member of JEDEC just to read their documentation.

[JSLEnterprises]

DDR3 SDRAM Reading

Would like to add Intel XMP (Introduced 2008) Information.
Intel XMP (Extreme Memory Profile) profiles are tested and verified to run on Intel based Motherboards/Chipsets from different vendors and conform to all JEDEC standards regarding to their specified speeds (up to 1600MHz). XMP is an overclocked memory profile that is designed for the 'Noob' in mind to not have to worry about adjusting various settings portion of the OC process and is guaranteed to function at the speed specified. Major manufacturers (initially Corsair, Kingston, and Micron) have Intel certified XMP profiles embeded into their enthusiast products and have been tested with a variety of setups and motherboard vendors. An increasing number of Memory manufacturers have adopted Intel XMP. Further Reading Intel's XMP webpage

[notdrugged]

Thanks for the additional information. That's very good to know. I didn't even know what XMP was. Just how long it's been since I've really been up on things.

[Fle@B@gL@ne]

I just put this setup together last week
I have 1800 DDR3(ASUS Sabertooth x58 MB)
ram speeds 1066 1333 1600 no 1800 but 1866 and higher
It reads if locked CPU DDR 800 and 1066 only can be used
I'm at 1866 no troubles all voltages stock but ram 1.70 from 1.65