DDR3: Kingston and OCZ at 1333MHz

We had a first look at DDR3 performance back in May with an engineering sample of Corsair's DDR3 DHX memory. However, as some of the first DDR3 modules to run off the production, Corsair's modules were rated at a CAS-9 latency, which is pretty high in comparison to the respectively lower latency CAS-7 modules we have on test today.

All right, let's not start whining about "latencies going up" or "how is seven low??" because it doesn't work quite like that. DDR accessed two bits of data per clock cycle, DDR2 accessed four per clock and now DDR3 accesses eight.

In addition to the change of topography making access take longer as the signal goes through every bank, the extra time is only in response to having more data being sent per clock and frequencies being scaled.

The latency criticism is true to a certain extent because 1,333MHz equates to 0.75ns per clock, which makes 7 CAS clocks = 5.25ns, where as 800MHz equates to a longer 1.25ns but just three CAS clocks are 3.75ns. You have to drop the 1,333MHz CAS latencies down to CAS-5 to reach the same 3.75ns latency, but you've also got nearly twice as much available bandwidth at 11,000 versus 6,400MB/sec.

This means in a high throughput, longer latency scenario DDR3 should work better. It's then somewhat ironic that this kind of system was employed with the long pipeline of the old Pentium 4, instead, from experience the Core 2 architecture works better with short, sharp accesses in a low latency memory environment rather than needing a ton of bandwidth. The Core 2 has a ton of cache and far better pre-fetchers than anything on the P4s, which is meant to "hide" memory access latency, however these pre-fetchers aren't perfect.

Today, we've got modules from Kingston and OCZ on the test bench and without further ado, let's have a look at the modules...

Kingston HyperX KHX11000D3LLK2/2G

Manufacturer: Kingston
UK Price (as reviewed): £302.68 (inc. VAT)
US Price (as reviewed): $452 (ex. Tax)

The string of letters and numbers for Kingston's model name translates to a PC11,000 "low" latency kit of 2GB DDR3. It's strange, because even though the industry continues to use the PC-bandwidth numbers to represent module speed, fewer and fewer consumers and businesses use it because we're not bandwidth limited in this day and age. PC3-11,000 translates to a memory clock of 1,375MHz, rather than the official JEDEC speed of 1,333MHz. Kingston either feels that 10,666 doesn't quite have the same ring to it or it wanted to get a slight clock edge over its rivals.

Kingston does have an ultra low (UL) latency pair of 1375MHz modules which are rated at an eye watering 5-7-5-15, instead of the 7-7-7-20 modules we have here, but what stops us just overvolting, overclocking these LL sticks and saving a ton of cash instead? In actual fact, during testing we managed to drop the timings down to 5-6-5-15-1T at 1.85V, fully stable, but it was only at 1,066MHz not 1,333MHz.

Kingston HyperX KHX11000D3LLK2/2G Details:

Kit: 2 x 240-pin DDR3 Double Sided DIMM
Module Size: 2GB Dual Channel Kit (2 x 1GB)
Module Code: Kingston HyperX KHX11000D3LLK2/2G
Rated Speed: 1,375MHz DDR3
Rated Timings: 7-7-7-20 (CAS-tRCD-tRP-tRAS)
Rated Voltage: 1.7V
Memory Chips: Elpida

Kingston uses the same style of heat-spreaders that it sells with the entire HyperX series, just simply changing the DDR2 to a DDR3. There's a new futuristic look and style of sleek heat-spreader shown on its website, but it's disappointingly not used on these modules.

Elpida DRAM chips are used on these modules, which seems like a common choice given many other companies use either this or Qimonda for 1,066MHz and 1,333MHz modules. The DDR3 Micro D9s seem to be the preferred choice for enthusiast modules like the Corsair Dominator and DHX ranges, so we can only assume that there might not be much headroom in these modules. However, low latency might be a more appropriate option for Core 2 CPUs like it is with DDR2 (since Intel is the only one supporting DDR3 until sometime next year).

Memory: Is more always better?

Have you ever wondered what all of the fuss is about when it comes to memory? Memory manufacturers are going out of their way to sell you what they consider to be the best memory on the market for gamers, while also trying to push you into spending more money on more memory because - according to the memory makers - 2GB of memory is now becoming the industry standard for gaming systems.

The question is, whether doubling your current 1GB configuration to 2GB will be beneficial to your gaming experience or not. There are several options to take, too. Should you keep your current 512MB DIMMs and add another two to give you a total of 2GB, accepting the drawbacks (if there are any) from using the 2T memory timing?

Or, should you attempt to sell your current modules and purchase a pair of 1GB DIMMs?

Alternatively, could you get away with just sticking with your existing setup?

There are so many options on the market today, and we're going to attempt to answer these questions over the next few pages, while deconstructing some of the confusion that surrounds memory timings.

What does all of the jargon mean?

In the current computer market, where comprehending the intricacies of the latest generation of videos cards requires a great deal of stamina, it's easy to be deceived by memory and its many names - it's actually one of the simplest components in your system, and understanding what's best is easy.

Fred's driving along the road in his car, he spots something in the road and has to stop in a hurry, how long does it take for him to stop? There are two important components to this problem. The first is the most obvious (assuming something about his brakes, tyres and road surface) his speed will determine how far he travels when he presses the brake pedal.

The second, more subtle component, is his reaction time. How long it takes him to press that pedal. Actual experienced memory bandwidth is determined by two analogous factors; how quickly data can be transferred from the memory (memory speed), and how long it takes this transfer to start (memory latency).

Here's a quick primer in memory jargon:

Memory Speed:

The link between the CPU and memory is called the memory bus. Often, it runs at the same speed as the Front Side Bus (FSB), which regulates the communication between the CPU and lots of other system components. The newer Intel Pentium Processors seem to run better with the memory using an asynchrous 4:5 memory divider, meaning that the memory is running faster than the front side bus. The bus speeds are measured in MHz, or million clock cycles (CC's) per second.

Modern processors transmit 8 bits of data on every clock cycle, and all Athlon 64 and the older Socket 478 Pentium 4 CPUs run with a 200MHz memory bus. Newer Intel Pentium CPUs that use the LGA775 socket use either a 266MHz or 333MHz memory bus speed. The memory bus speed depends on whether they're an Extreme Edition or not - standard Pentium CPUs use a 266MHz memory bus, while Extreme Editions use a 333MHz bus.

If you multiply 400 (200 times 2 as Double Data Rate (DDR) memory runs at twice the clock speed) by 8, and you get a theoretical maximum figure of 3200Mbits/s transfer - hence the memory rating speed PC3200 found on the label of most new sticks of DDR memory. With the newer Pentium CPUs, you will see modules labelled with PC2-4200 (DDR2-533), PC2-5400 (DDR2-667) and modules up to PC2-8000 (DDR2-1000).

Memory Latency:

Addressing memory is much like reading from a large, multiple page spreadsheet. It doesn't matter how quickly you can read, before you can start you have to find the page the data you want is on (this is known as tRAS), work your way to the row and column the data's stored on (tRCD), when you've found the cell you want it takes some time before you start reading (CAS) and when you get to the end of a row you have to switch to the next, which takes time (tRP).

tRAS is the time required between the bank active command and the precharge command. Or in simpler terms, how long the module must wait before the next memory access can start. It doesn't have a great impact on performance, but it can impact system stability if set incorrectly. The optimal setting ultimately depends on your platform - the best thing to do is to run Memtest86 on your system with variable tRAS settings to find the fastest setting for your system.

The tRCD timing relates to the number of clock cycles taken between the issuing of the active command and the read/write command. In this time, the internal row signal settles enough for the charge sensor to amplify it. The lower this is set, the better - the optimal setting is either 2 or 3, depending on how capable your memory is. As with any other memory timing, setting this too low for your memory can cause in system instabilities.

CAS Latency is the delay, in clock cycles, between sending a READ command and the moment the first piece of data is available on the outputs. Setting CAS to 2.0 seems to be the holy grail with memory manufacturers, but the difference between tight timings and high memory bus speeds is an arguement that we hope to settle over the course of this article.

The tRP timing is the number of clock cycles taken between the issuing of a precharge command and the active command. It could also be described as the delay required between deactivating the current row and selecting the next row. In conjunction with the tRCD timing, which relates to the time taken between the issuing of the active command and the read/write command, the time required to switch banks (or rows) and then select the next cell for reading/writing or refreshing is a combination of the two timings.

The Command Rate timing is another timing that is important to maximum theoretical memory bandwidth. It's the time needed between when a chip is selected and when commands can be issued to the selected chip. Typically, these are either 1 or 2 clocks, depending on a number of factors including the number of memory modules installed, the number of banks and the quality of the modules you've purchased. The majority of memory available today is claimed to run at the faster 1T memory timing.

Memory latencies are normally quoted in the following format, CAS-tRP-rRCD-tRAS Command Rate, an example being 3.0-4-4-8 1T, with the numbers corresponding to the individual latencies quoted in clock cycles. Lower numbers are better, though in theory tRAS should be tRCD added to CAS Latency plus 2.


The differences between the 2x1GB and 4x512MB, and essentially the differences between using a 1T command rate versus the slower 2T timing, are relatively small in a selection of today's most popular games. However, there were instances where we found that we were able to play games with less frame rate lag and hitching as a result of having 2GB of memory using the 1T command rate timing.

However, the bottom line is that you will not see the same 10% performance increase in real games that we saw in SiSoft Sandra.

Memory timings are going to make the same subtle performance differences too, so it's a question of whether you can afford the faster modules that are capable of using tigher timings. The choice will ultimately depend on what memory you are currently using in your system and also how much you're willing to spend.

We'd recommend making the upgrade to 4x512MB over 2x512MB, even with the slight drawbacks we experienced in one of the four games we tested. That's because there are many other uses for your computer (aside from gaming) where you'll see the benefit of 2GB of memory. The general desktop experience is improved by 2GB of RAM, and I'm sure you'll never look back if you make the jump.

If you're willing to take a bit of a gamble (dependant on whether you'll be able to sell your current memory or not), we'd recommend swapping out your current memory for 2GB - it's just a case of whether you choose to buy the cheaper modules with looser timings, or whether you opt for memory capable of reasonably tight timings at DDR400. That'll ulimately come down to whether you're planning to overclock or not.