RDRAM versus SDRAM...Your Opinion?

[The Finman]

I would really like to hear peoples opinions of one type versus the other. [img]graemlins/popcorn.gif[/img]

I found an interesting article on the subject and would like to hear peoples opinions that have actuallly worked with both types of memory.

<h2><font color=#003399>RDRAM versus SDRAM</font></h2>

During the development of RDRAM, the manufacturers promised a bandwidth of twice that of PC100, which is true to a certain extent, but this is only valid when comparing PC800 RDRAM with PC100 SDRAM. Confused about what "PC100" (or PC-100) and "PC800" really means? PC800 would lead one to believe that it should be 8 times the speed of PC100, but is it?

Upon closer examination, RDRAM uses a 2 byte (16 bit) wide databus versus SDRAM's 8byte (64 bit) wide databus. Obviously, this makes the PC800 rating a bit confusing, but there is an explanation. PC800 RDRAM is actually a double-pumped module operating at a 400 MHz clock speed. Double-pumped means that data is transferred to the RDRAM on both the rising and falling edges of the clock, which is often referred to as double data rate *(DDR), creating an effective 800 MHz memory rating. PC100 SDRAM, on the other hand, is referred to as single data rate (SDR) and operates at 100 MHz clock speed, which can only transfer data on the rising edge of the clock, thus having an effective 100 MHz memory rating.

*Not to be confused with DDR SDRAM.

Memory Bandwidth is Theoretical in Nature
If we were to compare theoretical bandwidth, without considering memory latency, we would see the following:

PC800 RDRAM : 800 MHz x 2 Bytes = 1600 MB/s = 1.6 GB/s
PC100 SDRAM : 100 MHz x 8 Bytes = 800 MB/s = 0.8 GB/s

However, there are a number of new chipsets that have been released lately by both VIA and Intel (beginning with the BX unofficially) that support 133 MHz, or PC133 memory. If we look at the theoretical bandwidth of PC133 memory, it appears as follows:

PC133 SDRAM : 133 MHz x 8 Bytes = 1064 MB/s = 1.064 GB/s

If we carefully compare these three sets of theoretical bandwidth figures, it would appear that PC133 is well above PC100, and nibbling at the heals of RDRAM performance, but is it really? It would seem that PC133 offers performance well above what we have seen in real world benchmarks. Okay, so what's missing?

What is missing is that these are theoretical bandwidth numbers that do not take into consideration memory latency, which makes a big difference in actual bandwidth. Unfortunately many companies, and all too often vendors, use these unadjusted numbers to promote one architecture's superiority over another.

As we note above, theoretical bandwidth alone cannot be used to measure memory architecture superiority. It is fact that memory latency imposes too much of a penalty on actual memory bandwidth, and is different for every architecture. Therefore, to be able to really determine architectural superiority all factors must be taken into consideration, including the latencies.

Differences in Architecture
RDRAM is a memory architecture that relies on a packet-based protocol with an access latency that largely depends its distance from the memory controller. Although systems with multiple RDRAMs have slightly increased latencies compared to single-RDRAM systems, RDRAM latency is still, in a manner of speaking, comparable to that of SDRAM systems. By comparison, RDRAM protocols and architecture facilitate memory concurrency and minimize latency, as opposed to SDRAM, which does not. This is especially beneficial when multiple memory references are being serviced simultaneously. The number of RDRAMs does not affect peak bandwidth, and an RDRAM-based memory system provides peak bandwidth twice that of PC100 SDRAM. The 1.6 GB/sec bandwidth of RDRAM is achieved with only a 16-bit data bus, and when combined with control signals the memory controller only needs about one third of I/O channels that SDRAM does.

SDRAM uses a different approach. It uses a parallel databus 64 bits wide, and adding modules to the system has no effect on memory latency. In addition to the 64-bit databus, the memory controller must drive a multiplexed row and column address to the SDRAMs along with control signals.

To accurately measure (within reason) SDRAM performance, two metrics must be considered, bandwidth and latency. Unlike SDRAM, RDRAM offers not only higher bandwidth, but its latency is much improved when compared to what we've come to expect from SDRAM. You might be surprised to note that PC133 SDRAM latency is actually worse than PC100. You may review a reference article from Samsung Semiconductor, Inc. here: Rambus Dram Performance.

Defining Component Latency
The accepted definition of latency is the time between the moment the RAS (Row Address Strobe) is activated (ACT command sampled) to the moment the first data bit becomes valid. Synchronous device timing is always a multiple of the device clock period. You can read more about memory latencies here: CAS Latency, what is it? and here: Application Performance and Loaded Memory Latency, both of which will open in a new window for you.

The fundamental latency of a DRAM is determined by the speed of the memory core. All SDRAMs use the same memory core technology, thus all SDRAMs are subject to the same latency. Any differences in latency between SDRAM types is, therefore, only the result of the differences in the speed of their interfaces.

At the 400 MHz databus, the interface to a RDRAM operates with an extremely fine timing granularity of 1.25ns, resulting in a component latency of 38.75ns. The PC100 SDRAM interface runs with a coarse timing granularity of 10ns, or about eight (8) times that of RDRAM. Its interface timing matches the memory core timing very well, therefore its component latency ends at 40ns. The PC133 SDRAM interface, with its coarse timing granularity of 7.5ns, incurs a mismatch with the timing of the memory core. This mismatch significantly increases the component latency to 45ns.

Latency timing values can be easily computed from the data sheets of the respective devices. For SDRAMs, specifically PC100 and PC133, the component latency is the sum of the tRCD and CL values (tRCD*CL = component latency). With respect to RDRAM's, the component latency is the sum of the tRCD and tCAC values, plus one half clock period for the data to become valid (tRCD*tCAC+(tCLK/2) = component latency.

Although component latency is an important factor in system performance, system latency is even more important, as it reduces overall performance. System latency is determined by adding external address and data delays to the component latency. In most personal computers today, the system latency is measured as the time to return 32-bytes of data, also referred to as the 'cache line fill' data, to the CPU. Cache issues are often overlooked, whether intentionally or unintentionally, and they play a large part in over all system performance. You will find our rather lengthy discussion of cache issues here: Cache Explained.

In a computer system, SDRAM suffers from what is referred to as the two-cycle addressing problem. The address must be driven for two clock cycles (20ns at 100 MHz) to provide sufficient time for the signals to settle or arrive entirely on the SDRAM's already loaded address bus. After both the two-cycle address delay and the component delay, three more clocks are required to return the 32 bytes of data. The system latency of PC100 and PC133 SDRAM add five clocks to the component latency. The total SDRAM system latency is calculated as follows:

40 + (2 x 10) + (3 x 10) = 90ns for PC100 SDRAM
45 + (2 x 7.5) + (3 x 7.5) = 82.5ns for PC133 SDRAM

RDRAMs superior electrical characteristics eliminate the two-cycle addressing problem, thereby requiring only 10ns to drive the address to the RDRAM. The 32 bytes of data are transferred back to the CPU at 1.6 GB/second, which works out to be 18.75ns. Adding in the component latency, the RDRAM system latency is calculated as follows:

38.75 + 10 + 18.75 = 67.5ns for PC800 RDRAM

Whether measured at the component or system level, RDRAMs has the fastest (or lowest) latency. And as mentioned earlier, as the result of the mismatch between interface and core timing, the latency of PC133 SDRAM is significantly higher than the PC100. RDRAM's low latency, coupled with its 1.6 gigabyte per second bandwidth, provides the highest possible sustained system performance. Granted, DDR SDRAM has demonstrated an entirely new concept regarding timing and latency issues, however the final judgment on that issue has not yet arrived even as we close upon the third quarter of 2001.

When we consider overall system performance, we must take note of the impact of L1 and L2 cache hits on memory architecture performance. (Review more about L1 and L2 cache issues in Cache Explained). In addition, individual programs vary widely in memory use, and as such have an array of different impacts on system performance. As an example, a program that uses a random database search using a large chunk of memory will so heavily impact the caches that the memory architecture having the lowest latency will have the advantage. On the other hand, some well written software that creates large sequential memory transfers that requires little CPU processing often easily saturates SDRAM bandwidth. RDRAM will have an advantage here as well with its higher bandwidth. In those situations where the software code fits nicely within the L1/L2 caches, memory type will have virtually no impact at all.

It has been quite some time now since Intel chose to implement support for the RDRAM memory architecture in its i820/i840 and upcoming chipsets. Most people in the industry thought that they were making a huge mistake, as the promised performance benefits weren't showing up in most of the "then current" benchmarks. In fairness to Intel and to RDRAM (Rambus), it wasn't all that long ago we were still using EDO RAM and SDRAM was the upstart technology, and at the time pretty expensive. We do not write benchmarking software, and with that we are qualified to ask the question of those that do, "is current benchmark software capable of accurately measuring the performance of RDRAM and DDR SDRAM?". Thus far, we haven't received a viable answer, and we believe that is due to the fact that each camp is more interested in promoting its own technology rather than developing the truth. In hindsight we've all seen the benefits of using SDRAM and its impact on overall system performance. If we look at the performance benefits SDRAM offered in its early days on applications that were popular then, it seems as though it didn't offer huge advantages. Yet, we've come a long way, and SDRAM performance has continued to improve even though the technology, at first, didn't seem to promise that much of an improvement.

Due to the growing demand in memory bandwidth, the arrival of GHz+ CPUs and the ever-growing demands of today's software, SDRAM has run into bandwidth limitations. Sure, DDR SDRAM and VCDRAM might be able to hold off the introduction of a new memory standard for a little while, but it's eventual arrival is inevitable. We have run many of the same comparative tests between RDRAM and DDR, and while DDR SDRAM might promise increased memory bandwidth, it will run into severe timing, latency and propagation delay problems due to its wide databus and the ever increasing clock speeds. At present (and for the foreseeable future) memory will be relatively cheap to produce, but as data rates increase, motherboards will need to have six or even eight PCB layers to be able to run these memory modules at these higher rates and clock speeds, thus increasing motherboard costs substantially.

While RDRAM is not perfect, it is one of the most promising solutions to bandwidth, latency and propagation delay problems. It's scalable, which gives it a distinct advantage! When it was first released more than a year and a half ago, it was extremely expensive, but that was partly because it was new and the market hadn't caught on yet. Although still expensive when compared to SDRAM, RDRAM pricing will drop further as more manufacturers begin recognizing its potential and begin start selling RDRAM. It will then become as commonplace as SDRAM is now. By the nature of the Rambus manufacturing process itself, in all probability Rambus will never be as affordable as SDRAM, but then again SDRAM doesn't offer the same performance. But isn't performance what you're actually paying for? Better technology always comes at a price. We are certain that you wouldn't expect your $1000 or less computer to perform nearly as well as a $5000 top-of-the-line model.

http://www.dewassoc.com/performance/...am_v_sdram.htm

[45semi]

I can only offer an opinion on that which I have experience with...DDR Ram in video cards.
The card I have in my main system:
VisionTek Xtasy GeForce4 MX 440

Difference in DDR -vs- SDRAM cards (I have both)
HUGE difference in anti-aliasing (jaggys)
No pixel shading capability
(Some of the tidbits in DX9 wont make a diff)
LARGE Frame per Second increase
Supposed to be good at DVD play, but I dunno personally


Now, a LOT of what I read says that full utilization of DDR is very dependent on the mobo.
Bus speeds, what chipset (Southbridge, VIA, etc) IDE interface, the new AGP speeds, etc

It's come to the point that you must do EXTENSIVE research to mate memory with processor & motherboard, in order to get the most out of components.
I considered myself a 'technophile' until recently, when I started doing research for a future gaming system, and saw that the variety & choices available for all the different components that make up a modern system have taken off with the advent of DDR.

If you want to do such research, in anticipation of buying/building a new system, I recommend my sources for info on up-to-date component evaluation.
My 1st choice:
http://anandtech.com
My 2nd choice: (Due to fact site leans somewhat to manufacturers who contribute freebies to eval. staff, IMO.)
http://www.tomshardware.com

Now, when you take in factors such as trying to stay in budget, preferences for a particular manufacturors product (I'm a fan of Asus mobo's!)
what a system is going to be used for primarily (gaming, process greedy programs-example>Photoshop, etc) then you have to take into consideration what parts you select to use.

Wish I could offer more, but the variety & differential's of DDR, suitable mobo's, chipsets, etc isnt easily encompassed in a forum post. And, at this time, none of my systems are using DDR memory, still, I am maxed out on the SDRAM my mobo's will utilize

[The Finman]

Cool...thanks 45semi! [img]smile.gif[/img]

[DoctorDoom]

My box is using PC2700 DDR (512 megs) on an Asus A7V333 mobo w/XP2000+. It's more than fast enough for anything I'm likely to do with it.

RDRAM is thusfar used for Pentium-based mobos, and it's not overly popular because of the higher prices for an equivalent capacity. There are a few mobos which take both RDRAM and SDRAM.

There's still life left in SDRAM.

One consideration for jacking up clock speeds is that circuit board run lengths become significant. The speed of electricity in a conductor is approximately the speed of light. That sounds like a rip-roaring rate until we start factoring in frequency.

At 1 GHz, the "period" or wavelength of the wave is one nanosecond. In that length of time, light travels about 11.8". At higher frequencies, the wavelength becomes shorter. Given that the physical length of a conductor run on a PC board could amount to several inches, there's a good chance of timing errors at extremely high clock rates due to propogation delays.

As such, the maximum clock rate for memory is limited by the length of the board runs that connect it to the system, the prop delays in the controller chips and the RAM itself, etcetera.

This same issue is one reason why I'm skeptical about this mad race for multi-GHz CPUs. They might be able to process at 10 GHz, but the rest of the system can't support that speed. In effect, it's a Lamborghini Countach being driven in city traffic.

System speed is nothing but a "Gee whiz!" factor in 99% of the computers in the real world.

[45semi]

Why my next systems just might be Pentiums instead of AMD's...
but to be honest, this test occured almost two years ago, ...alas, I havent read anything to suggest heat protection built into AMD's in the interim.
But it still is signifigant due to the many still working systems using these processors!
(If anyone knows differently, please point me to it!)

You need to go far back in time to remember a CPU microprocessor that was able to operate completely without a heat sink. Intel's first Pentium CPUs were already producing a considerable amount of heat, but the specifications allowed operation without any special heat removal. A little bit later processors required at least a passive heat sink for trouble free operation. However, for the last three years is has become state of the art that a CPU requires a heat sink as well as a fan that ensures reasonable air flow through the cooling fins.

The reason is quite simple. The high clock frequencies of today's processors, which are nowadays measured in Giga Hertz and not in Mega Hertz anymore, lead to a heat dissipation of 50-80 W. To remove this huge amount of thermal energy from a CPU you require large and highly sophisticated solutions. Those heat removal solutions need to be attached to the microprocessor extremely closely. Once the heat conducting connection to the CPU is lost, the heat cannot be dissipated anymore and the processor overheats, which leads to different scenarios.

Common Reasons For Heat Sink Failure

Not everybody assembles his PC by his own. Some people wouldn't even contemplate opening their PC case to take a look inside. Those PC users are typically buying complete systems, either in a big retail shop, or from an online mail order company. In both cases the PC needs to be transported to your doorstep, either by yourself or by a delivery service. It is no rare occasion that the processor heat sink fell off while the system was in transport. The result is a black screen when the system is started for the fist time. In some cases this problem can easily be solved with a simple re-fixation of the heat sink, but often enough the processor did not survive this first doomed run of the system. A replacement of the processor is required, which can cause a lot of hassle, should the retailer not be willing to do this replacement free of charge.

Computer enthusiasts are of course not afraid to fiddle around with their computer system. Many of them are also trying to squeeze some more performance out of their processors via overclocking. Those folks are fully aware of the fact that the cooling of the CPU is extremely important. Therefore they often invest in expensive huge heat sink solutions. Those large heat sinks have one big problem however. They are often extremely heavy as well. If the mounting mechanism for those devices is not very sturdy, the heavy heat sink can still fall off because the mounting mechanism breaks off.

The most common problem with today's heat sinks however is fan failure. A lot of the cheaper heat sink/fan combinations use very cheap bearings for the fan spindle. Together with the dust that collects in the fan the worn out bearing suddenly stops the rotation of the fan. The airflow around the heat sink drops down to almost zero, so that it doesn't take long until the processor has reached a temperature that won't allow reliable operation anymore. In the worst case the CPU takes damage.

We wanted to find out how well processors from AMD and Intel are able to cope with the worst-case scenario - a sudden complete removal of the heat sink. We tested an Intel Pentium 4 2 GHz, a Pentium III 1 GHz, an AMD Athlon 1.4 GHz with the Thunderbird core and an AthlonMP 1.2 GHz, which comes with the Palomino core, as also used in the AthlonXP processors.

The Pentium 4 ran in a Socket423 motherboard with i850 chipset, the Asus P4T. Pentium III was tested using a Socket370 motherboard with i815EP chipset, the Asus CUSL2. For both Athlon processors we used a Socket468 motherboard with VIA KT266 chipset from Siemens, the D1289, which is already equipped with the logic that handles the thermal diode of 'Palomino', because it was directly developed for this processor.

We booted up those systems normally, started Quake III Arena, running the NV15-demo, and then removed the heat sink.

Intel Pentium 4 2 GHz - System Slowdown:
After the removal of the heat sink, Quake 3 Arena is slowing down significantly, but the system remains fully operational. The surface temperature of the Pentium 4 processor is a mere 29 degrees Celsius or 84 degrees Fahrenheit. After we put the heat sink back in place, the system performance went up to the original level. This shows that Pentium 4 has an excellent thermal design. The processor does not take any damage and you are not even losing data, because the system remains operational.

The Pentium 4 core comes equipped with a thermal monitoring unit that permanently checks the temperature. As soon as the core temperature has reached a certain trigger value, the thermal unit throttles down the clock of Pentium 4 until a safe temperature has been reached.

This solution is clearly commendable and proves that Intel's idea of equipping Pentium 4 with clock-throttling was far from a bad idea, as some sources want to make us believe. It is pretty much impossible to 'fry' a Pentium 4 processor. Additionally, Pentium 4 remains operational even once the thermal catastrophe took place and the heat sink fell off.

It also needs to be said that Intel's heat sink design specs leave hardly any room for an accident in which the heat sink comes off. The method that keeps the Pentium 4 heat sink in place is a very sturdy solution.

Intel Pentium III 1 GHz - System Hang / CPU Alive :
Our next candidate is Intel's Pentium III processor. A few seconds after the heat sink was removed the system hangs. However, once you replace the heat sink and reboot the system, Pentium III will greet you healthily up and running. Intel's older processor is also equipped with a thermal diode and a thermal monitoring unit. This unit is not as advanced as what we find in Pentium 4, but it ensures that Pentium III stops operating as soon as a certain trigger temperature has been reached. Your system may hang and you might lose data, but the hardware of your system does not take any damage. Intel has been equipping its processors with this simpler thermal protection for more than two years.

AMD Athlon 1.4 GHz (Thunderbird) - System Crash / Death Of Processor

Test-CPU number three is AMD's Athlon 1400 processor, which comes equipped with the 'Thunderbird' core that was introduced in June 2000.

The removal of the heat sink proves to be fatal. In less than a second Athlon 1400 dies the heat death. It doesn't take long and the core reaches a temperature of extremely hefty 370 degrees Celsius / 698 degrees Fahrenheit. If the user of the Athlon system doesn't turn off his box immediately, the motherboard will be destroyed too. There's even the risk of a fire.

AMD did not bless the Thunderbird core with ANY thermal protection whatsoever. If the heat sink should come off, the owner is facing a significant financial loss. He requires a new processor and possibly a new motherboard too. Athlon (Thunderbird) owners should make sure that the processor heat sink in their system is fixed 100% safely. The fact that the vast majority of heat sinks is only fixed to the little notches of SocketA doesn't help. We have seen several occasions when those notches finally broke under the weight of the heat sink.

Just like AMD's Athlon4 processors, AthlonMP is based on AMD's Palomino core, also used in the AthlonXP processor. This core comes equipped with a thermal diode that is required for Athlon4's clock throttling abilities. Unfortunately Palomino is still lacking a proper on-die thermal protection logic. A motherboard that doesn't read the thermal diode is unable to protect the Athlon processor from a heat death. We used a specific Palomino motherboard, Siemens' D1289 with VIA's KT266 chipset. Siemens assured us that the thermal protection circuitry is definitely working on their motherboard. So far we only know of Asus' A7V266 motherboard to include the thermal protection circuitry also. Those pictures cannot show you what happened by far as good as our test-lab video. A split second after the heat sink had been taken off the Palomino-Athlon, the system crashed. We then watched in horror as smoke clouds rose from the overheating core. The temperature measurement ensured us of what we had feared. No semiconductor survives almost 300 degrees Celsius / 580 degrees Fahrenheit. Palomino was dead.

We rushed to the telephone to confer with Siemens. The engineers assured us that what we had seen was for real. The thermal diode of Palomino is unable to react quickly enough. Only 1 degree/s is what the thermal diode is able to handle. That might be good enough for failing fans. A fallen off heat sink however will ensure a dead Athlon processor and possibly a damaged motherboard as well. What a serious disappointment!

We speculate that the thermal diode was basically included into the Palomino core because of Mobile Athlon4. A mobile processor requires a thermal protection and a clock throttling logic. However, AMD never designed this minimal thermal protection management to handle incidents like the complete failure of a heat sink.

http://www.tomshardware.com/cpu/20010917/index.html