Bandwidth and Memory Scaling
AMD includes the memory controller on their processors, but Intel continues to make the memory controller a part of the motherboard chipset. While the on-chip memory controller is theoretically superior, Intel manages to continue to improve memory bandwidth on their new chipsets. You have already seen in DDR3 vs. DDR2 and Intel P35 Memory Performance: A Closer Look that the new P35 improves memory bandwidth 16% to 18% compared to the same speed and memory timings on P965.
Kingston KHX11000D3LLK2 is the first to offer lower latency in DDR3 and this should further improve memory bandwidth. We compared Standard or Buffered bandwidth on the P965 running DDR2, the new P35 running DDR2, and the new P35 running Kingston DDR3-1375.
At DDR3-800 the Kingston manages stable 5-4-3-10 timings at 1.75V. Based on current motherboards and JEDEC standards the fastest available DDR3 timings are 5-3-3, so the Kingston DDR3-1375 is very close to the theoretical limit at DDR3-800. With the improved timings DDR3 is about 3% faster than DDR3 at 6-6-6 timings. Fast DDR2 on the P35 supporting DDR2 is the widest buffered bandwidth, but the difference between P35 DDR2 3-3-3 and P35 DDR3 5-4-3 is less than 2%. All P35 results, even the lower 6-6-6 timings, exhibit higher bandwidth than P965 at DDR2 3-3-3.
At both 800 and 1066, Kingston DDR3-1375 bandwidth is very close to the bandwidth of fast DDR2 on the P35. Again, all P35 results, even slower DDR3, are faster than P965 with fast DDR2 memory. At 1333 the 7-7-6-15 timings improve bandwidth by around 2.5%, and using lower latency allows the Kingston to run as fast as DDR3-1520 at 8-8-8-22 timings. However, the best bandwidth was achieved at faster timings and slightly slower speed. As shown above, the Kingston managed 1500 speed at 7-7-7-15 timings where standard buffered bandwidth is almost 7500 MB/s.
We also test memory with buffering schemes like MMX, SSE, SSE2, SSE3, etc, turned off. While these features do provide apparent improved bandwidth, the unbuffered bandwidth tends to correlate better with actual gaming and application performance. Unbuffered performance does not always follow the patterns of buffered memory performance.
Unbuffered results show the same basic pattern as buffered results in this case, although the domination of P35 in bandwidth performance is not as pervasive. At 800 and 1066 speeds, best bandwidth is with fast DDR2 on the P35 chipset, next is this Kingston DDR3-1375, then Fast DDR2 on P965, and last is slower DDR3. Unbuffered bandwidth is a good mirror of real-world performance, and this is what we expect in gaming tests. It is interesting that the lower latency Kingston has now passed DDR2 on the P965 and is nearly the equal in unbuffered bandwidth to fast DDR2 on P35.
Of course DDR2 could not do the 1333 speed, so the higher speeds of 1333 and 1500+ are the domain of DDR3 and here the Kingston memory shows its true capabilities. Lower latency DDR3 appears to be able to close any gaps that might exist in the overlap speeds of 800 and 1066.
AMD includes the memory controller on their processors, but Intel continues to make the memory controller a part of the motherboard chipset. While the on-chip memory controller is theoretically superior, Intel manages to continue to improve memory bandwidth on their new chipsets. You have already seen in DDR3 vs. DDR2 and Intel P35 Memory Performance: A Closer Look that the new P35 improves memory bandwidth 16% to 18% compared to the same speed and memory timings on P965.
Kingston KHX11000D3LLK2 is the first to offer lower latency in DDR3 and this should further improve memory bandwidth. We compared Standard or Buffered bandwidth on the P965 running DDR2, the new P35 running DDR2, and the new P35 running Kingston DDR3-1375.
| Standard (Buffered) Sandra XI.SP2 Memory Bandwidth - 2.66GHz | ||||
| Memory | 800 | 1066 | 1333 | 1520 (380x7) |
| Kingston DDR3-1333 KHX11000D3LLK2 |
6341 5-4-3-10 1.75V |
6736 6-5-5-12 1.7V |
6928 7-7-6-15 1.7V |
7329 8-8-8-22 1.8V |
| Corsair DDR3-1066 CM3X1024-1066C7 |
6156 6-6-6-15 1.5V |
6613 7-7-7-20 1.5V |
6757 9-9-9-25 1.5V |
- |
| DDR2 - P35 Corsair Dominator |
6456 3-3-3-9 2.25V |
6811 4-4-3-11 2.3V |
- | - |
| DDR2 - P965 (10x266) Corsair Dominator |
5531 3-3-3-9 2.25V |
5782 4-4-3-11 2.3V |
- | - |
At DDR3-800 the Kingston manages stable 5-4-3-10 timings at 1.75V. Based on current motherboards and JEDEC standards the fastest available DDR3 timings are 5-3-3, so the Kingston DDR3-1375 is very close to the theoretical limit at DDR3-800. With the improved timings DDR3 is about 3% faster than DDR3 at 6-6-6 timings. Fast DDR2 on the P35 supporting DDR2 is the widest buffered bandwidth, but the difference between P35 DDR2 3-3-3 and P35 DDR3 5-4-3 is less than 2%. All P35 results, even the lower 6-6-6 timings, exhibit higher bandwidth than P965 at DDR2 3-3-3.
At both 800 and 1066, Kingston DDR3-1375 bandwidth is very close to the bandwidth of fast DDR2 on the P35. Again, all P35 results, even slower DDR3, are faster than P965 with fast DDR2 memory. At 1333 the 7-7-6-15 timings improve bandwidth by around 2.5%, and using lower latency allows the Kingston to run as fast as DDR3-1520 at 8-8-8-22 timings. However, the best bandwidth was achieved at faster timings and slightly slower speed. As shown above, the Kingston managed 1500 speed at 7-7-7-15 timings where standard buffered bandwidth is almost 7500 MB/s.
We also test memory with buffering schemes like MMX, SSE, SSE2, SSE3, etc, turned off. While these features do provide apparent improved bandwidth, the unbuffered bandwidth tends to correlate better with actual gaming and application performance. Unbuffered performance does not always follow the patterns of buffered memory performance.
| Unbuffered Sandra XI.SP2 Memory Bandwidth - 2.66GHz | ||||
| Memory | 800 | 1066 | 1333 | 1520 (380x7) |
| Kingston DDR3-1333 KHX11000D3LLK2 |
4411 5-4-3-10 1.75V |
4761 6-5-5-12 1.7V |
4936 7-7-6-15 1.7V |
5172 8-8-8-22 1.8V |
| Corsair DDR3-1066 CM3X1024-1066C7 |
4098 6-6-6-15 1.5V |
4547 7-7-7-20 1.5V |
4702 9-9-9-25 1.5V |
- |
| DDR2 - P35 Corsair Dominator |
4536 3-3-3-9 2.25V |
4926 4-4-3-11 2.3V |
- | - |
| DDR2 - P965 (10x266) Corsair Dominator |
4226 3-3-3-9 2.25V |
4608 4-4-3-11 2.3V |
- | - |
Unbuffered results show the same basic pattern as buffered results in this case, although the domination of P35 in bandwidth performance is not as pervasive. At 800 and 1066 speeds, best bandwidth is with fast DDR2 on the P35 chipset, next is this Kingston DDR3-1375, then Fast DDR2 on P965, and last is slower DDR3. Unbuffered bandwidth is a good mirror of real-world performance, and this is what we expect in gaming tests. It is interesting that the lower latency Kingston has now passed DDR2 on the P965 and is nearly the equal in unbuffered bandwidth to fast DDR2 on P35.
Of course DDR2 could not do the 1333 speed, so the higher speeds of 1333 and 1500+ are the domain of DDR3 and here the Kingston memory shows its true capabilities. Lower latency DDR3 appears to be able to close any gaps that might exist in the overlap speeds of 800 and 1066.
Facebook
Twitter
RSS
42 Comments
View All Comments
Wesley Fink - Thursday, May 24, 2007 - link
The Asus P5K3 Deluxe motherboard allows DDR3 to be adjusted to 2.2V in .05V increments from the stock voltage of 1.5V. We ran voltages as high as 1.8V in this review, as Kingston specifies the memory at 1.7V. We gained nothing at voltages higher than 1.8V so we did not use them for testing.You seem to forget that enthusiast memory makers often specify higher than stock voltage for modules and they warrant the memory running at those higher specified voltages.
We do agree you should be careful with higher voltages on memory, but when manufacturers warrant products at higher voltage we are a bit less concerned.
TA152H - Friday, May 25, 2007 - link
Wesley,You can warrant something all you want, but that doesn't mean running over spec doesn't shorten the lifespan of a product. It will, without a doubt. The only question is, will it lower it significantly enough to matter, meaning during the useful lifespan of the product. Probably not, if they warrant it, but it depends on how long you keep it. Since only the kiddies will buy this junk, and they will replace the machine when the next alien invasion comes from Zargon, in higher resolution, it's probably OK. I doubt any serious machines like servers will have this sub-standard memory.
But, do they warrant it against additional power use? Do they pay for the electricity it takes? I don't think so. Do they warrant your motherboard against the additional heat? Do they give you additional fans to cool them? Do they pay for the electricity for those fans? Heat and electricity is a big problem, and even if they warranty their part, it stresses other parts too. And just because they'll warranty something doesn't really help that much if it breaks; the big loss isn't the part, but the down time. Do you really think they KNOW how long this part will last anyway? It's not like they can test it for 5 years and say it lasted that long. It's a best guess. The only certainty is they are shortening the life span.
So, a warranty doesn't cover everything, and there is always a price for running overspec, but that's not even my real point. I remember buying some memory from Kingston, and they had specs listed on it. It was for a mini-ITX, and it didn't have the crazy voltages available (why would it, the whole point was to save power and noise?). Of course, I see 2-2-2-5 and assume, naturally, that this is the timing for it will run at, at spec. Except the voltage you need for this is higher, and it's entirely misleading. I returned it of course, after yelling at them, and am still annoyed that these companies help make a standard, and then disregard it. I mean, if you want to run memory at 2.2 or 2.3 volts, put that into the standard. And it's not like you can say they find out quite a bit later that the standard wasn't realistic. Kingston is breaking it right after it's been made! So why didn't they say in the meeting, let's create the spec for 1.7 volts? Or, create a range. It's absurd they create a spec and the first memory out breaks it. Of course, the other memory makers will do this too, but one of the points of the memory was low power use, so it's a bit conflicting. Also, as they go to PC-1600, and it naturally sucks more juice, how high can you really go with the voltage without creating an enormous amount of heat that can't be ignored? Naturally, they'll be going beyond PC-1600 at some point even though that's the spec, and it'll just get worse. So, being able to make memory with proper voltages will become more and more important.
bldckstark - Friday, May 25, 2007 - link
Ummm.. HyperX memory is not marketed to businesses. It is marketed to enthusiasts. Businesses keep using the standard parts, and enthusiasts keep using high performance parts. This would be the same reason that companies don't buy Corvettes for their salesmen to drive. It is not a reasonable business decision for several reasons, some of those being initial cost, maintenance costs, and normal usage costs like gas.How come you don't buy servers from Alienware for your company? Do you buy EE processors for your companies desktops? Your argument is similar to that of not having a space program, because we can't use rocket engines on jets.
TA152H - Friday, May 25, 2007 - link
Do you have a reading comprehension problem?I was saying the same thing. I guess most people are too simple to realize that even though, in a general sense, you are against something, you can make a point for why it exists. I pointed out that only kiddies will buy this memory, and it won't be used for servers, so it's not that bad. You couldn't understand that?
But the main thing is, why make a standard when you're going to break it.
Your remark about rocket engines on jets is purely idiotic. It's a terrible analogy, and makes no point at all. But, just so you know, there were in fact rocket propelled airplanes (German ME-163), but jets are a competing technology, you have one or the other. But again, you missed my point, because you naturally assumed everything I was saying was against this over-voltaged memory, but I was giving both sides. I still don't like it though. They should have made the spec 1.7volts, or whatever, if they fully intended to make it at that voltage, which clearly they did.
menting - Friday, May 25, 2007 - link
"But the main thing is, why make a standard when you're going to break it. "is same as why set a speed limit when people are going to speed.
standards are there just for a standard. It doesn't say if they are prohibited from doing more. If they sell memory and say it conforms to JEDEC standards, then it means they can run at specced speeds at the specced voltages. They could go faster at higher voltages if they want. If they dont even say they conform to JEDEC standards, they can spec whatever they want and it's up to the user to decide if they want to buy memory that runs at the manufacturer's settings.
TA152H - Friday, May 25, 2007 - link
Another bad analogy.You can get a ticket for going over the speed limit and get fined. It's proscribed. You do it at your own risk and it's illegal. That's good?
If they intended to go at higher voltages, why not spec it at that? Or create a range? Why create a spec if everyone breaks it? My big problem is how it's advertised, it's not so clear that the timings are for grossly inflated voltages. You have to look, and unless you know to and not make an assumption that memory is made according the specification, you can be fooled. You don't make a standard to break it, that's just plain asinine. You make a standard for conformity, that's the point of standards.
rjm55 - Friday, May 25, 2007 - link
You seem obsessed that Kingston sold you some memory sticks in the past that were rated at higher than JEDEC voltage. You learned that pretty much everyone in the memory industry does this and most people think nothing of it. You really need to get on with your life or seek professional help.PrinceGaz - Sunday, May 27, 2007 - link
He may perhaps have a point in that the DDR3 standard has only just arrived, and already modules are arriving which are intended to be used well above the rated voltage.I run my memory above voltage like most of us, but when a new standard arrives and the recommended voltage has already been exceeded by over 13% almost immediately, it makes a mockery of it.
yyrkoon - Friday, May 25, 2007 - link
Well, I kind of agree both ways here, but would err on the side of staying with the specification. Just like companies like Asrock releasing a motherboard with supposed SATAII ports, but they do not support NCQ, which is part of the SATAII spec!Granted this situation here is a bit different, they added to the spec, but not only did they add voltage capability, they added potetnial heat/overvoltage as well. This has impact on more than just the memory, this could adversely effect a motherbaord as well, and possibly even a PSU(over time).
bobsmith1492 - Thursday, May 24, 2007 - link
Why did you not put up some numbers at this speed for comparison? Granted the CPU may be running 35MHz slower, but might the RAM be enough to make up for it? At the very least, the bandwidth numbers would be impressive... assuming the latency affects the bandwidth?