Memory Scaling on Core i7 - Is DDR3-1066 Really the Best Choice?
by Gary Key on June 24, 2009 9:00 AM EST- Posted in
- Memory
Memory Prelude
Intel designed Core i7’s performance around triple-channel mode and low latency DDR3-1066. In fact, with three channels of DDR3-1066 memory, the platform will provide up to 25.6GB/s of memory bandwidth and nearly 38.4GB/s at JDEC approved DDR3-1600 memory speeds. Intel figures the bandwidth numbers available at DDR3-1066 coupled with the on-die memory controller and relatively fast secondary cache speeds should provide enough performance to satisfy anyone. For the most part, they are right.
However, we are never satisfied so we set out to determine if applications we utilize everyday would benefit from increased bandwidth, improved latencies, or a combination of both. Obviously synthetic memory benchmarks improve greatly as you increase bandwidth and decrease latencies. The memory companies (with help from us at times) tout their latest and greatest memory products with these benchmarks while they are staples on the benchmarking circuit.
We were discussing our initial overclocking results with a couple of BIOS engineers and came away surprised. We learned they typically tune memory performance utilizing a combination of Sandra/Everest and Super Pi. That did not shock us but it explains why time after time our various test systems would perform exceptionally well in the synthetic benchmarks and then fall flat on their collective faces when faced with heavy multitasking situations. We believe too much emphasis is placed on trying to win a benchmark that in no way reflects the true performance of applications or enhances the stability of the system.
That brings us to another story. We had planned to incorporate a full overclocking section in this article but our DDR3-1866 and DDR3-2000 kits based on the Elpida DJ1108BASE, err Hyper ICs, have been experiencing technical difficulties as of late. As in, the damn things are failing faster than we can keep replacements in the labs. This is not a vendor specific problem, nor does it seem to be a voltage problem either, as we have had modules from various suppliers fail at stock VDimm or at 1.8V when trying to push through to the DDR3-2300 level. We have asked the memory suppliers a few pointed questions and are waiting for answers. In the meantime, replacements are arriving shortly. We will follow up with additional DDR3-2000 overclocking scenarios in the near future.
Facebook
Twitter
RSS
47 Comments
View All Comments
Seikent - Wednesday, June 24, 2009 - link
I'm not very sure if it's relevant, but I missed a load times comparation. I know that the bottleneck there should be the hdd, but I still think that there can be a performance boost.deputc26 - Wednesday, June 24, 2009 - link
ave and min lines are mixed up.MadBoris - Wednesday, June 24, 2009 - link
I'll be considering upgrading in October at the same time I go from XP to Win 7.So this is good to know if/when I go Core I7.
I guess I can see how Winrar RAM workload sdtays high since it grabs the buffers of compressed data chunks and writes them to disk as fast as the HW permits, so bandwidth matters then.
While it looks like very few apps can saturate the bandwidth latency benefits/penalties are always having an effect as usual.
Maybe I missed it but I didn't see anywhere in the article that tried to explain the technical reasons "why" 2000 doesn't provide advantage over 1066.
I understand the differences of latency and bandwidth. Is it really because no software is using RAM workloads large enough to benefit from increased bandwidth (except compression) or is there another bottleneck in the subsystem or CPU that doesn't allow moving all the data the RAM is capable of?
vol7ron - Wednesday, June 24, 2009 - link
Your question is long, so i didn't read it all, but does bottom of pg2 answer:"That brings us to another story. We had planned to incorporate a full overclocking section in this article but our DDR3-1866 and DDR3-2000 kits based on the Elpida DJ1108BASE, err Hyper ICs, have been experiencing technical difficulties as of late."
They said some other stuff, but it seems like it wouldn't be right to post info on faulty chips.
TA152H - Wednesday, June 24, 2009 - link
I'd like to see a test between the crippled i5 memory controller with very fast memory, and the i7 with low cost 1333 Mhz memory. There's really no point in the 1066 memory, except for Dell, HP, etc... to throw in generic machines; it's not much cheaper than 1333 MHz, and the performance bump really seems to be biggest there. I think 1333 MHz (low latency) is a reasonable starting point for most people, the cost seems to warrant the performance. After that, you definitely see diminishing returns.It seems anyone buying an i5 with very expensive memory is probably a fool, but, a few benchmarks might be interesting to validate or invalidate that. Of course, the i5 might be better when released, so even then it wouldn't be proof.
Gary Key - Wednesday, June 24, 2009 - link
I wish I could show i5 numbers, but that ability is officially locked down now. I can say that our results today will not be that much different when i5 launches, low latency 1333 or possibly 1600 will satisfy just about everyone. :)strikeback03 - Thursday, June 25, 2009 - link
Of course, by the time you can share those numbers we will most likely have to specify whether we are talking about LGA-1366 i7 or LGA-1156 i7. Thanks Intel.kaoken - Wednesday, June 24, 2009 - link
I think there is a mistake with the farcry graph. The min and avg lines should be switched.hob196 - Thursday, June 25, 2009 - link
Looking closer it might be that you have the SLI min on there instead of the Non SLI min.halcyon - Wednesday, June 24, 2009 - link
It's so nice to see AT calling things as they are.This is why we come here.
Straight up honest talk from adults to adults, with very little marketing speech and numbers do most of the talking.
Excellent test round up, mucho kudos.