DDR3-2000+ Memory Kits - Fast but Flawed
by Rajinder Gill on July 8, 2009 12:15 AM EST- Posted in
- Memory
A few 3D results...
Let’s take a quick look as to how 3D benchmarks such as 3DMark Vantage, 3D Mark 06 and 3D Mark 05 are believed to be somewhat affected by memory bandwidth. Vista 64 allows us to see how the memory voltages we utilized for Super Pi apply to a 64 bit OS when all 6GB of memory is fully mapped.
No major gains here in 3DMark Vantage as we have a 100-point difference between CAS6 1836MHz and CAS8 2294MHz.. Voltage wise, we see the need for an elevated idle voltage in Vista due to increased voltage droop under load with 6GB fully mapped. The actual load Vdimm in these benchmarks is exactly the same as what we used for Super Pi 32m under the XP operating system.
3D Mark 06 scores more consistently run to run than Vantage and seems to prefer the tighter CAS latency of 6, giving a 400 point boost.
Switching over to XP for 3D Mark 05 we find that the GPU benches are not overly sensitive to CAS or memory bandwidth. The best overall scores come when using the 20X Uncore multiplier on our 920 CPU (one below the CPU core multiplier). Using higher Uncore multipliers than the CPU core multiplier actually resulted in lower average results for 3D Mark 05, due to the additional hold times asserted by the IMC to the clock crossing process. This performance offset can be seen in BIOS via the increased Round Trip Latency values. In essence you get faster L3 cache performance while the IMC slows down memory bus transfer times to levy the increased performance on the other side of the bus; giving up one for the other. Bottom line here is that higher Uncore frequencies don't really offer anything exciting in any benchmark that reads or writes more data than the CPU cache can hold. As an example running a 23X Uncore multiplier results in a 15ns shift for the worse in Round Trip Latency which cannot be reclaimed with the current range of settings availble withn the EVGA BIOS. Higher Uncore frequencies may appear attractive at first approximation, but they're not always best in every scenario.
Using CAS 6 also resulted in the GPU tests of 3D Mark 05 failing to pass at 229 BCLK regardless of voltages, while running 2294 MHZ CAS 8 with the same Uncore multiplier passed without a glitch.
Facebook
Twitter
RSS
13 Comments
View All Comments
justniz - Wednesday, July 22, 2009 - link
DONT BUY CORSAIR PRODUCTS.Their warranty is a scam and their customer support is insultingly useless.
I RMA'd $680 worth of memory that was advertised as fully covered with a lifetime waranty. Now they refuse to offer anything other than $90 worth of much lower performing memory. Basically I paid for a Ferrari under lifetime warranty and they will only offer a Daewoo as a replacement.
All emails to Corsair are either ignored or take days for a response which is mostly just an automated or canned reply that contains no information and is of no actual help. Their customer phone support takes ages to get through, then you speak to someone who hardly speaks English, is technically clueless, and has no authority to make any actual decisions.
Their phone and email system is designed to completely prevent you from getting to anyone that is actually empowered enough to help you.
Worst experience ever. I will never buy any Corsair products again.
My only recourse against them is to share my experience with you, the world, so that you won't make the same mistake I did.
DONT BUY CORSAIR PRODUCTS.
BlackDragon24 - Thursday, July 9, 2009 - link
I'm really glad you took the time to write this article Raja. I went thru two of the top end OCZ blade elpida kits in three days. First kit fried in the first hyperpi32M run only running 1.60 vdimm and 1600 memory speed. Second kit failed after three days while running only 1800mhz with 1.63 vdimm. Third set seems to be running like a champ so far (about 3 months), but I am afraid it won't last long.Luckily OCZ was a champ about the whole thing and took care of me very quickly. I would expect as much with $450 memory.
Hopefully if these fail they'll have a comparable replacement part.
bobsmith1492 - Wednesday, July 8, 2009 - link
I would recommend graphing the test results vs. CAS value and clock speed in a 3D graph where CAS is one axis, clock speed is the other axis, and the vertical axis is the test results. That could be a good visual aid to the first block of numbers of test results.Other combinations, like processor clock and RAM clock might help also - of course the non-graphed variables would have to remain constant.
mataichi - Wednesday, July 8, 2009 - link
I bought some Kingston HyperX T1, and they crapped out after a few weeks. I RMA'd them with Kingston and they said it would be about a month before they had any more in stock.Luckily they offered to refund the purchase price, so I took it.
kitfit1 - Wednesday, July 8, 2009 - link
Once all the others come into line and "pull" their Elpida as well, the question then is, will they replace ALL the Elpida ram in the wild.As no one knows if they have suspect ram (until it dies), the ram company's should do a re-call in the same way that car company's have to do. Then they can replace it with whatever comes after Elpida.
navilor - Wednesday, July 8, 2009 - link
According to Corsair:http://forum.corsair.com/forums/showthread.php?t=7...">http://forum.corsair.com/forums/showthread.php?t=7...
We have seen a number of reports across various forums about failures of modules (from Corsair as well as from other memory manufacturers) built with Elpida “Hyper” RAMs. Through lab testing, we have now been able to reproduce similar failures. We are continuing to test to determine the cause of these failures. Note that although a relatively small percentage of “Hyper” ICs appear to be affected, the rate of failure is not acceptable to Corsair or to our customers.
Due to these failures, we will no longer sell Hyper-based modules until the issue can be resolved. We have also have asked our retailers to return any modules they currently have on their shelves. Products impacted include TW3X4G1600C6GTF, TR3X6G1866C7GTF, TR3X6G2000C8GTF, TR3X3G2000C7GTF, and TR3X6G2000C7GTF. We are working on enhancing our manufacturing and testing process to be able to offer these parts again as soon as possible.
We continue to stand behind these modules 100% with our standard warranty, which can be found at http://www.corsair.com/warranty/default.aspx">http://www.corsair.com/warranty/default.aspx.
Rajinder Gill - Wednesday, July 8, 2009 - link
I think most automobile recalls are due to safety issues more than anything else. We can hope that if there is no such recall here, that the lifetime warranty that OCZ and Corsair offer on these modules will be honored should they happen to fail for users in the future.regards
Raja
JonnyDough - Wednesday, July 8, 2009 - link
The price/performance of these modules is so ludicrous that I find myself not even caring...Give me half a year and I'll leave your OC'd $5K+ system in the dust for half the price.
vol7ron - Wednesday, July 8, 2009 - link
That's because you're not one of the enthusiasts the article was written for. In 6 months, you will not leave a $5K system in the dust for only $2.5K. That's guaranteed.Now, if you meant that you could beat a $5K 2009-07-08 system in 2010-01-08 with $2.5K, then that's not really saying anything. In fact, I'm not sure 6 months is enough time these days that it would go down 1/2 price, I would imagine it'd be about 3/4 price in that time and probably a year to turn around; mainly because the bubble for performance enhancement is about to burst.
Even to say if your statement was true, you could do the same today with half the price of a system 6 mo. prior.
I guess I'm writing so much about this because I have a problem with your statement in general; especially considering the facts made and context of the article, which you're arguably not the target audience for.
Dreamwalker - Wednesday, July 8, 2009 - link
Just read at guru3d.comhttp://www.guru3d.com/news/corsair-pulls-dominator...">http://www.guru3d.com/news/corsair-pulls-dominator...