Date: April 3, 2006
Author(s): Rob Williams
So you think you have fast DDR2? Then you better check out the GeIL PC2-8000 which happens to be the fastest we’ve ever seen, and the benchmarks prove it! It’s also probably the most expensive ‘value’ ram you will set your eyes on, but is it worth the cash for the ultimate performance?
No matter what type of hardware you are talking about, there’s always a low, mid and high end version. Oh, and then there is extreme high-end. In the DDR2 market, PC2-8000 speeds are awarded a spot in the extreme high-end catagory. Do the 1000MHz speeds really help though, with the loose 5-5-5 timings? Because we all know, that whereas 3-3-3 for DDR2-667 is a sweet spot, 5-5-5 for DDR2-1000 may actually end up being slower. That’s what we will find out today.
PC2-8000 speeds for DDR2 is not that common, because not too many modules are actually capable of such speeds even with a great amount of voltage. In fact, as far as I know, Crucial, OCZ and Patriot are the only two other companies that provide 2GB kits at these speeds. All three of these kits use 5-5-5 timings, so it seems that’s a real barrier. Let’s take a quick look at what we received.
GeIL packages all of their Value memory in the same packaging, which is a cardboard box with a small plastic blister pack inside. The modules are kept quite secure this way. The modules themselves are extremely lightweight. They use a brushed aluminum heat spreader which is what helps the modules be ultra light. This could prove useful when trying to keep the modules cool.
I did not open up the modules to see the chips, and I have not received a reply from GeIL yet as to which chips these modules use.
It may sound funny to even want to overclock top of the line memory, but what’s the fun of not bothering? Chances are you will need to overclock your CPU in order to even reach the stock DDR2-1000 speeds. 250FSB worked for me and I was ready to get cracking. After running a 10 hour MemTest run, I found the modules to be completely error free. In order to have them error free though, I had to use 2.3v as even 2.25v would prove to not be enough. That’s a fair amount of voltage but it’s to be expected with modules this fast.
Overclocking is not really a huge option here because DDR2-1000 is no doubt a breaking point to begin with. One thing I did want to try though was simple under clocks to see if we could score better benchmarking results. Here is a quick sum-up of stable overclocks I have found:
One option I was unable to try was 300FSB, because my CPU was unable to handle the overclock using the stock cooler. I have another cooler en route, but it did not arrive in time for this review. Once my CPU is able to hit 4.2GHz stable, I am confident that 300FSB with DDR2-1000 speeds will prove to be that fastest setting.
To test out the modules, I used a selection of popular and not-so popular tools. If you are interested in using any of the tools for yourself, you can download them from their respective sites:
SANDRA is one of the most popular benchmarking programs on the planet, and for good reason. The memory benchmark specifically is for the power user because you can configure exactly how you want the tests to be performed, with whatever options turned on or off you want. Because the CPU can have a lot of overall say in how fast your memory is, I run the memory through a buffered and unbuffered tests. Unbuffered tests stresses the memory directly instead of a combination of the memory and CPU.
Usually I expect to see stock results hover around 5000/5000 in buffered mode, but these GeIL’s blew past that to hit 6391/6387. With a minor under clock and tighter timings we managed to break the 7,000 barrier. That same under clock managed 4,663/4,681 unbuffered. Amazing stuff.
EVEREST is another great memory benchmarking tool to use. It not only will give you a READ speed result, but also a WRITE and Latency. Since Latency plays a large roll in how responsive your memory is, it proves to be a very worthy benchmark. I am using a slightly older version though, because the newer versions give exaggerated results. For instance, whereas you may get a 9,000 MB/s READ in version 2.21, the new versions will give you a result of 11,500 MB/s READ or something similar. Those results are far too different for comfort, so I am sticking to the older version for the testing.
Around this time last year, I can recall my goal to be able to hit 7.300 READ in EVEREST. At the time, that seemed like a sweet number to hit. Not even a year later, we can easily hit 8,400 READ. Even the stock read at 7,500 is a nice spot.
Let’s move on to a few more benches.
MemTest is not exactly a benchmark per se, but it if you run it after any overclock, it will tell you how much memory bandwidth you are pushing out. This is a great way to immediately see if you have actually increased performance or hurt it before logging back into Windows to do more testing.
This one pretty much speaks for itself..
PC Mark has a unique way of benchmarking memory. It will perform four varying tests of Read, Write, Copy and Latency and then spit out a final score based on those results. There are also some 3D Mark 01 scores for your consumption if you enjoy them, but I personally find them to be a waste of time, really.
This memory just keeps setting personal records here. 5,401 is the higher amount of PC Mark’s I have ever scored for the memory side of things. The Corsair PC2-6400 kit I reviewed last week came very close, but the GeIL clearly wants to remain ahead of the pack.
Sciencemark is a program I only recently found out about thanks to a friend [Thanks Mark!]. Like PC Mark, it can do various benchmark tests on the memory and spit out both the bandwidth score and overall Sciencemarks score. The Sciencemarks overall score doesn’t really tell you much, but it’s yet another number you can use to compare your overclocks to.
I have not previously posted any results from Sciencemark before, but I have tested out my other modules to get some comparisons. Again, the only thing to come close to these at Max overclock was the Corsair PC2-6400. They had results of 6293MB/s and 1460.84SM’s and were clocked exactly the same as the best results for the GeIL found below.
Let’s move right along to our conclusions.
There’s not much to be said about the performance of these modules really, the results speak for themselves. This is the fastest memory I have ever tested and it looks to stay that way for a while… or at least until I get my hands on another PC2-8000 kit ;)
Corsair recently announced a PC2-8500 kit, but sadly it only comes in at 1GB, so it looks like this GeIL has nothing to really worry about. Throughout all of our benchmarks, these modules proved to be faster than any other we have tested… and by a fair margin. The humorous thing is that the modules performed better under clocked, but there are two reasons for this. The first is the tighter timings, 4-4-4 did help quite a bit compared to the stock 5-5-5. The second reason is that the FSB was higher because I could lower the memory frequency. Chances are that even this OC could be beat if I can get my FSB up to 300, and as soon as a new cooler arrives, I will be testing that out right away.
If you consider yourself to be a hardcore overclocker, these modules will love you. The ASUS P5WD2-E tops out at 2.4 VDIMM, so I was unable to attempt 4-4-4 @ DDR2-1000. Throughout the overclocking tests though, these modules seemed to love any amount of extra voltage, and a full MemTest run seemed very close to being error free at 4-4-4 with 2.4v. I could image that 2.5v or 2.6v especially may allow for 4-4-4 at stock frequency. I will test all of this in due time..
So.. are these modules worth it? If you are not an overclocker at all, and have no desire to do so, then these are the fastest modules you can buy. ‘Nuff said. If you are an overclocker, then there may be more reason to check these out because the overclocked results are nothing short of amazing. I was truly impressed with these modules and they were a blast to play with. I look forward to getting hooked up with better cooling because I’m confident that these can be pushed even further than I could currently manage.
Because of the awesome performance and great overclock ability, I award these modules a well deserved 9/10. Now for the part that hurts… the price. Currently these modules retail for around $400, but I have seen them as high as $460. This is a high price, but it’s competition is no lower. PC2-8000 is the extreme high-end and you must be ready to open that wallet wide in order to partake. That being said, it’s funny to even call these “Value Modules” because of the high price, and I am unsure why GeIL didn’t release these in their Ultra line-up because this is the only PC2-8000 kit they currently have.
Copyright © 2005-2020 Techgage Networks Inc. - All Rights Reserved.