Date: July 19, 2006
Author(s): Rob Williams
If you are thinking of jumping onto the PC2-8000 bandwagon, there is surprisingly much more choice now than there was just two months ago. Like other PC2-8000 kits we have tested, A-Datas is seriously fast, but is it worth your hard earned cash?
Next up on the Techgage workbench is some blazing fast memory from A-Data. This is the first A-Data review we’ve done, so I was looking forward to discovering the quality of their products. Though the company is not that popular over on these shores, at least in the enthusiast circle, they are a rather large company that owns around 6% of the DRAM market. I am unsure of what the market is like in Taiwan, but with a large market share as such, it would have to be quite massive.
Quick introduction aside, I am taking a look at their top of the line DDR2 kit, the Vitesta PC2-8000 2GB with 5-5-5-15 timings. Actually, these are triple spec’d modules, but I will get into that shortly. Before we get into the technical side of things, let’s skim a few shots of the modules themselves and their packaging.
Many memory manufacturers like to display their pride and joy in a clear plastic blister pack, and A-Data is no different. What is different here though, is how it’s designed. Instead of a single piece of plastic that opens like a book, this packaging is actually two separate plastic pieces that fit inside of each other. This makes it quite easy to remove the modules once you receive them.
The heatspreaders used are somewhat similar to what Mushkin uses on their Redline series, although the shade of red on the Vitesta is lighter. The front and back are held on with a simple clip, although it will still prove difficult to remove due to the thermal paste used. It’s never wise to remove spreaders off of FGBA modules anyway, unless you don’t mind risking your investment.
There is a silver sticker on both the front and back of the modules. The front includes the speed and timings, while the back is for the serial number and model code.
Ok, enough drooling. Time to get into the tech specs.
If you haven’t caught on by now… these are PC2-8000 modules, also known as DDR2-1000. They are classified as “Extreme Edition” for their speed, and probably their overclocking potential. Though, the stock 5-5-5-15 timings are nothing really to be called ‘extreme’. Upon asking around, these sticks seem to use Micron D9GKX chips, which are binned at 400MHz 2.5ns, CL5. These are some of the most expensive, if not the most expensive, chips that Micron has to offer. Why A-Data did not go with the equally performing, but less expensive D9GMH or similar chips is unknown. Regardless, I had a blast using the GKX chips in the Super Talent PC6400 kit, so I couldn’t wait to get down and dirty with my overclocking business with these.
One thing to note also is that these sticks are not dual spec’d, but triple spec’d! This means that they should work out of the box at any of the listed settings. However, who is going to fork out ~$400 for a kit of memory and run it at DDR2-667? Crazy people, that’s who. Out of the box though, they will function at DDR2-667, with 4-4-4-8 timings. It’s up to you to manually change the settings to allow for the DDR2-1000 speeds.
Though there are 1066MHz modules on the market, 1000MHz is still really the cream of the crop when it comes to DDR2. So how do these overclock?
Since these modules use the D9GKX chips, I knew I was in for a similar experience to what I had with the Super Talent sticks. That proved quite true.
If you check back to that review, you will see that these results are incredibly similar. Though I was able to push past 520MHz, I would error instantly in MemTest, so I left them alone. DDR2-1040 with 4-4-4-8 timings proved stable through a 32Million Super Pi run.
As you can see by the results also, these modules seem to love voltage. All of the results you see are with the lowest voltage required. While some settings seemed to function fine with lower voltage, MemTest would tell me different. One thing I have to question is as to why they don’t ship these with 4-4-4 stock timings. On both GKX kits I have used, they were ultra stable with 4-4-4 with no sign of any error. If I was a company who was responsible for buying these very expensive chips, I would be sure to make the modules look as appealing as possible, which 4-4-4 would help. Though on the other side of the coin, if they are indeed rated otherwise, it could drive up prices for consumers.
Throughout all of our benchmarks regardless of what we are reviewing, testing is done in a clean and stand-alone version of Windows XP Professional with SP2. Prior to testing, these conditions are met:
If you are interested in using the same benchmarks as us, feel free to visit the developers website:
The testing rig used for today’s benchmarking is as follows:
There are a few things to mention before checking out the results. First, realize that all testing is performed on a rather weak Intel. The 820 D is on a locked multiplier, so I am limited to how high my FSB can go before I have an instable system. In my case, my max stable FSB is 280. Also, bandwidth results will be -far- different here than what you see coming out of AM2 and Conroe. I will have a comparison with a similar PC2-8000 kit later on so you can see some differences on a single system.
First up is one of my personal favorites, Sandra 2007. If you are still using version 2005, don’t worry as the results are the same for the memory tests. DDR2-933 on my system always proves to take the crown of any CPU or memory benchmarks, due to the 280 FSB. There’s no doubt that this is the case here. Overall, these scores are quite similar to what I have achieved with other PC2-8000 kits. However, this one surpasses the performance of the OCZ VX2 by ~50MB/s at the DDR2-933 setting.
The unbuffered are no different really, with DDR2-933 (280FSB) taking the crown again. Note the differences at the stock speed, going from 5-5-5 to 4-4-4 though. Over 125MB/s difference for a tweak that can hardly even be considered an overclock.
Ahh, Everest. This is another great benchmark for simple Read/Write and Latency tests. The only downside to this tool is that Lavalys tends to change benchmark algorithms with each new release, so even micro-version updates should not be compared to one another. After testing though, I found versions 2.8 and 3.0 to have identical memory results, however CPU and FPU were varied.
At my optimum setting, OCZs VX2 beat out the Vitesta by a seriously low margin… around 30MB/s, or 0.5%. Great results overall though.
When it comes to latency, 60.0ns is usually a goal worth aiming for, though we didn’t really come close here. If my board could support more than 2.4vdimm, I could easily see this memory hitting 61.0ns – 62.0ns at 500MHz with tighter timings.
3D Mark 01 is considered outdated for a good reason. It is. One great thing it’s still good for though, is seeing how your CPU overclock scales. I like to pay more attention to PC Mark though, since it’s benchmark is completely memory specific. Looking at the DDR2-1000 reports, the scaling is quite interesting.
I am impressed that adjusting the tRP made any real affect, but it actually made a 40 point difference in the score… the same difference as going from 5-5-5 to 4-4-4.
Super Pi, like 3D Mark 01 is good for seeing how your CPU overclock scales. The primary benefit is that it does just that… but quickly, especially if you only run the 1 Million method. Memory plays a big part in the results though. The tighter the timings, the faster the roundtrip for the data between your CPU and memory.
All of the previous results have been done using only the A-Data modules. For a comparison test, I took the recently reviewed VX2 to see how they fared. Both sets of modules used the exact same settings. The FSB stayed at a constant 250… only the memory frequency was increased. Here is a quick list of the used settings and voltages required:
Interestingly enough, the A-Data modules cleaned up the DDR2-667 results, but were no match for the VX2 at the other settings. Why I find this somewhat humorous is because the Vitesta use more expensive chips, but seem to lack in the greater speeds.
As mentioned in the intro, A-Data is not really a common name thrown around when discussing memory with other enthusiasts. While they are popular in other Countries, it’s apparent that they want a taste of the North American market, which is clearly evidenced by these modules. They chose to use Microns best offering in terms of speed, which will please everyone from the laid-back computer guy to the hardcore overclocker. Without a doubt, these are very fast modules.
Do I recommend the Vitesta PC2-8000? I don’t even need to think before I answer with a wholehearted “Yes!”. There are a couple predicaments that may hold you back from an immediate purchase though. First, I have no idea where you can yet purchase them. I am awaiting a reply from A-Data on the matter, and once I hear a reply I will add a link to the end of this article. The second problem is the fact that there are -many- other enthusiast modules on the market. Not only PC2-8000 kits, but ones with the same chips. So you will really have to shop around if you want the best price.
If these modules hit an e-tailer for around $400, it will be considered a great buy. After a quick skim at a popular e-tailer, the lowest PC2-8000 modules I found were $395, and are of the ‘hit or miss with overclocking’ variety. Considering these modules use some of the best overclocking chips on the market, it’s hard to go wrong. If you want the “ultimate” module out right now though, you will need to spring up to $500 for Corsairs 6400C3… which are rarely in stock lately.
After it’s all said and done, I am awarding the Vitesta PC2-8000 a well deserved 9 out of 10. Because the modules do not appear to be released yet, I am not going to hold back any of the score due to the price. If it launches competitively priced to equivalent modules out there, then the score is right on the ‘money’. If it happens to launch for much higher, it would deserve an 8 out of 10.
If you have a comment you wish to make on this review, feel free to head on into our forums! There is no need to register in order to reply to such threads.
Copyright © 2005-2020 Techgage Networks Inc. - All Rights Reserved.