Crucial ballistix Tracer Red PC6400 4-4-4-12 4GB review

[K404]

Do you want candy?



Official specs and marketing:

BL2KIT25664AR804
• Module Size: 4GB kit (2GBx2)
• Package: Ballistix Tracer 240-pin DIMM (with LEDs)
• Feature: DDR2 PC2-6400
• Specs:
• DDR2 PC2-6400 4-4-4-12
• Unbuffered
• NON-ECC
• DDR2-800
• 2.2V
• 256Meg x 64

Are you ready to blow the hinges off your modules? Outfit your system with the new clipless, candy-apple red Ballistix™ Tracer™ modules from Crucial.
Our highest density Ballistix module yet, the Ballistix Tracer Red modules feature 2GB of 800MHz DDR2 goodness. Add to it the bright red heat spreaders, red PCB, red LED lights, plus a raised and diamond-cut logo and clipless design—our Crucial Ballistix Red 2GB DDR2-800 module or 4FG kit delivers outstanding results for PC performance enthusiasts everywhere.
For systems with dual-channel DDR2-based motherboards, Ballistix Tracer Red is ideal for users craving faster response time and improved reliability. It's a high-performance, low-latency memory option that's the perfect blend of sugar and spice—sweet looks on the outside, red-hot performance on the inside

_________________________________

The package and the contents:

Crucial continue their no-frills packaging with the Tracer Reds. These are also more minimalist than previous shipping boxes, being half the depth. Will good things come to those who don’t judge books by the cover?
[CENTER]








The no-frills packaging is in complete contrast to the contents which are a LOT more bling: Bright red PCB, bright red heatspreaders with polished aluminium-edges and raised logos.








The bling doesn’t stop there- heres some pictures of the RAM installed and powered up:







Pics of RAM installed and powered up

I suppose the visual aspect of RAM like this is fairly cut and dried- there it is. Beauty is in the eye of the beholder, so I will not comment. I do not think many people will buy this for chasing the highest possible RAM clocks or for aggressive benchmarking.
It should be popular with case modders and gamers, who expect excellent stability and reliability, which I associate with lower voltages. Because of this, most of the testing will be at voltages at which Crucial will honour warranties.


Going into detail.

The ICs are made using a 78nm process, which allows a higher density of cells to be packed into the same space whilst staying within an acceptable power consumption and heat envelope. The 64MB chip density of 2x1GB modules has doubled to 128MB to allow 2GB to be placed onto one stick.
Seeing as the numbers of cells that can be addressed has doubled, one would expect the maximum performance of these sticks to be slightly lower than older 1GB sticks. This theory has been hinted at by several members of assorted enthusiast forums (including this one.) I am entering this review with a fairly open mind. I believe that there is excellent performance to be had from these sticks if they are approached in the right way.



Test setup- notes, drawbacks and assumptions

All tests will be carried out on an Asus P5K-E motherboard, BIOS 1004. Two generic low-speed 80mm fans were positioned over the RAM at all times.


No VDIMM values lower than 1.95V were possible (board limitation.)


All voltages were measured using a Digital Multimeter which was placed across the relevant pins of a DIMM slot to provide the most accurate possible readings.


4GB kits of RAM need slacker tRAS and tRFC timings than 2GB kits. This was made apparently as testing progressed. I associate these timings with stability, not performance so I do not think that changing these timings will have a negative effect on performance.


tRRD, Write to Precharge and Write to Read delays were also considered “irrelevant” for the RAM. They have been tested and considered stable at 3,9,9 respectively to 693MHz using 2x1GB kits. I will assume for this review that IC density will not have an effect on this.


There appear to be few applications currently available that can make full use of 4GB of RAM. SuperPi 32M only uses around 600MB, so cannot be considered a full-blown stress test. I used Windows Memtest v3.6 as the primary test, using 4 instances of either 740MB or 750MB each (as other background apps would allow) and took 2 full passes of each instance as stable (partly due to time constraints.) SuperPi 32M was also used however.


I am limited to a 32-bit O.S. currently (Windows XP Pro,) so only 3.4GB was available for use. I appreciate this is not ideal, but it is still a popular O.S. and I believe several people will have a 4GB RAM setup in place before moving to a 64-bit O.S. so these tests should still carry some real-world parallels.

[K404]

Testing

I will not insult Crucials quality control or binning by testing it at 400MHz, 4-4-4 at 2.2V. Instead, I will start by seeing if it will run 400MHz 4-4-4 at lower voltages.

This RAM has EPP profiles which state that the RAM will, in addition to 400MHz 4-4-4, 2.2V, run the following setups:
400MHz @ 5-5-5-18, 1.8V
500MHz @ 5-5-5-15, 2.2V

Due to the amount of time it took to test each combination of settings, for this first set of results I did not squeeze every last MHz of stability out of the RAM but instead chose rough ballpark figures. Seeing as this is a review sample sent by Crucial, the possibility remains that the sticks are cherry-picked anyway so any results should only be taken as ballpark expectations of what all kits are capable of.


400MHz 4-4-4-12 2.05V






Despite numerous attempts and mixtures of settings, I could not pull any of the latencies down to “3” within 2.2V at 400MHz, which was a slight disappointment. This meant that the next step was to jack up the MHz. Initial attempts were unsuccessful due to tRFC being set too tight. It had to be loosened off to 42 to get even 425MHz stable.



450MHz 4-4-4-15 2.15V







500MHz 5-5-5-15 2.05V







550MHz 5-5-5-15 2.15V





Readers will notice another adjustment to the tRFC (Refresh Cycle Time.) This again was unavoidable in order to allow any form of stability at the MHz.



I will perform more tests at both higher and lower voltages in due time.



Initial Conclusion

After a quick check of some UK e-tailors, 4GB kits of PC6400 with these timings sell for between £76 and £99, while sets using slacker latencies (5-5-5) currently cost between £58 and £76. The price for this kit direct from Crucial is £110, with one e-tailor selling them for £94. That puts these kits pretty much at the top of the price range for the spec. Having said that, no other kit looks quite like this. Reports coming in from reviews of other RAM kits put overclocking ability at similar levels, so for performance, Im not 100% convinced the RAM is worth the extra. I believe the majority of people who buy this RAM will do so for the heatspreaders and LEDs, which are essentially unrivalled.

[K404]

Further testing

Stability looked slightly on the edge in some of the screenshots above, and this worry was reinforced by pushing things a bit further.

None of the following were considered stable.

2.25V 575MHz 5-5-5-15 tRFC=60
2.25V 475MHz 4-4-4-15 tRFC=60
2.20V 460MHz 4-4-4-15 tRFC=60

However, I loosened off tRAS to 18 and achieved the following (both at 2.25V):





Some of the original low-voltage tests were repeated with a higher tRFC value and produced fewer errors.

I will not start adding in mid-way latency sets like 5-4-4-15, even though I suspect they will help. The numbers I have posted should show the ballpark limits of the RAM, which was the intention of the review.

Stability dropped off very quickly- I could not get even SPi 1M to run reliably at 590MHz, with tRAS set to 21. Its possible that higher tRFC values will help, but 60 is the max I can set in BIOS (I did not consider Memset changes as valid, seeing as most people will want "fire and forget" settings.)

They key to 4GB RAM stability appears to be in tRAS and tRFC.


Thankyou very much to Crucial Europe for the review sample!

[K404]

Further update on results:

I removed the VDIMM mod from my mobo and re-ran some tests, using what i've learned so far about using 2x2GB RAM, (although I can use no lower than 1.95V.)

400MHz 4-4-4-15, 1.95V







500MHz 5-5-5-18, 1.95V






In my opinion, slackening off tRAS does not decrease performance, but can help increase stability.

[celemine1Gig]

Great! Exactly what Micron/Crucial needed to defeat all the PSC based Kits that are available right now. Competition is always good.

[Spyrus]

Yes but it cannot defeat Evo one apparently
Great ram though...

[K404]

Pics of RAM installed now up.

I am 100% aware of the errors- I had to manually "ok" each one to allow testing to continue. More voltage would probably remove them. They only occured in the 2nd instance to start and only in the x51%-x58% of the test, wondering if thats a bug of some kind? Not sure if they would create a fatal error, or a "try again" error.

[DerekT]

These seem to be PSC chips as well as denoted by the high TRFC. I have now worked with Corsair 8500C5 4GB, G.Skill 8500C5 5GB and the above specs are pretty much the same. Too close for me to think that they are different IC's.

[Spyrus]

tRFC is 60 here for 550MHz that is worst than Powerchips (52).

[xwards]

they look nice, oc isnt bad for 4GB kit

[K404]

More results up- now at 575MHz 5-5-5-18, 2.25V for Memtest.

[Spyrus]

Yes as you have noticed tRAS and tRFC is most important as you
get close to 580 with 2x2GB modules. tRFC needs >52 and tRAS>14 for sure
by my findings with Evo One.
I am interesting though to see the results with closed case and no fan over them
to see the difference with a 24/7 common setup without fans.

[K404]

added some more results at 1.95V. Mobo will not supply anything lower.

(400MHz 4-4-4-15 and 500MHz 5-5-5-18)

[KTE]

Pics of RAM installed now up.

I am 100% aware of the errors- I had to manually "ok" each one to allow testing to continue. More voltage would probably remove them. They only occured in the 2nd instance to start and only in the x51%-x58% of the test, wondering if thats a bug of some kind? Not sure if they would create a fatal error, or a "try again" error.

Great review, thanks
For me, that s/w errors erratically at times for no reason (memtest 24hr stable) and is buggy. I tested stock speeds with all software incl. RightMark Stability Tester and RM Memory Test and this in-Windows executable along with Orthos were the most bug-full and inaccurate, giving false errors randomly at any speeds. Especially when some other software tries to pinch some CPU time.

Most of the time it gave false positives, whilst there was no real error.

[Jamestuk]

Whenever I see pictures of the Tracer sticks I've always thought they looked a bit odd with the LEDs. Do they look any better in real life?

[Oliver]

If you in general like such things, it looks better irl, i have a crucial lanfest 2k7 kit and have seen tracer modules irl... but for me the virsual look have no importance, but how they overclock...

[Spyrus]

They do great with CAS4 that is for sure.

[TEDY]

what can we expect with CAS4?

[sonicsallen]

Thanks for the review, not impressed by the RAM though.

[geoffsthaboss]

these are pretty much the 2x2 ballistix right, tracer just means the addition of the leds?