Some Disconcerting News About The Ttgi Psu's

[Nohto]

I always thought the TTGI PSU's were a quality unit, especially since Opp has given them his seal of approval. But some new light has been shed on this subject showing a different story. 2003 Power Supply Roundup at Anandtech (http://www.anandtech.com/showdoc.html?i=1841)


Memory Errors (less is better)

[Nohto]

.

[Nohto]

and

[Nohto]

I wish I would of known this before I bought 3 of them. Looks like the PC Power & Cooling did as we all would expect. The fan switch in the rear of the PSU is suppose to cause some of the trouble, but in all 3 units and all 3 tests each?

[TheDude]

The 520ss of which I own several, did fairly well in all but memory errors. Still a good deal for $78. :D but not the best PSU out there for sure...plus I suspect that something has changed since I got mine a year ago...I have never had a problem, but hear of problems with them recently?

[Nohto]

bump

[sjohnson]

I don't know - while I'm still digesting the article here are some initial impressions:

The review and analysis were written in a style that makes you want to skip the reading and jump to Anand's conclusion rather than forming your own conclusion. I've been burnt by this style too many times in the past (I call it the "baffle them with bulls***" style). The review *MAY* be valid but the style of reporting approaches techno-babble.

I've pretty much gone over the "memory error" completely. Two problems:

1. Having memtest86 load the memory, waiting 6 hours, then having memtest86 check the memory contents is not reflective of any real-world use. I know they're trying to check EMI shielding of the PSU, but is this really the way to do it?

2. While I understand that time constraints enter into any review, testing the "memory error" portion only once for each PSU makes the results meaningless and should not be used for comparison of PSU's. Random errors or external influences could eaisly account for the discrepancies between PSU's.

Overall, the difficulty in digesting the word content of the review will probably keep me away from further study of the review. The charts are nice, but understanding what Anandtech did to acheive the results so that I can form my OWN opinion is simply a lot of work...

[shrae]

Originally posted by sjohnson
The review and analysis were written in a style that makes you want to skip the reading and jump to Anand's conclusion rather than forming your own conclusion.

He does this... a lot. And even (as sjohnson said) his information/premise may be completely valid and done in a fair and unbiased manner, it's this style that keeps me off his site. It's impossible for me to read many of his articles because there's an overwhelming sense of predisposition.

[Nohto]

It was difficult, to say the least, for me to read the article, but seeing how I have 4 of the WORST scoring PSU's in the review, I had a little incentive.

sjohnson
While I understand that time constraints enter into any review, testing the "memory error" portion only once for each PSU makes the results meaningless and should not be used for comparison of PSU's.

I think you misread something. They ran the test 3 times on each PSU. Also, it is a little more than just a coincidence that all 3 TTGI PSU's were pieces of crap. (That is in the interference dept.)


Unfortunately, it took 6 hours to run this test and we ran it three times for each supply to assure accuracy.

[sjohnson]

Crap is a bit strong. Even the number of errors reported for the TTGI is a tiny fraction of the millions of memory cells "tested" with that method. The test measured how well a charged memory cell maintained it's contents for 6 hours, without change. Is this real-world? And, the conclusion is that the PSU must be the cause of the "error" because nothing else changed. Other than using the same system (exception PSU), how are we assured that there were no other influences?

If the test were run three times (sorry, I missed that in my reading) then why is only a summary shown in the tables? How much variance existed between tests?

Again, I understand the amount of effort to compile such a large review, kudos to Anand for the effort! - but even three tests per PSU provides statistically insignificant results - EXPECIALLY when the sampled "errors" were from MILLIONS of memory cells. 100 samples is a minimum if you want loose data correlation, 1000 is better. No way they could do even 100 tests but they present the data as a valid comparison, as though the results are fact

Interesting? Yes!

Thought-provoking? Certainly!

Conclusive? IMHO, most definitely not! :)

I own or have owned Sparkle, Enhance, Enermax, Fortron and TTGI. All good PSU's. I only wish I'd save the money to get a PC Power unit...

[Nohto]

If the test were run three times (sorry, I missed that in my reading) then why is only a summary shown in the tables? How much variance existed between tests?
It shows you right in the table the results for each test. I don't know how you missed it. Trial 1,2,3.

Other than using the same system (exception PSU), how are we assured that there were no other influences?
We aren’t, but I highly doubt he has an ulterior motive, trying to make TTGI look like the worst PSU out of the entire test when it comes to interferance.

I know CRAP is a strong word, but you have to consider the fact that TTGI puts out a PSU that is lower priced than the competition and I should of thought about this also, but YOU GET WHAT YOU PAY FOR. A 520w PSU for $85.00 seems too good to be true, and it just might be if you look at it that way. The only thing that they should of done is bypassed the fan control switch and re-tested the units to see if that was the true factor in the interference problem.
I am curious how PC power and coling test their interferance. They claim to be the cleanest PSU's on the market and from what I hear they are.

[sjohnson]

Sparkle and Enhance are inexpensive as well, so the aphorism about getting what you pay for doesn't necessarily apply ;)

I don't think any ulterior motives existed so much as poor scientific method and reporting (the text, not the charts).

I'll admit that I only read the review once (not an enjoyable experience), and my remarks regarding the data presentation weren't accurate. My premise still stands that no one could test "interference" (as presented by Anandtech) on all those PSU's to within a 50% siginificance level, let alone the 90% or so needed to draw conclusions. The variance of errors is quite large on several of the tests, but this could be due to the small number of errors manifested in all the tests. 8 cells erring out of how many millions?

I guess my point is also that the review made some assumptions without backing up those assumptions with fact or with correlating evidence - such as the "memory errors" being caused by the PSU. 6, 7 or 8 one bit errors out of millions of cells is statistical noise. Good information may exist in that noise but it's drowning in the static. For all we know neutrinos knocked the erroring cells because the TTGI and other "crap" PSU's were tested when the rotation of the earth positioned the memory so as to present a better target for neutrino interference.

A far-fetched scenario, yes - but my point is that we don't know what caused the errors and the testing doesn't prove that the PSU caused them.

Getting zero, one or two errors is also statistically insignificant. You can tend to show failure but it's nearly impossible to show success via data analysis. The unknown is a serious impediment to drawing any conclusion(s) from the presentation.

Why not use standard RFI testing on the PSU's rather than jumping to the conclusion that memory degradation during 6 hours of memory stasis was due to the PSU?

In any case, the real proof will be in how well your TTGI behave for YOU, not Anandtech :D

[TheDude]

In any case, the real proof will be in how well your TTGI behave for YOU, not Anandtech

I agree 100%. It was an interesting read, but my 3 520w TTGI's are doing fine for me for the price I paid and the use I put them to. I upgraded to a truecontrol 550w in my prommie and noticed a significant improvement...pleased with it for the same reason.
For a top of the line choice for best Ocing, I would shell out the cash for the PC Power and cooling PSU.

[jinu117]

Considering how well the Fortron in general behaves... I am wondering how it would perform with same amount of airflow the PC P&C PSUs have. (ANY PSU can keep it cool if it need to be that noisy -_-... After all, heat is one of the worst enemy when it comes to efficiency of PSU...) Maybe I will get el cheapo fortron 400W and Put swap and put extra fan to see how well it does vs my Antec TruePower...

[daddy_fizz]

yeah i ordered one of those fortron, only th 550 watt one (only $79) on newegg, will try it out and see how it does compared to my TTGI 520W Triple fan

~Fizz

[jinu117]

WAIT>.... those are for dual procs...!
The largest fortron for typical use is 530W -_-;
Hope it didn't surprise you...

BTW, one funny thing...
Since I didn't like my MBM fluctuation with Antec TP 550W... I gave a shot with my trusty old Seasonic 350FS... (ahaha I know... 350W... but efficiency and coolness of PSU is unmatched other than its siblings... and newer generation). Wierd enough, I am getting MORE STABLE voltages (doesn't net to better OC though) and a little cooler case... -_-;
I don't really know what to think anymore. (they have 80% efficiency seasonics now... only up to 400W though)

[Tedinde]

I just switched from a 550 antec in my vapo to a ttgi 550 in opps review. 12v now @ over 12v without modding yet, My antec was @ 11.49. 80 watt tec on the gpu.

Opened up the ttgi, and i can get up to 13.3 volts if needed. I've seen no problem for the 4 days i've had it. 24/7

[dafatguy]

this sould be retested and then compared