19 Power Supply Round up on Anandtech!

[Aboroth]

Thanks JPSJPS, I was going to type something along those lines but your post was better than mine would have been.

I'm sorry, but this roundup is crap.

 

[jaeger66]

Originally posted by: JPSJPS


Your statement on page 27 ?With a power supply running so far out of specification, the CPU cache, system memory and video memory are in serious jeopardy of corruption.? Is just plain *COMPLETELY WRONG*!
These voltages are *EXCELLENT* and not close to being out of spec.

That's been mentioned, addressed, and he still feels he's right.

 

[KristopherKubicki]

JPSJPS:

I realize my "ripple measurements" were not. I am working to correct that. You are correct in assuming I am not an electrical engineer.

For point #3, the power supplies were the only things I changed from test to test. What else might cuase such problems?

As far as this thing about the variable fan switches, I'll stand by my comments. Several of the manufacturers who made those power supplies approached me before publication to tell me that the switches were giving them problems and users their problems. JPSJPS made it clear that my "ripple" test was done incorrectly anyway, so how would you know if its out of spec judging by what I wrote anyway? That being said, if you feel like running a 9800 Pro with 11.40V on the +12V rail (which is within the ATX12V spec), it will fail. Ill publish some of this data when i update the ripple part.

Cheers,

Kristopher

 

[KristopherKubicki]

JPSJPS and Jaeger66:

Please check the corrections I have made to the article. Thanks for your help guys.

Kristopher

 

[pelikan]

KristopherKubicki- That's the most professional response to criticism I've ever seen. I'm impressed.

 

[JPSJPS]

Originally posted by: jaeger66

Originally posted by: JPSJPS
Your statement on page 27 ?With a power supply running so far out of specification, the CPU cache, system memory and video memory are in serious jeopardy of corruption.? Is just plain *COMPLETELY WRONG*!
These voltages are *EXCELLENT* and not close to being out of spec.

That's been mentioned, addressed, and he still feels he's right.

The problem is that in the technical world, "feeling right" means nothing! Power supplies are designed to the specs in the links that that I posted (and the tolerances you posted earlier). In fact, what we have here is a hopeless guy with virtually no technical knowledge who is wishfully trying to define a ?good power supply?. (Hopeless means that he does not know enough to realize that he does not know enough)
End of this story.

In fact, in his earlier Power supply review he made a glaring amateurish error explaining the value of Power Factor Correction and never acknowledged that either.

You posted to a thread last night:
"2003 Power Supply Roundup Part II: Better Faster Cheaper"
but I do not see that thread anymore.
If his 9800 does not work at +12VDC ± 5%, then the problem lies with the 9800 and *NOT* with the power supply.

I have never seen a review that is as pathetic as this one. I am shocked that Anandtech would post this review and make a complete fool of themselves.

EDIT: Kristopher - Sorry I posted this while you were responding. I will get back to you.

 

[jeffrey]

Another victory for Enermax power supplies.
Why can't most other vendors power ALL the rails like Enermax can?

 

[JPSJPS]

Kristopher - Thanks for the PM! I see you got stuck doing a job out of your field of expertise and that explains the whole story.

Unfortunately, blunt criticism sounds hostile but that is not my intent either.

As I said earlier, obviously you are a nice guy working hard trying to do a good job but you are too short on technical knowledge.

RE: On the response page of your review I offered this:
?On a positive note, I volunteer to proof/edit/correct any future articles that you propose if you desire. You have my email.?

John

 

[Mem]

KristopherKubicki, I always use Enermax PSUs,however would it be possible in the next PSU roundup to have some other popular models like Chieftec PSUs(made by Channel Well Tech)included since they are good on price and fairly easy to find online,I`ve fitted quite a few for friends and they are great for price/perfomance for those that want a good decent PSU without breaking the bank so to speak.

 

[KristopherKubicki]

[i[You posted to a thread last night:
"2003 Power Supply Roundup Part II: Better Faster Cheaper"[/i[

Our webmaster did away with the whole AnandTech Articles forum (not just the thread). I have contacted the powers that may be about this, you guys might want to drop a line in the Forum Issues forum if you feel the same.

Cheers,

Kristopher

PS: Thanks for the understanding JPSJPS

 

[TomNoddy]

I'm absolutely stunned by the number of memory errors that interference test threw up, especially when these were all brand name power supplies. When building PCs for people I try to make them as stable as possible, but this makes it look like registered ECC memory is the only way to go. If you want an AMD system I guess that means going back to the AMD 760 chipset..

It would be great to know a few more details about how they set that test up. Specifically ;

1) Were the power supplies loaded to their maximum capacity in this test, or just powering the motherboard and floppy drive so you could boot into the memtest program?
2) Was the motherboard you used from a first-tier manufacturer? I'd entirely understand if you don't want to mention specific models/brands.
3) Was the motherboard powering the fastest CPU it could run or was it running under spec (eg. Barton 2500 in a board that'd take a 3000)
4) Was the memory from a tier-one brand i.e. the stuff that finds itself in comparative performance reviews across the web, or was it el cheapo generic
5) Were all DIMM slots filled?
6) Was the RAM running at it's rated speed or under spec
7) Were chipset timings by SPD or Agressive.
8) Were the tests conducted onboard the ISS, or on Three Mile Island


Thanks

 

[KristopherKubicki]

Tom:

Most if not all of these questions were answered in the review.

1.) We used 350W of load.
2.) ECS K7VMM+ (We ended up breaking a few when trying to build our resistor platform. They were the cheapest things I could find. Do keep in mind that we used the same motherboard for the entire test, AND the test was done in alphabetical order by manufacturer).
3.) No, it was running an Athlon 1700+ at spec. We had the resistor platform attached to the power supply though.
4.) Yes, the memory was cheap PC2100 from Crucial. Since we ended up frying a lot of hardware before getting this right, we werent going to throw Mushkin DDR500 in the test.
5.) 1 stick.
6.) Spec.
6.) Default.
7.) I did the first test inside our breeder reactor but my good luck goldfish kept dying, so I figured it was bad for my Fung Shui.

Kristopher

 

[MercenaryForHire]

:beer:

- M4H

 

[sharkeeper]

I'm absolutely stunned by the number of memory errors that interference test threw up, especially when these were all brand name power supplies. When building PCs for people I try to make them as stable as possible, but this makes it look like registered ECC memory is the only way to go. If you want an AMD system I guess that means going back to the AMD 760 chipset..

I am too since I've run Memtest86 on quite a few systems and have only seen errata reported when there is a memory problem. I usually run it for 24 hours regardless of how much ram is in the system. Perhaps the board was crappy?

EDIT: Ok, I see you're using a modified version of MT86. Can you be more specific on what was changed?

Our workstations use ECC exclusively and rarely do we see bit flips in the logs.

-DAK-

 

[KristopherKubicki]

We just took one test, and let it write as much addressable space as possible (as opposed to just a few bits). We then just inserted a sleep() for six hours and let the test count the errors. The big difference here is we are using as much of that 256MB stick as possible as opposed to just a few blocks.

Ill probably post a diff in the review somewhere for all you nix folk.

Cheers,

Kristopher

 

[JPSJPS]

Originally posted by: TomNoddy: I'm absolutely stunned by the number of memory errors that interference test threw up, especially when these were all brand name power supplies.

I am surprised too and I do not recall seeing any other similar published test data. While three test runs probably are not 100% conclusive, they would tend to validate the error rate data differences vs power supplies.
I am *not* an expert on the theory for this type of memory and have not researched predicted error rates , so that is something that I plan to do when I have the time. The "small" number of errors strongly suggests that this is a problem caused by transients or random noise. If these errors were caused by interference from the DC-DC switching circuits or any other repeating steady state signals in the power supplies, the error rates would be *much* higher.
If we consider the clock speeds for modern RAM, even a noise spike would have to be no more than a few nanoseconds to cause only a single error. Also, if I recall correctly?? the typical Auto/Self Refresh interval for this type of volatile memory is about every 10-20 microseconds so an interfering signal of that width could wipe out the entire data stored in the memory.

Thus, labeling the test as "Interference" (from the power supply) is probably wrong but I don't have a better reason/name for it without guessing.

Kristopher - How much data was loaded into RAM for these tests?
EDIT: THANKS - I was typing while you posted so ignore this

John

 

[VidarL]

Kristopher:


In the article, you write that the Sparkle w/120mm fan had one of the lowest pitch levels, and perceived to be very quiet. Was it the one you perceived to be most quiet of the bunch? Or was the Zalman/Fortron even more quiet?

Vidar

 

[KristopherKubicki]

Vidar, In my opinon, scientific measurements aside, the ThermalTake unit was probably the quietest. I also thought the Enermax units were very quiet, but their rear fans were only spinning half the time.

The Sparkle FSP350 was spinning constantly but it was more of a low rumble rather than a whine. The Fortron/Zalman were not the quietest, but I felt they were a good blend of low noise, low price, and solid performance. Thats why they won the editors choice.

Cheers,

Kristopher

 

[Rufus210]

A couple of things:

1) Can anyone find the Sparkle FSP350-60PN for sale? I can't find it (or Fortron or any other re-badged one).

2) Are there any differences at all between the Fortron (and Sparkle and I think A-Open also has a re-badged one) and the Zalman units, other than price? Also any hint as to where to find the Zalman for the Fortron price?

3) Since the ThermalTake is variable speed, were the Noise/Temp testing done at full speed? Also what sort of difference does dropping it to slow speed make?

4) You mention how hot the Enermax's get (44-45 C), but the ThermalTake is very close to them (42 C) and signifficantly above the rest of the pack without mention. Any reason?

5) You state "the Fortron/Zalman were not the quietest," but the noise graph lists them as the same dB for low and 0.5 dB louder for high, which should be around margin of error. Are the numbers off or does the dual fan vs. single fan make some sort of difference?

Like the others I'm not the most impressed with your EE skills, but I'm more looking for noise and quality than bare numbers. I'm currently looking for a replacement for my current Antec 400w that's around the same quality but a good deal quiter and I'm trying to decide between the Fortron/Zalman and ThermalTake.

-Peter

 

[KristopherKubicki]

1.) I think Aopen sells the FSP350 as well. I have stumbled across it in a few places, (I think Directron.com was carrying them for a while)

2.) I have noticed the Zalman units increase in price significantly over the last couple weeks (Perhaps due to Toms review as well.) I am leaning a little bit more to the fortron unit now.

3.) All tests were done at full speed unless stated otherwise.

4.) Enermax's rear fans only turned on occasionally (motherboard controlled).

5.) I am only stating the facts. You can reinterperet the numbers anyway you like.


(Edit: I think you mean my lack of EE skills.) Hope that helps.

Kristopher

 

[pelikan]

Originally posted by: Rufus210
A couple of things:

1) Can anyone find the Sparkle FSP350-60PN for sale? I can't find it (or Fortron or any other re-badged one).

NewEgg has the Fortron FSP350-60PN.

 

[Rufus210]

Originally posted by: KristopherKubicki

3.) All tests were done at full speed unless stated otherwise.

How much of an effect did the slower speed have?

5.) I am only stating the facts. You can reinterperet the numbers anyway you like.

I was more asking for your opinion if you can actually hear a difference between the two.

Thanks again.

 

[Khalad]

Hello All,

All this technical stuff is more than I understand at present, but I appreciate the effort Kristopher made with the article and that the rest of you are contributing.

I am seeking to build a quieter than average pc that still runs at a medium to low end case temp. I was looking at www.silent.se and ended up finding www.nexustek.nl

Now I would really like to see the NX-3000 and/or NX-4000's tested up against other reliable competitive PSU's.

Since the listed 22dB idle for the NX4000 is great but I'd like to see them in a review.

If anyone has as experience with these models or HSFU from NexusTek I'd love to hear your thoughts.

Khalad

 

[jaeger66]

Originally posted by: Rufus210


Like the others I'm not the most impressed with your EE skills, but I'm more looking for noise and quality than bare numbers. I'm currently looking for a replacement for my current Antec 400w that's around the same quality but a good deal quiter and I'm trying to decide between the Fortron/Zalman and ThermalTake.

-Peter

FWIW, I chose a 365W Enermax because it was quieter to my ears than comparable units from Antec and Sparkle. Mine has the thermal control fans, not the manual. It does run very hot under load, but that's not a big deal to me.

 

[Chaotic42]

Originally posted by: RalfHutter


That's weird. The Anandtech PSU article I read gave the Fortron top honors ("Editors Choice"). The PCP&C "suprised them the most" but "sounds like a train" when running.

It was a joke, but apparently not a funny one.

Anyway, I have a PCP&C 350W in my FreeBSD system and it's very loud. Louder than my Antec 550, 2 CPU fans, SCSI hard drive, and 3 case fans put together... and I sit next to that, the BSD system is about 10 feet away.