Yeah, it's different power supplies from the same company. But is that the sort of product that a reputable company would release? Or even a kind of shady one?
Originally posted by grndzro Power supply 62$ Looking online I can't find a single bad review of Diablotek's UL series
With power supplies, it's not about trying to find bad reviews.
It's about trying to find legitimate good reviews. THere are tons of user reviews, and open-box reviews, and places where they plug a power supply into a computer and say "Yup, it turns on". There are very few sites that can actually test power supplies with load testers, and see things like efficiency, droop, ripple, and get the actual performance of the unit. My favorite site is HardOCP - if a power supply passes their test, it's an excellent unit, even if it doesn't win any awards, because their testing is pretty brutal. There are (a very few) other sites that do good testing as well, but generally, if HardOCP didn't give the unit a Pass, I won't even consider it.
I couldn't find a single legitimate review of a UL unit. I did find this one on a Diablotek PHD series:
HardOCP With the PHD650, Diablotek seemingly tried to polish a turd with a turd and as such we got the expected outcome, a pile of crap. Given the horrible build quality of the unit and poor component selection of the unit that is hardly a startling result. With a current price of ~$49.99, users would be better off taking their cash and setting it on fire rather than buying this unit. There is quite literally no price point or situation where the Diablotek PHD650 should be a consideration for users as it failed almost every test possible today. Retailers and e-tailers should be ashamed to be selling such a product.
They basically say it's a crappy power supply, but it didn't explode. The only good mark it got was that it was cheap, and they remark that it's great for a "Backup Crappy Machine." They didn't run enough testing on it to specify if it was actually a decent performing power supply though (unlike HardOCP, which did).
You don't really need 650W in that machine - the 6870 has a TDP of 151W, the CPU is another 100W give or take, the rest of the system is another 100W give or take. That's 350W total peak. AMD recommends 500W.
I know this is a budget system, but really your putting a lot of faith in a power supply, no matter what budget the system is. For less money than the 650W Diablotek costs, you could get a Corsair or Silverstone 500W, or any other various high quality 500W: spend less money for demonstrably better quality. Newegg has the Corsair CX500 with a $20 rebate right now, with a list price of $59 - cheaper even before the rebate, and has the required 6-pin connectors for the 6870. It's not their best power supply, but I'd be more willing to bet on that than anything by Diablotek.
Yeah, it's different power supplies from the same company. But is that the sort of product that a reputable company would release? Or even a kind of shady one?
Google Dioblotek UL series. they are a different breed than their other lines. Otherwise I agree with Diabloitek in general.
Personally my latest build has a Diablotek UL in it and so far 1 year under heavy load no problems.t
As far as their PHD series I wouldn't trust it in a bucket of water near anything flamable.
Go ahead google "Diablotek UL" and click everywhere one is available. From 575w,675w775w. Read the comments. Companies can turn over a new leaf.
Do you want a power supply that demonstrably is good? Or merely one you hope might be good.
If you've got a good product, you don't keep it secret. You send it to reputable sites to get good reviews and show the world that it is a good product. Some of Diablotek's power supplies have been so awful that I'd assume all of their power supplies are garbage until proven otherwise. And a user review that says "I pulled 200 W from a 600 W power supply and it didn't blow up" doesn't count as proven otherwise.
Google Dioblotek UL series. they are a different breed than their other lines. Otherwise I agree with Diabloitek in general. Personally my latest build has a Diablotek UL in it and so far 1 year under heavy load no problems.t As far as their PHD series I wouldn't trust it in a bucket of water near anything flamable. Go ahead google "Diablotek UL" and click everywhere one is available. From 575w,675w775w. Read the comments. Companies can turn over a new leaf.
I did, where do you think I found the review saying it was a crappy unit but it didn't explode.
The post you didn't read http://www.thinkcomputers.org/diablotek-ul-series-575w-power-supply-review/4/ They basically say it's a crappy power supply, but it didn't explode. The only good mark it got was that it was cheap, and they remark that it's great for a "Backup Crappy Machine." They didn't run enough testing on it to specify if it was actually a decent performing power supply though (unlike HardOCP, which did).
It's your money, your rig. You couldn't pay me to put a Diablotek in my personal computer though.
I'v seen plenty of cases where DDR3 tested for only Intel XMP didn't run at the rated specs on an AMD system. But usually it runs fine. eh, I couldn't find a direct comparison between an OC PII-X4 and an OC 4100. But for bang $ the 4100 wins IMO.
I looked up the memory thing, because I had never heard of anyone having issue with XMP RAM in a AM3 or AM3+ motherboard. Turns out, some people have had issue, in most cases it appears not because the RAM sucks, but rather, because they didn't configure the BIOS to match the specifications of the RAM. XMP is an Intel-specific thing, where they can embed into the memory 3 different timings (stock, performance, "extreme"), and the motherboard can then use that information to auto-configure the timings. Non-XMP motherboards won't have this information, so they won't auto-configure the DIMM to run at the proper timings and voltages. XMP is totally optional, it just provides a mechanism to auto-setup the motherboard. The memory should work at it's rated specifications, provided you tell the motherboard what those are.
I will give you, most of the time, you plug in memory, and if it isn't certain, it goes to some low default clock speed and works. I don't know if XMP interferes with that on AM3 motherboards or not - the user reports were unclear, other than many were able to get the RAM to work fine by manually setting the timings, or perhaps more importantly, the voltage.
As far as Bang for the Buck CPU...
Now I'm curious, so I'm looking as I haven't priced AMD CPU's in a while (Intel's prices never really change).
An FX-4100 is going for $109 right now on Newegg. That's a stock 3.6G.
The only Zosma (4-core Thuban) they carry is $124 at 3.0G. A stock 3.6 Deneb PII X4 is going for $149 - that's a BE, which adds a lil bit to the cost. A low-speed (2.8) Deneb X4 can be had for $94. A Thuban X6 2.8 is $149.
Given the ~20% per-core performance difference, the FX-4100 at 3.6G should be comparable to the 960T PII X4 at 3.0G.
So I have to hand it to you, bang for the buck, looks like the FX-4100 does win if you go clock-for-clock. The last time I had actually looked at Zambezi prices, the FX-4100 was still near $125.
Since the old X4 940 BE has been discontinued, I think you may be right, the FX-4100 may be the Bang for the Buck budget CPU. I don't know that $40 is worth the extra performance to get the only BE Denebs that are still in inventory, and your only looking at equivalent stock performance out of the Zosma or any of the X6 Thubans really.
Ridelynn you obviously like bending the truth to fit your needs.
No where did the UL review say anything about exploding. I's a budget gaming power supply that happens to be reliable.
Those 12v ripples were pretty minor and easily ironed out by the motherboard.
The Seasonic had the same 12v ripples and could get loud under full load. was based on older PS designs.
The Diablotek 675-UL is 20$ cheaper. Is reliable or you would see more people complainng since it is widely available online and in retail. In general people have nothing bad to say about it, and there are plenty of comments if you look at the online retailers. It has a 3 year warranty that covers component loss. What more do you want.
It's cheap, it won't blow up, Runs cool, and it's quiet. It has over voltage and overcurrent protection, 105c capacitors, Plenty of connectors. Only downfall is 3% or so less efficiency.
(THE UL SERIES. I'M NOT TALKING ABOUT THE PREVIOUS GENERATION)
I doubt the OP will be pulling 675w. This supply will do fine. For just 63$
I just read a review of the Rosewill 630w green and it is much better at 70$
Originally posted by grndzro Ridelynn you obviously like bending the truth to fit your needs. No where did the UL review say anything about exploding. I's a budget gaming power supply that happens to be reliable. Those 12v ripples were pretty minor and easily ironed out by the motherboard. The Seasonic had the same 12v ripples and could get loud under full load. was based on older PS designs. The Diablotek 675-UL is 20$ cheaper. Is reliable or you would see more people complainng since it is widely available online and in retail. In general people have nothing bad to say about it, and there are plenty of comments if you look at the online retailers. It has a 3 year warranty that covers component loss. What more do you want. It's cheap, it won't blow up, Runs cool, and it's quiet. It has over voltage and overcurrent protection, 105c capacitors, Plenty of connectors. Only downfall is 3% or so less efficiency. (THE UL SERIES. I'M NOT TALKING ABOUT THE PREVIOUS GENERATION) I doubt the OP will be pulling 675w. This supply will do fine. For just 63$
The Zosma X4 is a failed Thuban. I'm kinda thinkin I don't want a failed anything.
I do? I paraphrased in my response - as it wasn't in quotes like the other review quote was... however, other Diablotek power supplies ~have~ exploded. The UL is apparently the only one that doesn't. It is a budget power supply. I have yet to see an actual review that deems it reliable, and from all indications the Diablotek UL is a rebranded Great Wall PSU, but I can't confirm the model. The closest I could find is an OCZ ZT 650W by Great Wall, which failed the HardOCP power supply testing because it couldn't deliver past 75% under nominal conditions without completely shutting down, nor could it meet it's rated usage temperature. Granted - I don't know if this is the same as Diablotek UL - I suspect it may be, but that's as reliable as I can say right now.
And my take is, no power supply is considered reliable until it's been tested by a reputable site and proven otherwise. I can't prove anything about the Diablotek UL series - I can't even prove who the OEM is.
To bend your own quote... the UL is possibly/allegedly the only Diablotek product that doesn't truly suck. I'm kinda thinkin' I don't want to buy anything from a company that puts out 99% suck and 1 maybe ok product. Especially when there is better for cheaper. I don't know why you even still defend them. It's like trying to say Charlie Manson isn't all bad because he wrote a couple of songs.
Although, the Rosewill (rebranded ATNG) is a much better power supply - I could at least find a real review on it (as a Nexus brand) that passed, I agree. But.. I still don't know why you want a 600+W power supply in the first place. A really good 500W power supply is still overkill for an AMD system with a 6870. This is a budget system after all right? Why pay more for a power supply from a junk manufacturer just because it has a sticker on it that has a bigger number, or pay for more power than your ever going to use. If you have it - put that money towards quality or longevity, not toward a number on a sticker.
I really wouldn't look past the Corsair CX500 or Seasonic at $59 (with $20 rebate) or Antec Basiq 500W for $54. Quiz can probably find a suitable decent power supply for even cheaper, I don't do enough research on what is rebranded what.
Every CPU (and GPU really) is a failed something or another - that's how they make them. Dozens on a wafer, those that test well get binned high, those that test worse binned lower. All the same silicon. There are many, many reports of people taking Zosma's and unlocking the 5th and 6th cores and having a full-blown X6. Just like people unlock 6950's to 6970's. Just like nVidia snips ROPs and TMU's and magically has the 580, the 570, the 560Ti, etc... Just like Intel has 5 different Core i5 1155's and 3 different Core i7's and the entire i7/i5/i3 lineup in the first place...
I'm not debating that Diablotek as a startup company was a complete failure, It was.
I'v worked at several IC companies in phoenix so I know that internal managment can drastically affect the quality of components as well as work conditions.
Such as having quality through hole soldering equipment and personell.
Upgraded QC, Higher standards for quality components. Better testing equipment.
I don't know where the problem was with older Diablo PS lines. The reviews were such that something must have changed if the UL's aren't blowing people's computers up by the hundreds.
Companies can change when managment is replaced or a big investor pops up, or it's bought out.
It appears Diablotek is owned by Evercool which makes some decent stuff. I would imagine the parent company tossed Diablotek managment out and put someone decent in charge, who knows. UL is a step in the right direction. mabye we'll see a new line come out with good reviews. At least UL passed UL lol
Something isn't quite right with the Zozma imo. I know they are Thuban cores but it's performance stinks.
How do you know that the Diablotek UL series happens to be reliable? There are a lot of problems that a power supply can cause short of exploding. If the UL series is so good, then why isn't it 80 PLUS certified? 80 PLUS certification doesn't automatically mean that the power supply is good, but not having it almost invariably means that it's junk. The requirements for certification are pretty low, so if a power supply can't even meet such a low bar, something is amiss.
Hardware Secrets does do good power supply reviews, but you should be aware that they're pretty generous with their "golden award". That's why you have to read the review. If Hardware Secrets gives a power supply a golden award and Hard OCP rates the same power supply a "fail", they're not necessarily contradicting each other.
For the Rosewill Green, the review says the power supply is kind of all right, but hey, it's cheap, so we'll give it a golden award. The review is also from 2009, and a part that was pretty good in 2009 would be considered significantly less good by today's standards.
But let's look at the review. All Teapo capacitors, which aren't that bad, but aren't that great, either. Some rated at 85 C, and some at 105 C, which again, isn't that bad, but isn't that great, either. Ripple on the +5 V rail is about 3/4 of spec, while on the +12 V rails, it's about 2/3 of spec. That does constitute staying within spec, but a lot of power supplies can do a lot better than that. Energy efficiency barely stays over 80%.
And look at the wattage rating. Supposedly it's 630 W total, but only 611.5 of that is on the +12 V, +3.3 V, and +5 V rails, so it's pretty much impossible to pull 630 W from it without running it out of spec. That's not the sort of thing that reputable power supply companies do. The +12 V rail is officially rated at 600 W, but a component analysis finds that its rectifiers are only rated at 514 W. The power supply supposedly has a "peak" power of 730 W (any mention of peak power is a bad sign, by the way), but shuts down if you pull 670 W from it.
Now, much of that is merely overly optimistic wattage ratings. If they call it a 550 W power supply rated at 42 A on the +12 V rail, then the criticisms of the above paragraph disappear. But again, inflating wattages like that is something that reputable power supply companies don't do.
Originally posted by grndzro Something isn't quite right with the Zozma imo. I know they are Thuban cores but it's performance stinks. http://www.cpubenchmark.net/cpu_lookup.php?cpu=AMD+Phenom+II+X4+960T 954 per core for the zozma and 971 per core for Thuban @ 3ghz. Memory bandwidth would dictate 4 core scaling should be better than 6.
CPUMark doesn't measure memory bandwidth - it's strictly CPU (which is a big part of the reason why it only tells a very small part of the story). A difference of 17 per core comes out to 1.75% - that's within the margin of error.
The X4 960 falls exactly where it should between the 955 and the 840 benchmarks, and is in line with the X6 1075T benchmark.
The real outlier is that the FX-4100 gets a passmark score that is significantly higher - 4343 vs 3819, or 1086 per core, yet in real world use, is about ~20% per core slower than Thuban (or rather, needs ~20% higher clock speed to get the same results). So take that difference of 17 between the 960 and the 1075T for what it's worth, and take passmark scores in general for what they are worth.
I don't see anything at all wrong with it, or with any of the Phenom II line really. But at a price of $124, you can get the same performance out of a FX-4100 for $109 on the same motherboard. That is the only point I was illustrating, mainly to myself. Which is what you were arguing for in the first place.
Bulldozer is slower per core than Zosma, but not necessarily 20% slower. How well threaded the program is will affect the per core performance, too. Each Bulldozer module has one integer scheduler shared between two cores, so using both cores from a module won't get you double the performance of only using one core. AMD estimates that two cores should get you 80% more performance than one in programs that scale perfectly to two cores. Thus, one would expect an FX-4100 to fare better against a Phenom II X4 in programs that only scale to one or two cores (so that they don't have to use both cores from a module and deal with the shared scheduler) than in programs that scale flawlessly to four cores.
Bulldozer also has some instructions that Zosma doesn't, most notably SSE4, AVX, and FMA4. (Also AES-NI, but that's very specialized and will never be used for games.) AVX and FMA4 are barely used at all thus far, and SSE4 is only a few years old as well. But if those instructions become more commonly used in the future, Bulldozer will see performance gains from it (as will Sandy Bridge for SSE4 and AVX, but not FMA4) and Zosma won't.
There are a lot of pros and cons for FX-4100 versus Phenom II X4 960T. As I said earlier, I'd be shocked if there aren't a lot of programs in which each beats the other. So I just picked whichever of the two was cheaper.
It's showing the FX8150 to be slower per core by nearly 20% in almost every test, to a Thuban 1100T. The FX8120 is more than 20% in every test, although it has a lower base clock.
1100T is a 3.3G base clock, FX8150 is 3.6, FX4100 is 3.6
So I interpret that to mean that a 3.0-3.2 Thuban will perform about the same as a 3.6 Zambezi, core for core.
Now going through the rest of the benchmarks on that page, none of which are actual gaming unfortunately, it does go back and forth a bit.
Gaming, however, is mainly based on per-core performance, which is why I place so much emphasis on that first link.
Here are a few real gaming benchmark numbers that test my hypothesis:
This one is interesting, because they test 4, 6 and 8-core Zambezi's. In pretty much every case, it just came down to a matter of Ghz rather than cores (no surprise there), and the Zambezi's were ~generally~ behind the Denebs and Thubans by about the same margin - roughly 15-20%, with a couple of outliers.
Anadtech - same thing. The same delta in per-core performance more or less.
Core for core, for gaming, Zambezi needs more Ghz to accomplish the same job. Fortunately, it can attain higher speeds in general, but in many cases, it can't get fast enough. Sure there are a lot of exceptions, especially when you look outside of gaming, but just looking at gaming, Zambezi is generally slower clock for clock and core for core than Thuban/Zosma, and in many cases, even Deneb - and that's what matters in gaming.
All of that being said... if the price for a 3.6G Zambezi is cheaper than a 3.0G Zosma/Thuban/Deneb, then it wins, even if it is slower per clock, it has the base clock to make up for it and come ahead in performance per dollar. Just comes down to which is on sale that day, in the budget arena.
Comments
New build 689$ I think it's about as good as it gets for under 700$ before shipping
Power supply 62$ No more Diablotek, updated to use Rosewill green 630
Case 30$ Put a 80cm fan on top and side blowing in and 1 in the back. airflow looks pretty good then.
Case fans 10$ 4 in case one is a lemon, and the price drop almost pays for the 4th one
OS and Motherboard 185$ (Quizical)
Processor And Ram 138$ (Quizical)
Hard drive 80$(Quizical)
Cd/DVD 20$
Video card 165$
I'v seen plenty of cases where DDR3 tested for only Intel XMP didn't run at the rated specs on an AMD system. But usually it runs fine.
eh, I couldn't find a direct comparison between an OC PII-X4 and an OC 4100. But for bang $ the 4100 wins IMO.
Most people's opinios are that the Stars cores will always be faster. But at 4.6-5 ghz FX might win out. But then a PHII-X4 at 4ghz might stomp it.
That Schedueler update actually helped my X6 a bit.
No, no, no. Don't go Diablotek anything.
http://hardocp.com/article/2010/12/17/diablotek_phd650_650w_power_supply_review/
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=187
Yeah, it's different power supplies from the same company. But is that the sort of product that a reputable company would release? Or even a kind of shady one?
With power supplies, it's not about trying to find bad reviews.
It's about trying to find legitimate good reviews. THere are tons of user reviews, and open-box reviews, and places where they plug a power supply into a computer and say "Yup, it turns on". There are very few sites that can actually test power supplies with load testers, and see things like efficiency, droop, ripple, and get the actual performance of the unit. My favorite site is HardOCP - if a power supply passes their test, it's an excellent unit, even if it doesn't win any awards, because their testing is pretty brutal. There are (a very few) other sites that do good testing as well, but generally, if HardOCP didn't give the unit a Pass, I won't even consider it.
I couldn't find a single legitimate review of a UL unit. I did find this one on a Diablotek PHD series:
http://hardocp.com/article/2010/12/17/diablotek_phd650_650w_power_supply_review/9
That may be enough to make you reconsider.
I did find this review of the UL series:
http://www.thinkcomputers.org/diablotek-ul-series-575w-power-supply-review/4/
They basically say it's a crappy power supply, but it didn't explode. The only good mark it got was that it was cheap, and they remark that it's great for a "Backup Crappy Machine." They didn't run enough testing on it to specify if it was actually a decent performing power supply though (unlike HardOCP, which did).
You don't really need 650W in that machine - the 6870 has a TDP of 151W, the CPU is another 100W give or take, the rest of the system is another 100W give or take. That's 350W total peak. AMD recommends 500W.
I know this is a budget system, but really your putting a lot of faith in a power supply, no matter what budget the system is. For less money than the 650W Diablotek costs, you could get a Corsair or Silverstone 500W, or any other various high quality 500W: spend less money for demonstrably better quality. Newegg has the Corsair CX500 with a $20 rebate right now, with a list price of $59 - cheaper even before the rebate, and has the required 6-pin connectors for the 6870. It's not their best power supply, but I'd be more willing to bet on that than anything by Diablotek.
Google Dioblotek UL series. they are a different breed than their other lines. Otherwise I agree with Diabloitek in general.
Personally my latest build has a Diablotek UL in it and so far 1 year under heavy load no problems.t
As far as their PHD series I wouldn't trust it in a bucket of water near anything flamable.
Go ahead google "Diablotek UL" and click everywhere one is available. From 575w,675w775w. Read the comments. Companies can turn over a new leaf.
Do you want a power supply that demonstrably is good? Or merely one you hope might be good.
If you've got a good product, you don't keep it secret. You send it to reputable sites to get good reviews and show the world that it is a good product. Some of Diablotek's power supplies have been so awful that I'd assume all of their power supplies are garbage until proven otherwise. And a user review that says "I pulled 200 W from a 600 W power supply and it didn't blow up" doesn't count as proven otherwise.
I did, where do you think I found the review saying it was a crappy unit but it didn't explode.
It's your money, your rig. You couldn't pay me to put a Diablotek in my personal computer though.
I looked up the memory thing, because I had never heard of anyone having issue with XMP RAM in a AM3 or AM3+ motherboard. Turns out, some people have had issue, in most cases it appears not because the RAM sucks, but rather, because they didn't configure the BIOS to match the specifications of the RAM. XMP is an Intel-specific thing, where they can embed into the memory 3 different timings (stock, performance, "extreme"), and the motherboard can then use that information to auto-configure the timings. Non-XMP motherboards won't have this information, so they won't auto-configure the DIMM to run at the proper timings and voltages. XMP is totally optional, it just provides a mechanism to auto-setup the motherboard. The memory should work at it's rated specifications, provided you tell the motherboard what those are.
I will give you, most of the time, you plug in memory, and if it isn't certain, it goes to some low default clock speed and works. I don't know if XMP interferes with that on AM3 motherboards or not - the user reports were unclear, other than many were able to get the RAM to work fine by manually setting the timings, or perhaps more importantly, the voltage.
As far as Bang for the Buck CPU...
Now I'm curious, so I'm looking as I haven't priced AMD CPU's in a while (Intel's prices never really change).
An FX-4100 is going for $109 right now on Newegg. That's a stock 3.6G.
The only Zosma (4-core Thuban) they carry is $124 at 3.0G. A stock 3.6 Deneb PII X4 is going for $149 - that's a BE, which adds a lil bit to the cost. A low-speed (2.8) Deneb X4 can be had for $94. A Thuban X6 2.8 is $149.
Given the ~20% per-core performance difference, the FX-4100 at 3.6G should be comparable to the 960T PII X4 at 3.0G.
So I have to hand it to you, bang for the buck, looks like the FX-4100 does win if you go clock-for-clock. The last time I had actually looked at Zambezi prices, the FX-4100 was still near $125.
Since the old X4 940 BE has been discontinued, I think you may be right, the FX-4100 may be the Bang for the Buck budget CPU. I don't know that $40 is worth the extra performance to get the only BE Denebs that are still in inventory, and your only looking at equivalent stock performance out of the Zosma or any of the X6 Thubans really.
Decided to throw something together just for fun.
Antec 300 Case $59.99
Hitachi 500 GB $79.99
Lite On DVD Burner $17.00
Asrock 970 Extreme 3 Motherboard $89.99
HIS 6870 Graphics Card $169.99
HEC Zephyr 580W PS $64.99
G Skill Ripjaw Series Memory $29.99
AMD FX-4100 3.6Ghz Quad Core $109.99
Total $621.93
Freebies-
Dirt 3 Game
Black Wired Mouse
Rebates-
$20 Mail in rebate on the Graphics Card
$15 Mail in rebate on the Antec Case
$10 Newegg promotional gift Card
That puts you at about $580 after rebates and still needing an operating system.
Ridelynn you obviously like bending the truth to fit your needs.
No where did the UL review say anything about exploding. I's a budget gaming power supply that happens to be reliable.
Those 12v ripples were pretty minor and easily ironed out by the motherboard.
The Seasonic had the same 12v ripples and could get loud under full load. was based on older PS designs.
The Diablotek 675-UL is 20$ cheaper. Is reliable or you would see more people complainng since it is widely available online and in retail. In general people have nothing bad to say about it, and there are plenty of comments if you look at the online retailers. It has a 3 year warranty that covers component loss. What more do you want.
It's cheap, it won't blow up, Runs cool, and it's quiet. It has over voltage and overcurrent protection, 105c capacitors, Plenty of connectors. Only downfall is 3% or so less efficiency.
(THE UL SERIES. I'M NOT TALKING ABOUT THE PREVIOUS GENERATION)
I doubt the OP will be pulling 675w. This supply will do fine. For just 63$
I just read a review of the Rosewill 630w green and it is much better at 70$
http://www.hardwaresecrets.com/article/Rosewill-Green-Series-630-W-RG630-S12-Power-Supply-Review/881/1
http://www.newegg.com/Product/Product.aspx?Item=N82E16817182200
There is room in the under 700$ for this in my last build, and it is a better PS than the Diablo UL
The Zosma X4 is a failed Thuban. I'm kinda thinkin I don't want a failed anything.
I do? I paraphrased in my response - as it wasn't in quotes like the other review quote was... however, other Diablotek power supplies ~have~ exploded. The UL is apparently the only one that doesn't. It is a budget power supply. I have yet to see an actual review that deems it reliable, and from all indications the Diablotek UL is a rebranded Great Wall PSU, but I can't confirm the model. The closest I could find is an OCZ ZT 650W by Great Wall, which failed the HardOCP power supply testing because it couldn't deliver past 75% under nominal conditions without completely shutting down, nor could it meet it's rated usage temperature. Granted - I don't know if this is the same as Diablotek UL - I suspect it may be, but that's as reliable as I can say right now.
And my take is, no power supply is considered reliable until it's been tested by a reputable site and proven otherwise. I can't prove anything about the Diablotek UL series - I can't even prove who the OEM is.
To bend your own quote... the UL is possibly/allegedly the only Diablotek product that doesn't truly suck. I'm kinda thinkin' I don't want to buy anything from a company that puts out 99% suck and 1 maybe ok product. Especially when there is better for cheaper. I don't know why you even still defend them. It's like trying to say Charlie Manson isn't all bad because he wrote a couple of songs.
Although, the Rosewill (rebranded ATNG) is a much better power supply - I could at least find a real review on it (as a Nexus brand) that passed, I agree. But..
I still don't know why you want a 600+W power supply in the first place. A really good 500W power supply is still overkill for an AMD system with a 6870. This is a budget system after all right? Why pay more for a power supply from a junk manufacturer just because it has a sticker on it that has a bigger number, or pay for more power than your ever going to use. If you have it - put that money towards quality or longevity, not toward a number on a sticker.
I really wouldn't look past the Corsair CX500 or Seasonic at $59 (with $20 rebate) or Antec Basiq 500W for $54. Quiz can probably find a suitable decent power supply for even cheaper, I don't do enough research on what is rebranded what.
http://www.hardocp.com/article/2007/10/03/50_power_supply_roundup/17
Every CPU (and GPU really) is a failed something or another - that's how they make them. Dozens on a wafer, those that test well get binned high, those that test worse binned lower. All the same silicon. There are many, many reports of people taking Zosma's and unlocking the 5th and 6th cores and having a full-blown X6. Just like people unlock 6950's to 6970's. Just like nVidia snips ROPs and TMU's and magically has the 580, the 570, the 560Ti, etc... Just like Intel has 5 different Core i5 1155's and 3 different Core i7's and the entire i7/i5/i3 lineup in the first place...
I'm not debating that Diablotek as a startup company was a complete failure, It was.
I'v worked at several IC companies in phoenix so I know that internal managment can drastically affect the quality of components as well as work conditions.
Such as having quality through hole soldering equipment and personell.
Upgraded QC, Higher standards for quality components. Better testing equipment.
I don't know where the problem was with older Diablo PS lines. The reviews were such that something must have changed if the UL's aren't blowing people's computers up by the hundreds.
Companies can change when managment is replaced or a big investor pops up, or it's bought out.
It appears Diablotek is owned by Evercool which makes some decent stuff. I would imagine the parent company tossed Diablotek managment out and put someone decent in charge, who knows. UL is a step in the right direction. mabye we'll see a new line come out with good reviews. At least UL passed UL lol
Something isn't quite right with the Zozma imo. I know they are Thuban cores but it's performance stinks.
http://www.cpubenchmark.net/cpu_lookup.php?cpu=AMD+Phenom+II+X4+960T
954 per core for the zozma and 971 per core for Thuban @ 3ghz. Memory bandwidth would dictate 4 core scaling should be better than 6.
They did a pretty good number on the Rosewill green. and the reviewer seems really good. he even goes into component analysis.
http://www.hardwaresecrets.com/article/Rosewill-Green-Series-630-W-RG630-S12-Power-Supply-Review/881/1
How do you know that the Diablotek UL series happens to be reliable? There are a lot of problems that a power supply can cause short of exploding. If the UL series is so good, then why isn't it 80 PLUS certified? 80 PLUS certification doesn't automatically mean that the power supply is good, but not having it almost invariably means that it's junk. The requirements for certification are pretty low, so if a power supply can't even meet such a low bar, something is amiss.
Hardware Secrets does do good power supply reviews, but you should be aware that they're pretty generous with their "golden award". That's why you have to read the review. If Hardware Secrets gives a power supply a golden award and Hard OCP rates the same power supply a "fail", they're not necessarily contradicting each other.
For the Rosewill Green, the review says the power supply is kind of all right, but hey, it's cheap, so we'll give it a golden award. The review is also from 2009, and a part that was pretty good in 2009 would be considered significantly less good by today's standards.
But let's look at the review. All Teapo capacitors, which aren't that bad, but aren't that great, either. Some rated at 85 C, and some at 105 C, which again, isn't that bad, but isn't that great, either. Ripple on the +5 V rail is about 3/4 of spec, while on the +12 V rails, it's about 2/3 of spec. That does constitute staying within spec, but a lot of power supplies can do a lot better than that. Energy efficiency barely stays over 80%.
And look at the wattage rating. Supposedly it's 630 W total, but only 611.5 of that is on the +12 V, +3.3 V, and +5 V rails, so it's pretty much impossible to pull 630 W from it without running it out of spec. That's not the sort of thing that reputable power supply companies do. The +12 V rail is officially rated at 600 W, but a component analysis finds that its rectifiers are only rated at 514 W. The power supply supposedly has a "peak" power of 730 W (any mention of peak power is a bad sign, by the way), but shuts down if you pull 670 W from it.
Now, much of that is merely overly optimistic wattage ratings. If they call it a 550 W power supply rated at 42 A on the +12 V rail, then the criticisms of the above paragraph disappear. But again, inflating wattages like that is something that reputable power supply companies don't do.
CPUMark doesn't measure memory bandwidth - it's strictly CPU (which is a big part of the reason why it only tells a very small part of the story). A difference of 17 per core comes out to 1.75% - that's within the margin of error.
The X4 960 falls exactly where it should between the 955 and the 840 benchmarks, and is in line with the X6 1075T benchmark.
The real outlier is that the FX-4100 gets a passmark score that is significantly higher - 4343 vs 3819, or 1086 per core, yet in real world use, is about ~20% per core slower than Thuban (or rather, needs ~20% higher clock speed to get the same results). So take that difference of 17 between the 960 and the 1075T for what it's worth, and take passmark scores in general for what they are worth.
I don't see anything at all wrong with it, or with any of the Phenom II line really. But at a price of $124, you can get the same performance out of a FX-4100 for $109 on the same motherboard. That is the only point I was illustrating, mainly to myself. Which is what you were arguing for in the first place.
Bulldozer is slower per core than Zosma, but not necessarily 20% slower. How well threaded the program is will affect the per core performance, too. Each Bulldozer module has one integer scheduler shared between two cores, so using both cores from a module won't get you double the performance of only using one core. AMD estimates that two cores should get you 80% more performance than one in programs that scale perfectly to two cores. Thus, one would expect an FX-4100 to fare better against a Phenom II X4 in programs that only scale to one or two cores (so that they don't have to use both cores from a module and deal with the shared scheduler) than in programs that scale flawlessly to four cores.
Bulldozer also has some instructions that Zosma doesn't, most notably SSE4, AVX, and FMA4. (Also AES-NI, but that's very specialized and will never be used for games.) AVX and FMA4 are barely used at all thus far, and SSE4 is only a few years old as well. But if those instructions become more commonly used in the future, Bulldozer will see performance gains from it (as will Sandy Bridge for SSE4 and AVX, but not FMA4) and Zosma won't.
There are a lot of pros and cons for FX-4100 versus Phenom II X4 960T. As I said earlier, I'd be shocked if there aren't a lot of programs in which each beats the other. So I just picked whichever of the two was cheaper.
These are just some artifical benchmarks, I'll grant you. But
http://www.hardocp.com/article/2011/10/11/amd_bulldozer_fx8150_desktop_performance_review/4
It's showing the FX8150 to be slower per core by nearly 20% in almost every test, to a Thuban 1100T. The FX8120 is more than 20% in every test, although it has a lower base clock.
1100T is a 3.3G base clock, FX8150 is 3.6, FX4100 is 3.6
So I interpret that to mean that a 3.0-3.2 Thuban will perform about the same as a 3.6 Zambezi, core for core.
Now going through the rest of the benchmarks on that page, none of which are actual gaming unfortunately, it does go back and forth a bit.
Gaming, however, is mainly based on per-core performance, which is why I place so much emphasis on that first link.
Here are a few real gaming benchmark numbers that test my hypothesis:
http://www.bit-tech.net/hardware/cpus/2011/10/12/amd-fx-8150-review/9
ARMA II - an FX-8150 OCed to 4.8 roughly matches a 1100T OCed to 4.2 - a difference of about 14% in clock speed.
http://www.legionhardware.com/articles_pages/amd_fx_8150fx_8120fx_6100_and_fx_4170,6.html
This one is interesting, because they test 4, 6 and 8-core Zambezi's. In pretty much every case, it just came down to a matter of Ghz rather than cores (no surprise there), and the Zambezi's were ~generally~ behind the Denebs and Thubans by about the same margin - roughly 15-20%, with a couple of outliers.
http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/8
Anadtech - same thing. The same delta in per-core performance more or less.
Core for core, for gaming, Zambezi needs more Ghz to accomplish the same job. Fortunately, it can attain higher speeds in general, but in many cases, it can't get fast enough. Sure there are a lot of exceptions, especially when you look outside of gaming, but just looking at gaming, Zambezi is generally slower clock for clock and core for core than Thuban/Zosma, and in many cases, even Deneb - and that's what matters in gaming.
All of that being said... if the price for a 3.6G Zambezi is cheaper than a 3.0G Zosma/Thuban/Deneb, then it wins, even if it is slower per clock, it has the base clock to make up for it and come ahead in performance per dollar. Just comes down to which is on sale that day, in the budget arena.