Oh So Efficient: FSP Announces 400 W 80 Plus Titanium PSU

[JackNaylorPE]

Again you refute your own statements./ First you claim that it doesn't exceed 140 watts, then post a reference saying it's 195 watts @ 1.319 V..... You can't argue both sides of the coin

As you correctly indicated chips vary and the example I used required 1.5 volts. Now we know that increases in voltage represent an exponential increase in power as you also previously opined .

195 * (1.50 / 1.319)^2 = 252 watts

Let's look at some core temperature heat numbers:

4.3 Ghz @ 1.225 volts Delta T = 28.25
4.6 Ghz @ 1.385 volts Delta T = 45.25 (1.131 times the voltage = 1.602 times the heat. Based upon temps, the power generated would appear to be greater than a square function.

The reference also refutes your RAM claims as it shows a total of 6.14 watts for 4 modules.

So let's get back the point..... please explain, using your own numbers, how an overclocked 5960 at 1.5 volts works on a 400 watt PSU and how separating the numbers into any categories you want is going to change the size of the PSU or a Radiator serving the system.

Even your own reference.... at the lower voltage and average load has it exceeding the power rating of the PSU at average load

Your Source data @ average not peak load and reduced voltage
192 watts for CPU
48 watts for VRM controlling the CPU

Uncontested
287 watts for overclocked GPU

Your quotes
30 watts for fans
15 watts + say 5 base load (assumed) for RAM

Other
SSD + HD 10 watts
Optical 5 watts

Subtotal = 592 watts
+ 15% for capacity aging = 89 watts
Total = 681 watts

So looks like no matter how you try and divide up the numbers, .... which has no relevance to this topic .... according to **your** own numbers, this PSU can NOT power the stated system.

 

[smeezekitty]

85W is lost in the motherboard's VRM.

I find that hard to believe. 70% efficiency is dismal for a buck converter.

 

[RedJaron]

As Invalid points out, these aren't really suited for budget builds. Super-high efficiency comes at a cost. A 450W FSP Aurum 92+ Platinum PSU costs around $100 - this Titanium unit would probably be at LEAST that expensive. A 450W FSP 80 Plus Silver PSU can be had for $60 or less. In a budget build $40-50 can make a big difference. Outside of FSP you can get also find good-quality 80 Plus Bronze or 80 Plus units for even less (or with higher output per dollar if your build necessitates it).

That really depends on final pricing as well as user budget and use case, now doesn't it? Not to mention I already stated as much. Yes, I was probably a bit overeager in saying this would be a top recommendation for low-budget builds. At the time I was hoping this would be somewhat close to FSP's 80 Gold 400W unit that is often available ~$70, so a $80-$90 range. But yes, their Plat 450 is usually around $100 so who knows.

Someone who wants to completely min/max a gaming rig for a given budget would likely be content with the biggest capacity 80 Bronze unit they could get for $60. But some people value efficiency and quiet over raw capacity. A highly efficient PSU will generate less heat, therefore ( hopefully ) the fan doesn't need to spin as much, and all of this adds up to less wear on the unit, hopefully equating to a longer life. Such specs would be highly desirable to someone looking to build a modest spec'd computer on a low(-ish) budget, particularly if it were going to stay powered on for long periods of time. The low-heat would also be nice in a smaller case where cooling and air-flow can be at a premium.


BTW, why do you up-vote every single one of your posts?

 

[InvalidError]

I find that hard to believe. 70% efficiency is dismal for a buck converter.

You will always have some configuration of buck and synchronous rectifier FETs active and passing the sum of load current peaks when they occur. During those peaks, you get instantaneous power that goes something like this: 0.003 ohms RDSon * 150A^2 = 68W split between 6-10 phases depending on VRM design. This does not include switching losses, copper losses on the PCB and through buck inductors, socket losses, etc.

When dealing with very low voltages and very high currents, efficiency can quickly take a dump - it does not take much to add 0.001 ohm or more along a power trace or plane.

 

[smeezekitty]

When dealing with very low voltages and very high currents, efficiency can quickly take a dump - it does not take much to add 0.001 ohm or more along a power trace or plane.

That's true. It is easy to forget that it is dealing with less than 1.5V at 50-150A where every milliohm counts.

Yet from what I have seen, most VRMs only have small or even no heatsinks on the FETS. If it were dissipating that much, how
could it be without a large heatsink?

 

[InvalidError]

Yet from what I have seen, most VRMs only have small or even no heatsinks on the FETS. If it were dissipating that much, how could it be without a large heatsink?

How many DPak or similar packages do you see around a 6-10 phase VRM? At least a dozen since each phase has a high-side and a low-side driver. That averages out to less than 5W per device, which is not so horrible - it should be within the realm of what you can get away with by stitching the motherboard's ground and power planes together to spread heat around without a heatsink - stitched copper pours/planes can be surprisingly effective heatsinks. To move over 100A peak between PCB layers/planes, you will need tons of via stitching anyway, might as well put them to good use.