Why not both? The DUAL current steam machine.

[Runamok81]

It occurs to me that all of the smallest, most powerful designs produced this past year came way of pairing an mITX graphics card with an internal DC-DC power board instead of a traditional power supply. By now, we are all aware that both AMD and nVidia will have flaghsip cards of the mini variety, thanks to HBM2 memory.

I don't anticipate the HDPLEX 250W will be able to power these new cards. So, I'm wondering if a dual current steam machine could be the solution. What a I mean by dual current, is a chassis that could be configured to use either

THIS

OR THIS

So the chassis would have both mounting points inside for the HDPlex and mounting points on the I/O plate for the ATX bridge. Allowing a user to pick power solution that meets his/her design choices. The ATX bridges are sold by HDPLEX and they are reasonably price. Several passively cooled external ATX/SFX power supplies exist which could be used for this.

Thoughts? Anyone ever seen this implemented?

 

[Bieberfever]

That bridge is sick. It's exactly what I've been looking for.

 

[Runamok81]

I know, but I haven't seen anyone to complete a build with one yet. Which case would you use it with? Which power supply would you pair with it?

 

[SaperPL]

I don't get it - What's the point of making nice SFF case when you'll put a huge ATX/SFX PSU brick next to it?

 

[Runamok81]

I don't get it - What's the point of making nice SFF case when you'll put a huge ATX/SFX PSU brick next to it?

It would be the only nice SFF case with a flagship level GPU.

I'm watching the GPU market. AMD is moving down from 28nm to 16/14nm FinFet. That means reduced TDP across the baord for them. AN area where they are currently struggling. I'm expecting the majority of their line up will now fit into an mITX constrained case. Same with nVidia, but on the size front instead of efficiency. They are adopting HBM2, meaning their cards will be smaller. They are already efficient. So, this year, both manufactures will have lineups that more suited for an mITX constrained design, thanks to process improvments and HBM2 memory. However, there is a chance that the tippy-top flaghsip cards (next gen Fury and Titan) will still need more than 400W. In that case, this case design could house them, albeit with an external power supply. That "flagship option" is something that no current DC-DC case offers. It's either go big or go home. This fixes that. Create a next gen case that covers the whole gamut. From power-sippers, to power users all in the smallest mITX case

 

[SaperPL]

Once again - what's the point if you're leaving a brick standing next to it?

I can understand it only for the mobility of such config if you were to easily split those and put into the luggage but except for that it's madness.

If you want go this way then you could go for smallest case with support for low profile card and move all other things to the external box - gpu, psu and hard drives but that doesn't mean that solution is reasonable

dc bricks make sense because they are made for external use and are quite small.

 

[TheHobbyist]

The other concept that is being missed is that the size of the card is not only a factor of the pcb, but a factor of the cooler that is required to adequately cool it.

 

[Runamok81]

Once again - what's the point if you're leaving a brick standing next to it?

I can understand it only for the mobility of such config if you were to easily split those and put into the luggage but except for that it's madness.

If you want go this way then you could go for smallest case with support for low profile card and move all other things to the external box - gpu, psu and hard drives but that doesn't mean that solution is reasonable

dc bricks make sense because they are made for external use and are quite small.

Do you think that people would be willing to increase the size of their adapter in exchange for a more powerful GPU? Which is larger, an SFX power supply or 300W+ laptop power supply? What's the size difference?

The other concept that is being missed is that the size of the card is not only a factor of the pcb, but a factor of the cooler that is required to adequately cool it.

What size coolers do you expect will be on the next generation of cards? Do you think the next generation of cards will be more or less power efficient?

 

[Aibohphobia]

It's either same performance for lower TDP or more performance for the same TDP. I'm betting they'll go for the latter.

With VR around the corner both camps will want to wring as much graphics horsepower out of their next-gen cards as possible.

 

[Runamok81]

It's either same performance for lower TDP or more performance for the same TDP. I'm betting they'll go for the latter.

With VR around the corner both camps will want to wring as much graphics horsepower out of their next-gen cards as possible.

The power savings going from 28nm to 16nm/14nm paired with a move from GDDR5 into more efficient HBM2 is going to be bigger than most folks anticipate. Remember, this is the first transistor shrink for GPUs in four years. What to expect? From TSMC's website.

TSMC’s 16FF+ (FinFET Plus) technology can provide above 65 percent higher speed, around 2 times the density, or 70 percent less power than its 28HPM technology. Comparing with 20SoC technology, 16FF+ provides extra 40% higher speed and 60% power saving.

So, sure... Manufacturers could use that power savings to produce a chip that runs twice as fast at the same TDP, but that is a hypothetical that market trends have NEVER followed. Instead what we've seen with both CPU and ARM are chips with lower power requirements and lower power ceilings.

 

[Phuncz]

I truely hope Nvidia and AMD both will use this to lower power consumption and increase performance at the same time. If we take the hypothetical 70% increase (hard to believe), rougly 35% lower power consumption and 35% performance increase is nothing to sneeze at. It would basically mean a GTX 970 with the performance of a GTX 980 Ti and the power consumption of a GTX 960. It's basically what the Radeon Nano has done, but moar extreme.

 

[SaperPL]

Do you think that people would be willing to increase the size of their adapter in exchange for a more powerful GPU? Which is larger, an SFX power supply or 300W+ laptop power supply? What's the size difference?

This is the 330W Alienware brick

I can't find the precise dimensions but I'd say it's 160 x 70 x 35 mm tops which is around 0.4L. The SFX is 125 x 100 x 64 which is 0.8L and SFX-L is 125 x 130 x 64 which is 1L.

Other than volume there are few different problems. Even if you get "passively" cooled psu it still need some minimal airflow made by other fans. This means you can't just enclose the psu blocking it's holes and that compared to brick is worse at least in terms of the looks.

Next thing is that since the bricks are quite slim and there are usually some tight spots below and behind furniture you can easily hide them. Once again you can't even think about doing that with SFX because you'd block the airflow.

Last thing is that instead of nice and elegant single cable you get bunch of thick sleeved ones and even if you sleeve them together it'll still end up looking like a sausage.

 

[Runamok81]

This is the 330W Alienware brick

I can't find the precise dimensions but I'd say it's 160 x 70 x 35 mm tops which is around 0.4L. The SFX is 125 x 100 x 64 which is 0.8L and SFX-L is 125 x 130 x 64 which is 1L.

Other than volume there are few different problems. Even if you get "passively" cooled psu it still need some minimal airflow made by other fans. This means you can't just enclose the psu blocking it's holes and that compared to brick is worse at least in terms of the looks.

Next thing is that since the bricks are quite slim and there are usually some tight spots below and behind furniture you can easily hide them. Once again you can't even think about doing that with SFX because you'd block the airflow.

Last thing is that instead of nice and elegant single cable you get bunch of thick sleeved ones and even if you sleeve them together it'll still end up looking like a sausage.

I do own the 350W Vodoo Firebird, and it is the same volume as SFX-L. One liter. I think that the 330W Dell adapter would be smaller, but not by much. Until someone can provide measurements, lets assume it is 0.8 liters, the same size as SFX, but to your point a different dimension. If you wanted to dangle a passive PSU behind a desk, yes, you would have have to pull the desk out slightly.

As far as passive PSU's needing collateral airflow, I wasn't aware that was required. I would think that requirement would preclude these types of PSUs from being used in passively cooled PCs, their target market. Do we have any examples stating that collateral airflow is required for passive PSUs to function?

About the cables, yes, they would have a larger circumference. And I do think it is possible to have a passive PSU. Take a look at Project Quantum and its external, passive PSU.

That power brick is pushing 1200W passively cooled. It stands as an example of what is currently possible, and its something I think has room for improvement. Make a horizontal design, use some of the available SFX/ATX PSUs and keep the wattage under 500W. I guess I don't view the concerns you've voiced as deal breakers. I think it's quite possible, and we have existing examples that simply need refinement.

 

[SaperPL]

Here's another photo:

http://www.amazon.co.uk/GENUINE-Original-Alienware-Adapter-Dell/dp/B00AK75R6U

And your brick for comparison:

It looks twice the width of alienware one.

I can understand that this may be favorable for your brick because of better cooling or peak wattage etc, anyway volume is not everything, thickness also matters as I said before.

As for passively cooled psu's - if they weren't requiring any airflow then they wouldn't have inlets and outlets. Of course if you have a psu with a lot of headroom and low power components in your system then total passive solution might work somehow since hot air will slowly escape itself from the psu by convection.

Project quantum had custom brick and you want to use the atx/sfx standard psu so you'll have a bit more cables that won't be packed so gracefully. Also note that it wasn't Fury X2 but Fury X with single gpu:

http://www.windowscentral.com/heres-whats-inside-amds-incredibly-powerful-project-quantum

So the brick was like 600W not 1200W.

Anyway the big external brick is still not an elegant solution.

 

[Runamok81]

I truely hope Nvidia and AMD both will use this to lower power consumption and increase performance at the same time. If we take the hypothetical 70% increase (hard to believe), rougly 35% lower power consumption and 35% performance increase is nothing to sneeze at. It would basically mean a GTX 970 with the performance of a GTX 980 Ti and the power consumption of a GTX 960. It's basically what the Radeon Nano has done, but moar extreme.

Historically, that node jump has yielded a real-world 30% power draw reduction. The current flagship cards recommend a 600W PSU. I'm predicting that the next gen flagships will only require a 500W or less PSU. SFX territory.

 

[Aibohphobia]

So, sure... Manufacturers could use that power savings to produce a chip that runs twice as fast at the same TDP, but that is a hypothetical that market trends have NEVER followed. Instead what we've seen with both CPU and ARM are chips with lower power requirements and lower power ceilings.

Oops, I should have qualified that the flagship cards I think will favor performance over lower power consumption. As a whole it'll probably drop in power consumption.

About the cables, yes, they would have a larger circumference. And I do think it is possible to have a passive PSU. Take a look at Project Quantum and its external, passive PSU.

Do we know if it's actually passive? I've never seen a close examination of the Quantum brick and it's big enough to just have a standard ATX PSU in there.

 

[Runamok81]

Here's another photo:

http://www.amazon.co.uk/GENUINE-Original-Alienware-Adapter-Dell/dp/B00AK75R6U

And your brick for comparison:

It looks twice the width of alienware one.

I can understand that this may be favorable for your brick because of better cooling or peak wattage etc, anyway volume is not everything, thickness also matters as I said before.

As for passively cooled psu's - if they weren't requiring any airflow then they wouldn't have inlets and outlets. Of course if you have a psu with a lot of headroom and low power components in your system then total passive solution might work somehow since hot air will slowly escape itself from the psu by convection.

Project quantum had custom brick and you want to use the atx/sfx standard psu so you'll have a bit more cables that won't be packed so gracefully. Also note that it wasn't Fury X2 but Fury X with single gpu:

http://www.windowscentral.com/heres-whats-inside-amds-incredibly-powerful-project-quantum

So the brick was like 600W not 1200W.

Anyway the big external brick is still not an elegant solution.

Project quantum used a dual Fiji GPU. Photos are not measurements. Passive PSUs don't require fans.

 

[Ocellaris]

Project quantum used a dual Fiji GPU. Photos are not measurements. Passive PSUs don't require fans.

Keep in mind Project Quantum was announced at the same press event where AMD told everyone the Fury video cards were overclocking beasts. AMD also told us you need 8GB VRAM for 4K gaming while launching Fury with 4GB as a 4K card. And furthermore they acted like the 390 and 390X were not rebadged 290 and 290X cards.

I have high hope for Fiji, I just wouldn't place a lot of confidence in press information for unreleased AMD products.

 

[Runamok81]

Here's another photo:

http://www.amazon.co.uk/GENUINE-Original-Alienware-Adapter-Dell/dp/B00AK75R6U

And your brick for comparison:

It looks twice the width of alienware one.

I can understand that this may be favorable for your brick because of better cooling or peak wattage etc, anyway volume is not everything, thickness also matters as I said before.

As for passively cooled psu's - if they weren't requiring any airflow then they wouldn't have inlets and outlets. Of course if you have a psu with a lot of headroom and low power components in your system then total passive solution might work somehow since hot air will slowly escape itself from the psu by convection.

Project quantum had custom brick and you want to use the atx/sfx standard psu so you'll have a bit more cables that won't be packed so gracefully. Also note that it wasn't Fury X2 but Fury X with single gpu:

http://www.windowscentral.com/heres-whats-inside-amds-incredibly-powerful-project-quantum

So the brick was like 600W not 1200W.

Anyway the big external brick is still not an elegant solution.

Sorry to be short with you, but Project quantum used a dual Fiji GPU. Photos are not measurements. Passive PSUs don't require fans.

I can see that this is becoming a "pro vs con" power brick discussion for you. let's not have that. Many folks are effectively using external laptop adapters in their designs. I'm looking for non-biased input about the engineering challenges. Hopefully using accurate information and measurements.

EDIT: Also, I do want show some level of appreciation for your expertise. I need someone attacking the feasibility of using an external PSU in order to refine the idea. Thank you. Maybe, I'm a bit defensive about someone calling my imaginary design an ugly baby. So, thanks for the reality check, but I still think its feasible.

 

[Runamok81]

Keep in mind Project Quantum was announced at the same press event where AMD told everyone the Fury video cards were overclocking beasts. AMD also told us you need 8GB VRAM for 4K gaming while launching Fury with 4GB as a 4K card. And furthermore they acted like the 390 and 390X were not rebadged 290 and 290X cards.

Sounds like you're rallying against AMD marketing here. I can sympathize. nVidia did similar things with their 3.5GB 970 GTX fiasco. That's marketing. I don't think it would be fair to say that these marketing tactics mean that they lied about quantum? You think it only had a single Fiji GPU?

 

[Ocellaris]

Sounds like you're rallying against AMD marketing here. I can sympathize. nVidia did similar things with their 3.5GB 970 GTX fiasco. That's marketing. I don't think it would be fair to say that these marketing tactics mean that they lied about quantum? You think it only had a single Fiji GPU?

I definitely think AMD has had 2x Fiji cores in a SFF system for engineering tests. Both Fiji cores would need to be very close together on one PCB. I just don't think that means a retail product will come from it as presented with Project Quantum with 2x Fiji. The way PQ is put together, it isn't something that would transition to hardware a user could assemble themselves. The water block sandwich between the GPU and CPU would almost certainly be problematic for custom systems.

 

[SaperPL]

As for the photos - I could go for pixel measurements based IEC connector reference dimensions to prove that, but that's not the point. Roughly it should be close to what I've already noted.

The thing about 2x Fiji inside Project Quantum is that AMD stated they had them inside but every reviewer was given the unit with single fury X. Also note that cad images shown on videos and photos include only single chip card like the single core fury X. If I remember correctly Fury X2 was shown at the same time the Project Quantum was presented and AMD didn't provide TDP info at that time yet for the X2.

https://www.youtube.com/watch?v=fBu5KjepWTw

And no - you cannot fit two cards in there because it's simply something like QinX watercooling sandwitch with single block for both gpu and motherboard.

To me it looks like its AMD's pure marketing and maybe they even came last minute with card from team making fury X2 to project quantum team and asked "hey guys, will it fit this card?" And they said "well .. umm... yeah, it will physically fit, but..." and at this point the marketers took off and went with that to the stage without regard for the power consumption and cooling.

 

[Runamok81]

As for the photos - I could go for pixel measurements based IEC connector reference dimensions to prove that, but that's not the point. Roughly it should be close to what I've already noted.

The thing about 2x Fiji inside Project Quantum is that AMD stated they had them inside but every reviewer was given the unit with single fury X. Also note that cad images shown on videos and photos include only single chip card like the single core fury X. If I remember correctly Fury X2 was shown at the same time the Project Quantum was presented and AMD didn't provide TDP info at that time yet for the X2.

https://www.youtube.com/watch?v=fBu5KjepWTw

And no - you cannot fit two cards in there because it's simply something like QinX watercooling sandwitch with single block for both gpu and motherboard.

To me it looks like its AMD's pure marketing and maybe they even came last minute with card from team making fury X2 to project quantum team and asked "hey guys, will it fit this card?" And they said "well .. umm... yeah, it will physically fit, but..." and at this point the marketers took off and went with that to the stage without regard for the power consumption and cooling.

If you could use pixel measurements to show that the 330W Dell is 0.4L instead of my guess at 0.8L, that would be helpful. Also, where did you get that photo of the single fury Quantum. Maybe the reviewer that took the photo would have adressed why they got a single Fury? If we're lucky the same author may even have gotten a look at the PSU to guess at its wattage. Thanks for your help.

 

[Aibohphobia]

The photo and video are from PC World's tear-down: http://www.pcworld.com/article/2973...y-and-powerful-project-quantum-dissected.html

Basically one of the Quantum demo units failed for whatever reason so AMD let them take it apart to examine it. The power brick wasn't sent with the unit.

The serial number of the unit was "5" and most of the custom parts on the unit PC World disassembled were also marked "5" except the Fury X which was marked "4". Could be coincidence, could be AMD didn't want the original dual-GPU card out in the wild but I suspect that most of the prototype Quantums are actually just using a single Fury X. All of the CAD models shown in the AMD video for the Quantum show a single-GPU card.

 

[Runamok81]

Thanks Aibohphobia. I think Saper linked to a Windows Central article that didn't have the relevant details about the single fury. Looks like they addressed it in the original article. Right below the photo was this text

Conspiracy theorists unite
I know that’ll inspire the tinfoil-hat conspiracy theorists to say AMD never based any of the Project Quantum machines on a dual Fury X, and they faked the Apollo Moon landing too.

While I have nothing to contradict that theory, there are plenty of logical reasons why Number 5 isn’t a dual Fury X machine. The most obvious one is that AMD didn’t say all Project Quantum machines were dual Fury X-based. Even if only one Project Quantum were dual-GPU, that’d be enough. And because a dual card won’t ship for another few months, it’s unlikely AMD would leave its upcoming hotness outside its control.

Later on, they had this blurb about estimated PSU wattage.

One way AMD saved space was to move the PSU outside of the case. This meant a lot of tricks internally, but an external power brick on a desktop case isn’t unheard of, nor a bad idea. It gets the heat from the PSU out of the system and lets AMD run a much larger unit. Unfortunately, AMD didn’t make the brick available for this autopsy, but from pictures I’ve seen of it, it’s very hefty and I suspect fairly high wattage.

I think Windows Central just haphazardly reposted that article without caring about accurate, engineering details. Shame on them.

Anyways, I found a retailer listing the 330W Dell Alienware power brick 0.9L. Hopefully Saper can do his pixel measurements to confirm.

 

[Aibohphobia]

I came up with ~0.6L

 

[Runamok81]

I came up with ~0.6L

Thanks. Those measurements look more accurate than any of the guesses.

 

[SaperPL]

I stand corrected. My pixel measurements add up to the same thing except for the height which i thing is 35~36mm. Dimensions from a retailer look to me like it's the size of the carton box it's packed in.

Anyway that's still not as thick as SFX and still passively cooled brick.

Also from what I checked it looks like fury X2 would barely but fit with some different angled PEG connectors than the card shown by AMD.

My problem is that they still didn't show any images of that either made physically or the cad assemblies and the brick doesn't look like 1200 passively cooled monster to me that would be released to everyday user. Also the cooling using 100mm rad doesn't sound like something that can cool 600W of heat off the gpu's and cpu.

 

[QinX]

I believe they used a 180mm radiator, the footprint of Project Quantum is greater than 170x170.
Also the Fury X2 is more then likely dual R9 Nano so the thermal output shouldn't be much higher then a single Fury X. All in all I believe they did have a Fury X2 system making the rounds as it is technically doable on a 600W powersupply.

 

[SaperPL]

That's right, sorry, I've listened to what the guy in video said :|

So considering nano being power starving fury X, fury X2 is going to be two power starved fury X's...
Anyway it kind of makes sense with 180mm radiator's volume to cool that stuff vs 120mm rad for single fury X.

I'm still not sold on the 600W psu pulling doing that considering the power spikes on nano.

 

[QinX]

That's right, sorry, I've listened to what the guy in video said :|

So considering nano being power starving fury X, fury X2 is going to be two power starved fury X's...
Anyway it kind of makes sense with 180mm radiator's volume to cool that stuff vs 120mm rad for single fury X.

I'm still not sold on the 600W psu pulling doing that considering the power spikes on nano.

Considering Linus just build a dual CPU system wtih 7 R9 Nano's that are watercooled(So not thermally limited, always power limited) on 1 EVGA 1600W PSU I'm pretty damn sure that a 600W PSU is going to be plenty.

4790K @ ~100W
Fury X2 @ ~400W
Mobo/Cooling @ ~50W

The secondary capacitors in the PSU can handle the short power spikes.

 

[NFC]

Voodoo 350w = 6.4 x 6.3 x 1.5"
Dell 330w = 7.8 x 4 x 1.7"
PA-9E 240w = 7.8 x 4 x 1"

And because numbers don't really do it justice:

I personally think that SFX and ATX PSUs aren't comparable to AC/DC bricks--not just in terms of size (the Dell 330w although huge is the perfect shape to fit in my backpack's side sleeve) but also in simplicity. You have a single cable that plugs into your PC, rather than a slew of short, messy, MOLEX connectors. I'm not saying having the option is a bad thing, I just don't think it is an apples to apples comparison regardless of volume measurements. This is one reason I am not a fan of the Voodoo brick...it just is too awkward for me (although it looks nice and is the only brick that I have a stable Nano build on).

 

[NFC]

I came up with ~0.6L

Seems dead on.

 

[SaperPL]

Voodoo 350w = 6.4 x 6.3 x 1.5"
Dell 330w = 7.8 x 4 x 1.7"
PA-9E 240w = 7.8 x 4 x 1"

Why imperial units?

Voodoo 350w = 162.6 x 160 x 38.1 mm
Dell 330w = 198.1 x 101.6 x 43.2 mm
PA-9E 240w = 198.1 x 101.6 x 25.4 mm

I personally think that SFX and ATX PSUs aren't comparable to AC/DC bricks--not just in terms of size [..] but also in simplicity. You have a single cable that plugs into your PC, rather than a slew of short, messy, MOLEX connectors. I'm not saying having the option is a bad thing, I just don't think it is an apples to apples comparison regardless of volume measurements.

I agree completely - If it weren't for the messy cables then you could simply take U1 flex-atx psu, sleeve it from the outside with nice thick heat shrinking sleeve to cover metal and sharp edges, and be good to go.

This is one reason I am not a fan of the Voodoo brick...it just is too awkward for me (although it looks nice and is the only brick that I have a stable Nano build on).

Wouldn't it simply fit in that big secondary pocket of the bag?

 

[QinX]

(although it looks nice and is the only brick that I have a stable Nano build on)

Wait, you got it stable? How! What have you tested with?

 

[NFC]

Wouldn't it simply fit in that big secondary pocket of the bag?

It's a close fit--but in general it is so awkward. I used to ship out pelican cases with my full system buildouts and the voodoo I am having a hard time finding one that fits nicely. There are cases deep enough but then the chassis rides on the brick through the second foam layer.

Wait, you got it stable? How! What have you tested with?

Yeah...by at least one definition of stable--predictable. This isn't the thread for it, but I spent a very long day and basically confirmed all your findings.

35w CPU
M.2 SSD
Visiontek R9 Nano
16GB DDR3 1333
R.2 HDPLEX
350w Voodoo (the Dell units failed)

STOCK VOLTAGES FOR EVERYTHING

I could run Valley 1.0 all day long (even ran it overnight) on any combination of settings (Nano scored an average of 1829 on Ultra @ 2560x1440 8xAA) and I could play many of my games without crashing (though MechWarrior Online would trip it every now and then).

I moved up to Furmark and it would run flawlessly (ran it for 2 hours before I gave up) at 2560x1440 @ 8x. But when I turned off AA....BLAMO. Black screen.

Repeated this test ad nauseam with consistent results.

But I have an R9 Nano that performs great and doesn't get too hot...so I need a way to make it work...

*forces 8xAA as override in global settings*



Works perfect so far...Furmark runs stable with AA disabled (or so it thinks) and what I like is that I can run the system at 1080p too...which is WEIRD and I actually want your opinion on...you'd think that halving the resolution would have a huge effect on the power management, but it seems it is the HBM (AA) that causes the issue?

Anyway, my score in Valley (ultra) at 1080p is 3541 at an average watt pull of 259.

That being said, I still am not recommending the GPU as all it takes is one game with gaphical glitching caused by 8x override and your system might not work. Yuck. But I have to make it work for me, and right now, it is!

 

[SaperPL]

What was the framerate with AA? It looks like the AA was making cpu bottleneck the gpu so it's kind of not sure whether you're really taking all power the gpu could really take or was it just running low.

 

[NFC]

What was the framerate with AA? It looks like the AA was making cpu bottleneck the gpu so it's kind of not sure whether you're really taking all power the gpu could really take or was it just running low.

Well at 35w that is all the CPU the particular power brick could handle anyways, so you get what you get. While I did record 44FPS and 66FPS (with 8x and 0x AA, respectively) I didn't record average FPS on my X99 rig because it was irrelevant for my tests.

System 1:

X99 2011v3 @ 4.2Ghz
16GB DDR4 @ 2133
SFX 600w PSU

R9 Nano Valley average (3 runs) 1871 @ 310w
390 Nitro 8GB Valley average (3 runs) 1687@ 430w
285 ITX 941 @ 275w

System 2:

35w i7 @ 2Ghz
16GB DDR3 @ 1333
HDPLEX r.2 + Voodoo
R9 Nano Valley average (3 runs) 1829 @ 255w
285 ITX Valley average (3 runs) 908 @ 221w

Settings: Ultra, 2560x1440, 8x AA

With Crimson it seems that the Nano pulls ahead of the 390 Nitro!

 

[SaperPL]

I meant for the rig with nano and hd-plex.

To me it looks like the nano might be bottlenecked by cpu loaded with AA since there's a big drop - 44fps vs 66fps.

On the other hand I understand you've got i7-4765T which isn't too sloppy and shouldn't bottleneck too much.

Anyway to me, if you want to be sure it's not power starving then you need to let it run without AA on furmark or OCCT at 1080p and that should run at more than 100fps - if it's less then there's something wrong here.

Black screen is definitely something to investigate properly and for me so far it was only due to power starving either because of faulty riser or low power psu.

I've also encountered such "trip overs" with titanfall on slightly power starving R9-270X when either the game crashed to desktop or the driver did reboot.

 

[Runamok81]

NFC, I'm getting a total volume of 0.87L for the Dell 330W from those measurements, correct? I feel like I'm kicking a dead horse here, but external bricks being "too large" was a major concern at the start of this discussion. I want to get that out of the way before I move onto this PSU suggestion. Take a look, please.

The image above is a TFX-350 unit released by Seasonic. It's a hair smaller than SFX. It's comparable in size and shape to that of a laptop brick. It's semi-passive, and it received favourable reviews at SPCR. So, it is very quiet. Power delivery is great, 80+ Gold, and Johnny Guru thinks highly of the unit. This little guy can tame the Nano at price point well below other DC-DC solutions.

Add some custom wires and/or a nice sleeve job and I could see this PSU reliably powering a new design. Anyone have any experience with this unit?