AMD Project Quantum

deanx0r

Limp Gawd
Joined
Apr 24, 2005
Messages
274
https://www.youtube.com/watch?v=1RdISWkIUmM

I hope AMD will partner up with some major case manufacturers to bring this to market. It definitively has the whoa factor.

More info from wccftech:
http://wccftech.com/amd-intros-project-quantum-powered-dual-fiji-chip/

Dual Fury card under the hood.

AMD-Project-Quantum-with-Dual-Fiji-635x423.jpg

AMD-Project-Quantum-ith-Dual-Fiji-GPU.png
 
Last edited:
I hope it takes any ITX board rather than being stuck with an AMD CPU.
 
The prototype they were showing uses a modified/custom Z97 mitx board and an Intel CPU.
True, but I can't see the final product shipping with a competitors CPU. I understand the previous posters sentiment though.
 
Supporting different motherboard might be difficult as it seems the port cutouts are on the case, not to mention the rigid "pipes" that move the water to the tubes and then to the upper chamber. (I might be wrong about that second one though)
 
This looks to me more like a proof of concept/marketing exercise. I'd be very surprised if AMD had any plans to bring anything like this to market.
 
The motherboard has been heavily customized:

fAS9j2f.jpg


Props to djotter for noting that the board is an ASRock Z97E-ITX/AC
http://www.pcper.com/news/General-T...actor-Project-Quantum-PC-Dual-Fiji-Powerhouse


There's a PicoPSU attached to the motherboard and there's a separate power board that connects to the power socket on the rear of the case that probably has the connector for the GPU:

g0DYFrt.jpg



Also, you can see here that the water blocks are sandwiched by the GPU and motherboard:

ggSdYca.jpg


Interestingly the motherboard and GPU are backwards, the GPU connectors are facing rearward as normal but the motherboard IO is facing forward.


Also, the fan is an Enermax Apollish 180mm:

tSjIN66.jpg


Which also explains the red LEDs, those things are bright :p

Also means we're looking at a 180mm rad, though I don't recognize the one they're using.


And finally, going off this picture of the back, here's my estimated dimensions of the unit:
pfC6xkq.jpg

(Expreview.com reporter, please note this is a guess not official specs)

Plenty of room in that box for a 180mm ATX PSU.
 
This actually looks pretty cool. Only interested if it comes with an Intel CPU however.
 
(Expreview.com reporter, please note this is a guess not official specs).
LOL I see what tou did there ;)

A very nice concept this is, good for SFF that AMD focuses on ITX sizes this launch. Full Steam(box) ahead !
 
Also, you can see here that the water blocks are sandwiched by the GPU and motherboard:

So it's like QinX's 3 litre build? Or am I missing something? It seems much the same, except a 180mm rad on top instead of 2x80 on the side.
 
maybe this;

Sure, it's got a longer card :rolleyes:

Read again what I quoted from Aiboh, and remember that the central idea of sandwiching waterblocks between the PCBs was first implemented by QinX in a build log thread on this forum.

What I'm driving at, in case it wasn't obvious: is this another situation of big manufacturers cherry-picking innovative ideas from enthusiast forums?
 
I've thought about sandwiching a block between two components (even three) too over a year ago, I can't believe QinX and I were the only people that did. Apple did recently with the new MacPro. This concept PC seems less "stolen idea" and more "enthusiast engineer" to me.
 
This Quantum prototype gives me so many ideas, i may just have to build one for every room! :)
 
I've thought about sandwiching a block between two components (even three) too over a year ago, I can't believe QinX and I were the only people that did. Apple did recently with the new MacPro. This concept PC seems less "stolen idea" and more "enthusiast engineer" to me.

Shit I thought of it less than a week ago just with bullshit fury rumor mongering. A dual PCB GPU like 7950/9800Gx2/295 with a full block between them.

Oh man....Dual Nano's like that, that'd be 6 powerful inches.
 
I've thought about sandwiching a block between two components (even three) too over a year ago, I can't believe QinX and I were the only people that did. Apple did recently with the new MacPro. This concept PC seems less "stolen idea" and more "enthusiast engineer" to me.

Nice try :p

But dreaming of it and actually getting one together is slightly different things. QinX proved the viability and showed good thermals. This is what I imagine tips an R&D manager between saying "no budget for such nonsense" and "someone did what?! let's get on that asap".
 
This looks to me more like a proof of concept/marketing exercise. I'd be very surprised if AMD had any plans to bring anything like this to market.

I just re-watched the stream, the head of marketing said they were working with partners to bring it to market in the not too distant future. Still things could change.
 
Sure, it's got a longer card :rolleyes:

Read again what I quoted from Aiboh, and remember that the central idea of sandwiching waterblocks between the PCBs was first implemented by QinX in a build log thread on this forum.

What I'm driving at, in case it wasn't obvious: is this another situation of big manufacturers cherry-picking innovative ideas from enthusiast forums?

The way AMD implements it is quite different to the way QuinX did it.

AMD uses a riser that turns the mainboard head over heels, so the I/O of GPU and board are opposite to each other and the two main component sides of their PCBs are facing each other while in QuinX build, they both faced the same way. So what AMD could do is use a single waterblock for both components with no tubing in between while Qunix had to use seperate blocks and use (very short) tubing to connect them.

Also keep in mind that AMD might be working on PQ for quite some time, maybe they started drawing up first concepts way before QuinX did his first post.

EDIT: Holy crap I just noticed the comically large power brick. They're not impressing me with that one :D
 
Last edited:
You can build a smaller system with the A4-SFX if they release the fury x2 with aircooler.
Also they don't use default parts for the Quantum, so the system is not really upgradeable.

By the way, if you look at the pictures you see that they use two pcbs for the gpu only the bottom pcb has videooutputs. Both pcbs are connected to one riser.
 
I don't care about the power brick. I'll probably really like it if it turns out to be of quite high wattage.
 
We should maybe zoom out for a bit, since this is just a project AMD did to show off a proof of concept that won't even see the light of day until next year. We'll see this with Zen CPUs most likely in it's real form we should be expecting a more serviceable product too. No need to bash this into the ground because of some "how it's made" photos.
 
AMD uses a riser that turns the mainboard head over heels, so the I/O of GPU and board are opposite to each other and the two main component sides of their PCBs are facing each other while in QuinX build, they both faced the same way. So what AMD could do is use a single waterblock for both components with no tubing in between while Qunix had to use seperate blocks and use (very short) tubing to connect them.

Thank you, this helps my understanding of the differences better :)
 
You can build a smaller system with the A4-SFX if they release the fury x2 with aircooler.
Also they don't use default parts for the Quantum, so the system is not really upgradeable.

By the way, if you look at the pictures you see that they use two pcbs for the gpu only the bottom pcb has videooutputs. Both pcbs are connected to one riser.

I don't think size was the main concern here. Their system will probably be quite a bit quieter than one in the A4, and the design is a huge point as well. But yeah, you're right in that the A4 is a lot smaller.

Well it depends on what you mean by default. The GPU is their own Fury X2, the mainboard is a modded X99-ITX/ac, so you can upgrade the CPU easily and could at least sidegrade the GPU to a third party version if they keep the layout. The SSDs seem to be standard components as well, so those can be upgraded as well.
But yeah, not as easily upgradable as a PC in a regular case.

I can't really follow you with your last point. Where do they have two PCBs for the GPU? I only see a mainboard at the top and a GPU at the bottom.

We should maybe zoom out for a bit, since this is just a project AMD did to show off a proof of concept that won't even see the light of day until next year. We'll see this with Zen CPUs most likely in it's real form we should be expecting a more serviceable product too. No need to bash this into the ground because of some "how it's made" photos.

I hope we'll see mITX boards for AMDs high end sockets in that light.
 
Before Broadwell, AMD had a good solution for the low-end gaming market with it's APUs. After Broadwell, not so much.
So yeah, 'desktop market' is the correct term these days.

Enterprise server market have been for a long time Intel's monopoly and I personally care less. It's a whole different world and the two are very different, especially with VM's allowing any combination of server and client to work.

But I hope AMD is also going to be present when Living Room PC's start entering the consumer's minds. Zen and HBM APU's do promise amazing stuff, along with DX12 allowing the use of all that GPU power.
So I hope this Project Quantum is AMD's way of saying that they'll be there and ready.
 
Last edited:
Sure, it's got a longer card :rolleyes:

Read again what I quoted from Aiboh, and remember that the central idea of sandwiching waterblocks between the PCBs was first implemented by QinX in a build log thread on this forum.

What I'm driving at, in case it wasn't obvious: is this another situation of big manufacturers cherry-picking innovative ideas from enthusiast forums?

It could be possible buddy, that's why if I was a an author and a creator of an SFF project, I would definitely make patents for my own designs and creations....
 
Before Broadwell, AMD had a good solution for the low-end gaming market with it's APUs. After Broadwell, not so much.

True, but their market share still was miniscule. They had nothing for laptops, nothing for servers that could hold up against Intel.
Most HPC and Server applications use Intel Xenons, most consumer PCs use Core i CPUs, most laptops use Core i CPUs. That is a huge problem, a lot of people are awaiting something new and good from AMD, and I dearly hope that they can deliver, but Intel has so much more R&D budget.
That's another reason why we can be so excited about the R9 Nano and Fury cards. If AMD gets back in the game with GPUs, they can hold up longer and hopefully get out of the red. Else, there's nothing in the way to stop Intels monopoly except law.

It could be possible buddy, that's why if I was a an author and a creator of an SFF project, I would definitely make patents for my own designs and creations....

Yeah good luck paying for the patent lawyer and the yearly fees when you don't even sell your product or it barely makes a profit. And paying the lawyers when you defend your patents in court.
 
I've thought about sandwiching a block between two components (even three) too over a year ago, I can't believe QinX and I were the only people that did.
The old 9800GX and the initial GTX295 both used a 'sandwiched' heatsink (between two GPUs rather than a GPU and a CPU), and both had aftermarket double-sided waterblocks. It's not a new idea by any means.

::EDIT:: Herp-a-derp. I even googled "9800GTX" before correcting myself, then made the exact same mistake again...
 
Last edited:
The old 9800GTX and the initial GTX295 both used a 'sandwiched' heatsink (between two GPUs rather than a GPU and a CPU), and both had aftermarket double-sided waterblocks. It's not a new idea by any means.

9800GX2.

Also the 7950GX2 had that design but I don't know if it had a full waterblock for sale.
 
Oh wow, I didn't realize using AIO water coolers for video cards was such an old concept: http://www.geeks3d.com/20090922/bfgs-geforce-gtx-295-h2oc-ready-to-use-water-cooled-graphics/

2009 old heh... Before AIOs, there was the Vapochill - a sort of refrigerator for CPUs. People used to attach them to GPUs too.

Edit: also found review of stylish Swiftech "all-in-one" from 2002 which has CPU and GPU cooling blocks http://www.hardocp.com/article/2002/12/07/swiftech_qpower_barebones/1

Swiftech said:
This line of products is designed to support extreme thermal loads, while operating at very low noise levels. It features pre-installed, plug-and-play liquid cooling systems using our H202 series liquid cooling kits. The system ships with the cooling fluid already filled, either in single or dual processor configuration (see MSRP for various configurations). Optional graphics card cooling is also available, and can be factory installed using the MCW40 graphics cooler.

103910599114ilIXHycc_2_15_l.jpg
 
Last edited:
Yeah I think some things are making a comeback, like in fashion, lol :D

It's funny how timing is so important. Somethings introduced at the wrong time were deemed failures (or freaks), then they come back to be big hits.

Tablets, large screen smartphones (Dell Streak), and AIO GPU cooling, to name a few :)

At least in AIO GPU cooling, it's still the same company reaping the benefits. In the others, not so much :D
 
It's funny how a project that wants to show what the future of computing looks like just lets everybody see how old its ideas are :D
 
The take away I get is simply "some of the best ideas take time to brew." Sandwiched coolers have been around forever in one machine or another, the idea of stacking memory chips, local to the component that'll use them springs back to at least the 90s when we started looking into die stacking as a means of sidestepping the inevitable end of Moore's scaling: Good ideas sometimes just have old roots. With all that said, I'm curious to see how things play out over the coming decade. I'm fairly sure the stacking we see now will extend to the dies their self, and in several years time we may see dozens of gpu dies linked in stacked networks. They could have on-site memory and on-site cpu-like cores to semi-automate their selves, freeing up the main CPU and alleviating the need for faster connections back to board. These dies would have far lower clocks, maybe even asynchronous clocks to keep heat down, while simultaneously being orders of magnitude faster than anything we have now.
 
Back
Top