-
Save 15% on ALL SFF Network merch, until Dec 31st! Use code SFF2024 at checkout. Click here!
GPU VEGA NANO
- Thread starter VisualStim
- Start date
While it isn't necessarily cancelled, the outlook for it does seem rather bleak. Thermal concerns aside, recall the original price of the R9 Nano - the card was not received well outside of the SFF community and pricing was the major concern. HBM wasn't cheap then, and HBM2 isn't now.
Vega's pricing is a reflection of the manufacturing costs and difficulty associated with HBM2 - an RX Nano would likely cost about the same as a Vega 64 while offering less performance, offering a repeat scenario of the R9 Nano.
Vega's pricing is a reflection of the manufacturing costs and difficulty associated with HBM2 - an RX Nano would likely cost about the same as a Vega 64 while offering less performance, offering a repeat scenario of the R9 Nano.
Same scenario, but higher temperatures.
With a SFF dedicated GPU they can't admit such high temperatures.
It was basically a mock-up of what it might look like though
And sure, it might've been possible to get Vega under or including 175W, but yeah that would've been a little too hard.While it may very well be cancelled, I do want to note a couple of other points:
1) Vega is clearly supply constrained. The fact that there are no open air cards from partners points to this- they're only getting the supply now to do that, and Nano would be a lower priority.
2) There have been plenty of people successfully undervolting Vega cards and losing only minimal, if any, performance. I don't know what the TDPs are after undervolting, but you could potentially get something more reasonable if you binned chips for ones that undervolt well.
1) Vega is clearly supply constrained. The fact that there are no open air cards from partners points to this- they're only getting the supply now to do that, and Nano would be a lower priority.
2) There have been plenty of people successfully undervolting Vega cards and losing only minimal, if any, performance. I don't know what the TDPs are after undervolting, but you could potentially get something more reasonable if you binned chips for ones that undervolt well.
They could easily get a Vega chip under 175W, or down to almost any arbitrary power envelope. From testing, undervolting and downclocking Vega seems to work quite well to improve efficiency.It was basically a mock-up of what it might look like though
The problem is it wouldn't reduce the cost of the chip at all, but it would still reduce the performance. Vega 64 at RRP is not a good value proposition for anyone, and Vega 56 is only just competitive depending on market. A Vega Nano at the same price as even the Vega 54, but with even lower performance is not going to sell when there are competitive GPUs in the same size class but cheaper and more performant.
Hmm, well alright. How about a further cut down chip? Imo that'd be worth the trade off between heat and power consumption compared to performance. Something like Vega 40, 44 or 52 could work quite well, and it would allow for AMD/RTG to use more of the chips that they'd normally just throw away due to defective shader clusters and whatnot.
Cutting down the chip saves you no money unless you have a surplus of chips not meeting Vega 56's bin level. I'd imagine the Vega 56 bin level was set to harvest the majority of dies that could not be used for Vega 64 (and the equivalent Radeon Pro cards). Failures for HBM assemblies (GPU + HBM dies + interposer) seem predominantly to be failures during final assembly when the chips are bonded to the interposer, so unless you were willing to produce a SKU with a single functioning HBM2 die (4GB, half bandwidth) there's not much opportunity for cost saving.
Invented (not false) specs based on news, like this https://www.techpowerup.com/gpudb/2997/radeon-rx-vega-64-nano and this https://videocardz.com/amd/radeon-500/radeon-rx-vega-nano
Not directly related, but this vega aib card from xfx is using an itx form factor pcb. Interesting aesthetic with the power connectors right in between the fans.
now they just need to slap an aio on there, fury x style and i'm good.Not directly related, but this vega aib card from xfx is using an itx form factor pcb. Interesting aesthetic with the power connectors right in between the fans.
Not directly related, but this vega aib card from xfx is using an itx form factor pcb. Interesting aesthetic with the power connectors right in between the fans.
Fairly obvious that the PCB stops right after the PCI connector; see line of screws to the left of massive notch on backplate in first picture & yellow line that shows edge of PCB in second picture...
If anything, it is slightly shorter than the 170mm that is an ITX motherboard...
yeah i was thinking about that cooler. you got a point there ._.
wonder how hard would it really be to air cool within the itx card outline
wonder how hard would it really be to air cool within the itx card outline
I don't see why as you can put directly an aircooled vega 56 in it..
watercooling is a great addition for Vega 64 highly OC, but not really on Vega 56 as you are limited by bios.
I don't see why as you can put directly an aircooled vega 56 in it..
watercooling is a great addition for Vega 64 highly OC, but not really on Vega 56 as you are limited by bios.
Yes. I have a 56 and when you leave the power limit set to 0 in After burner it will peak at 165w card only per afterburner. Clocks will hover in the 1233-1340 range based on temps and power limits. --With my card anyway. I know that tool is not exact but it's something to reference. Once I OC the memory to 950 (v64 speed) and under volt the core by -75 AND increase the power limit to +50 I get 1450-1550 core and 250w shown in afterburner. Since the 64 bios /card only gives more voltage /power headroom It gets worse for sff.
V56 is the card to "Nano" for sure.
Taming this beast I think is doable with a Nano type cooler but it wil l dump serious heat into your chassis. Just like the OG Nano.
O yes first R9 Nano could be highly overclocked at expense of more heat...and noise...
However, i still scared about Vega Nano...I don't see the point as efficiency per watt has not increased from fury to vega on AMD side (fury was quite ok on this).
I hope AMD will release a vega refresh in 2018 (time to wait for Navi in 7mm) with better efficiency..
However, i still scared about Vega Nano...I don't see the point as efficiency per watt has not increased from fury to vega on AMD side (fury was quite ok on this).
I hope AMD will release a vega refresh in 2018 (time to wait for Navi in 7mm) with better efficiency..
