GPU VEGA NANO
- Thread starter VisualStim
- Start date
Hopefully we'll be proven wrong but I'm not holding my breath: https://videocardz.com/70465/msi-damn-rx-vega-needs-a-lot-of-power
This is my hope for sure. 1070-1080 performance with decent enough power needs. Maybe even 400 USD or so!?
Ok perhaps hopes turning into dreams here! But C'mon AMD. Please?
I witnessed that forum post spiral out of control since I'm a regular on the forum where it's discussed. At the moment it's not clear what this actually entails since I've seen that person be very sceptical on AMD GPUs in the past.
From what I recall, Polaris energy consumption was not as impressive as Raja was "boasting" about over a year ago. When people tried o overclock it, the wattage skyrocketed which gave an indication about how the architecure scales. And as far as I recall, Polaris was not more efficient than their previous Fury Nano, which was strange. It seems to me that Vega with its bigger die is just an extrapolation of the consumption of Polaris on such a die size, i.e. way high. With this said, someone demonstrated that water cooling a Polaris card did wonder for the power consumption. Probably because being able to keep the temperatures down helped with wattage consumption.
Nvidia is still more energy efficient, perhaps because of their break through with tile based rendering, playing an important part. Either way, I welcome another Nano card with HBM memory, because Nvidia seems to have no interest in HBM yet. It's going to be the card for SFF enthusiasts.
Nvidia is still more energy efficient, perhaps because of their break through with tile based rendering, playing an important part. Either way, I welcome another Nano card with HBM memory, because Nvidia seems to have no interest in HBM yet. It's going to be the card for SFF enthusiasts.
Holy hell. Hopefully, if it exists, it'll only be around $499 or something for an 8GB card. Also, to the people talking about power consumption, aren't AMD including some technologies with Vega that Maxwell had which made it more efficient? Maybe the TDP of the cards is due to just how the architecture works, and that it wont actually draw that much wattage?
One month to go...
All I want is a RX Vega Nano, because, SFF...!
I would pay about 500 bucks for something with GTX 1080Ti performance; but I would want something in the 400+ bucks range if it were GTX 1080 performance...?
nano, Nano, NANO...!!!
All I want is a RX Vega Nano, because, SFF...!
I would pay about 500 bucks for something with GTX 1080Ti performance; but I would want something in the 400+ bucks range if it were GTX 1080 performance...?
nano, Nano, NANO...!!!
(starts at the interesting part)
With so much bare PCB it's safe to say short cards won't be a problem.
With so much bare PCB it's safe to say short cards won't be a problem.
gtx 1070 performance at 300 is what think the nano would be. just me tossing a coin, considering all the current information.
PCPer have some benchmarks of the Vega FE, as well as photos of the board.
Looks like there's enough 'wasted' space to just about cram into a slightly shorter PCB, but the VRMs required to power Vega's 270W-300W hunger are going to take up more room than the R9 Nanos (and we all know the power spike penalty the Nano paid for reduced phases). That high power consumption nets performance as expected, usually around the 1070, sometimes between the 1070 and 1080, so dropping power to move to a smaller PCB is likely to net performance not far off of the R9 Nano's.
Looks like there's enough 'wasted' space to just about cram into a slightly shorter PCB, but the VRMs required to power Vega's 270W-300W hunger are going to take up more room than the R9 Nanos (and we all know the power spike penalty the Nano paid for reduced phases). That high power consumption nets performance as expected, usually around the 1070, sometimes between the 1070 and 1080, so dropping power to move to a smaller PCB is likely to net performance not far off of the R9 Nano's.
so why would you go amd if you get more performance for less power from a mini gtx card? the price would have to be very compelling
Right now? The only compelling reason I can think of is water blocks, the mini GTX cards are usually vendor specific which means no custom water blocks.
I wouldn't draw too many conclusions on performance just yet. I've yet to see a workstation card's performance come close to a gaming card.
True, but the Vega FE is closes to a Titan than a Quadro. It lacks application certification for one thing, and the drivers have an explicit 'gaming mode' (though AMD themselves admit that all that does is change the UI). Vega FE doesn't have big FP16 (that's taken up by FP32 packed math) or FP64 units that can be fused off to claw back power, nor does it have ECC hobbling memory performance. And unlike FirePro/Radeon Pro cards, the Vega FE does not appear to be running at lower clocks than are possible: from the cursory OC testing done so far with the air-cooled Vega FE at 100% fan there is just not all that much headroom. The caveat here is that with Fury and Fury X the watercooler on the X allowed for lower core temperatures, which allowed for lower driving voltages per-clock which raised clock headroom. RX Vega with watercooling may exceed air-cooled Vega FE clocks, but I wouldn't expect the same from an air-cooled card.
