Soldato
- Joined
- 6 Jun 2009
- Posts
- 5,474
- Location
- No Mans Land
In 6-12 months it will have Volta to contend with and will be severely spanked making Vega look even more poo than it currently is.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Poll: ** The AMD VEGA Thread **
- Thread starter TNA
- Start date
- Status
In 6-12 months it will have Volta to contend with and will be severely spanked making Vega look even more poo than it currently is.
No, I understand. The FE seems be to pulling a lot of power and I don't think the RX will be much different.
If I was buying a PSU for one of the high-end cards, Titan, Ti, FE or RX I'd be looking at an 850W or more.
Last edited:
Pretty sure you don't need anywhere near an 850w for the nvidia cards.
Will also future proof you a bit if you decide to sli/crossfire or add loads more bits
Caporegime
- Joined
- 8 Jul 2003
- Posts
- 30,063
- Location
- In a house
Just read through the last 10 pages of this thread, funny as hell 
Associate
- Joined
- 10 Apr 2015
- Posts
- 1,033
- Location
- airdrie, scotland
yeah i was advised a 550w would be fone for either an rx580 or 1060/70
Depending on the CPU and overclocking headroom. A 550W seems low for a high-end machine. Why skimp?
Doubtful, I'm pretty sure 1600 MHz was the target and they've hit that. Increasing clocks more would make the already high power consumption even worse.
Associate
- Joined
- 10 Apr 2015
- Posts
- 1,033
- Location
- airdrie, scotland
i went for a ryzen 1600 and either rx 580 or 1060 and was adviced that a good quality 550w psu was more than sufficient
Also the heat. Wasn't the blower card's fan revving at 4000 rpm while the gpu was at 80+ degrees during one of the benchmarks?
An draw 500w, I don't think so.
If RX Vega is really in some way faster than FE then it can only be because it is a new revision and I doubt clocks will go above 1600Mhz.
Fine for those cards but the RX is in a totally different category for power consumption.
I take it you're overclocking your Ryzen 1600?
Supposedly. I dunno why anyone who cared about heat output, throttling, and noise would buy a "founders edition" or "frontier edition" card anyway, they are always terrible in that respect. It makes me laugh that nVidia had the cheek to portray them as the premier version of their products.
Associate
- Joined
- 10 Apr 2015
- Posts
- 1,033
- Location
- airdrie, scotland
yes i will try for a 3.8 or 3.9.
They're normally a little better than AMD blowers, only a little
I wonder if there will actually be custom versions of the FE? I don't suppose there's a massive market out there for manufacturers to bother.
You have to factor in the extra power draw.
Associate
- Joined
- 10 Apr 2015
- Posts
- 1,033
- Location
- airdrie, scotland
yeah, as i mentioned im not too clever when it comes to all this sort of stuff and opened a spec thread on the forum and a 550w psu was what was recommended. still not all lost worst case scenario i can change it over with my sons psu
A hypothesis putting bits and pieces together.
There is a complication with the HBCC hardware that rockets power draw when trying to extract full performance or otherwise interferes with gaming workloads. In such a case I wonder whether RX Vega will disable a lot of functionality of the HBCC reducing it to a dumb controller that only interfaces with the HBM cache. Perhaps retaining some mild latency and power penalty over a simpler dedicated controller but significantly less than when fully functional. For workstation Vega and this somewhere-in-the-middle FE Vega, this inefficient feature is left fully enabled both gutting and capping performance (hence the even lower clocked workstation cards to drop into a manageable power envelope and the easy promise that RX Vega will have higher gaming performance than FE Vega). The only reasons the FE edition exists is to meet the H1 launch window, satisfy investor expectations and avoid a potential avenue of litigation. Which would explain it's conveniently niche use-cases, somewhat artificial target market and confusing marketing. If there is an issue with the silicon it would also make some sense to not provide a recent or most optimized gaming driver for FE Vega, but you could equally argue it the other way.
If it is the case it isn't necessarily a huge concern for subsequent GPU archs. On the fabrication process side alone '7nm' will help address the issue by reducing the cost of hbcc implementation (smaller proportion of the die's power and transistor budget) regardless of whether or not they are able to make an improvements through design.
It's already been said but their mistake (as usual) is not properly setting up some initial guiding boundaries to the early outflow of information. A few key points need to be well communicated so that everything that then unfolds automatically takes the preceding info into account. I think their AMA was an attempt but failed to do it adequately or clearly.
Thoughts?
Edit:
From Raja's comments in the AMA:
RX Vega being faster (and cheaper) and the one gamers should go for with reduced HBM cache at 8GB. 16GB not planned as states they will consider a 16GB RX Vega. Software/driver team finding Vega arch a challenge (as they would apparently attest to).
So some inferences:
16GB a differentiator to help justify FE existance and price. Smaller stacks or fewer active dies save power for the gaming card, 16GB not targeted by AMD for gaming. Software comment points to some difficulties on the software side with Vega. The most drastic change from previous GPU designs, from our limited information, appears to be the memory arch and high bandwith cache controller. A more sophisticated controller is going to draw more power overall and probably more for the equivalent tasks, likely introduce latency and definitely provide greater opportunity for hardware bugs with more complex data handling scenarios over simpler designs.
There is a complication with the HBCC hardware that rockets power draw when trying to extract full performance or otherwise interferes with gaming workloads. In such a case I wonder whether RX Vega will disable a lot of functionality of the HBCC reducing it to a dumb controller that only interfaces with the HBM cache. Perhaps retaining some mild latency and power penalty over a simpler dedicated controller but significantly less than when fully functional. For workstation Vega and this somewhere-in-the-middle FE Vega, this inefficient feature is left fully enabled both gutting and capping performance (hence the even lower clocked workstation cards to drop into a manageable power envelope and the easy promise that RX Vega will have higher gaming performance than FE Vega). The only reasons the FE edition exists is to meet the H1 launch window, satisfy investor expectations and avoid a potential avenue of litigation. Which would explain it's conveniently niche use-cases, somewhat artificial target market and confusing marketing. If there is an issue with the silicon it would also make some sense to not provide a recent or most optimized gaming driver for FE Vega, but you could equally argue it the other way.
If it is the case it isn't necessarily a huge concern for subsequent GPU archs. On the fabrication process side alone '7nm' will help address the issue by reducing the cost of hbcc implementation (smaller proportion of the die's power and transistor budget) regardless of whether or not they are able to make an improvements through design.
It's already been said but their mistake (as usual) is not properly setting up some initial guiding boundaries to the early outflow of information. A few key points need to be well communicated so that everything that then unfolds automatically takes the preceding info into account. I think their AMA was an attempt but failed to do it adequately or clearly.
Thoughts?
Edit:
From Raja's comments in the AMA:
RX Vega being faster (and cheaper) and the one gamers should go for with reduced HBM cache at 8GB. 16GB not planned as states they will consider a 16GB RX Vega. Software/driver team finding Vega arch a challenge (as they would apparently attest to).
So some inferences:
16GB a differentiator to help justify FE existance and price. Smaller stacks or fewer active dies save power for the gaming card, 16GB not targeted by AMD for gaming. Software comment points to some difficulties on the software side with Vega. The most drastic change from previous GPU designs, from our limited information, appears to be the memory arch and high bandwith cache controller. A more sophisticated controller is going to draw more power overall and probably more for the equivalent tasks, likely introduce latency and definitely provide greater opportunity for hardware bugs with more complex data handling scenarios over simpler designs.
Last edited:
- Status