Associate
+1 to that. I'm using them right now. Currently not a patch (excuse the pun) on Radeon drivers at the moment.
-
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.
- Status
The PG278Q I got was perfect, absolutely stellar and I was so happy. Sold it to a friend last year, and a week after he had it a dead pixel appeared right in the centre. It's a little grey one, so he wasn't too miffed at least.
I really wish Dell would embrace the "gaming" monitors more. Take their fantastic build quality, and put out a 100Hz 34" Ultrawide at 3440x1440 and I'd buy it right away.
The only reason I'm considering the ROG UW is because it seems to be the best of the UW G-Sync ones. The Acer can "overclock" to 100Hz; but a friend of a friend's cannot manage it at all. Acer told him it's not guaranteed either. I know ASUS RMA system is an horrific mess, but damn Acer also takes the cake.
I see the difference between the LG and Dell 38" UWs, is that LG manages to get to 75Hz and has Freesync, as opposed to normal 60hz on Dell. That and my previous LG being great is the reason I have that in my basket as opposed to the Dell there.
Hopefully top Vega doesn't disappoint, since it means I can old onto my current MG278Q a little longer and wait for 100+Hz UWs to become more available. Otherwise I'll just hop to green again.
Associate
- Joined
- 29 May 2014
- Posts
- 390
- delayed one year
- inferior performance compared to high-end GPU's from the competition
- inferior watt/performance
- inferior price/performance
AMD fans are still going to buy it, because a driver will be released someday in the future and probably in another universe that will give 100% more performance.
- inferior performance compared to high-end GPU's from the competition
- inferior watt/performance
- inferior price/performance
AMD fans are still going to buy it, because a driver will be released someday in the future and probably in another universe that will give 100% more performance.
Soldato
After Sunday you'll be able to buy Vega, pop a chicken in your PC and combine gaming and roasting
Winner Winner Chicken Dinner
Its just getting better this thread ..i don't care for vega when we have banter/jokes like this
it's either that, or cry at the disappointment
+1 all of the Pro ones I've used (and i'm using one right now) have been very good.
Permabanned
Absolutely not, There's no way £999 would be a fair price for a Vega card even with a 5 year warranty, watercooled or not.
AMD's problem is hawaii was the last AMD GPU that had proper 1:2 GP64 support, so for HPC and compute applications it is the most recent GPU they can offer. Fiji had ridiculous amounts of Fp32 compute, but had a crippled VRAM buffer for HPC applications and professional software, 4GB just isn't enough for many application but if you could get away with 4GB and FP32 then Fiji was extremely powerful. Then Polaris came along that is more efficient than Hawaii and a newer process, could adress 32GB of ram so for some professional software this is more useful. With Vega10 again we have no decent FP64 support, and although it can have 16GB GBM2 it doesn't offer anything special bandwidth wise and appears to have its own efficiency issues, however the FP16 double rate is useful for deeap learning so there are some merits to it. Vega20 is supposed to be AMD's first return to 1:2 double precision support, so that will really be the first proper HPC GPU from AMD since Hawaii. So the HPC/compute/professional side of AMD GPUs are just a complete mess.
Then at the gaming side AMD always seem to have very staggered updates. Nvidia prefers a top-to-bottom refresh. AMD have been doing lots of re-brands, 290->390, 480->580, instead of making a smaller Vega.
It is a mess but probably simply explaine by lack of R&D budget meaning they have to reuse old technology for the low end and can't afford to refresh an entire range of GPUS in one go.
Unless it smashes the Titan Xp in games as well as it does Pro workloads.
Associate
- Joined
- 19 Jun 2017
- Posts
- 1,029
Soldato
With well over 90% of gamers (as per the Steam hardware survey July 2017) using a 1060 or lower card, AMD are just targeting the majority of gamers by releasing the non enthusiast cards before the higher end cards.
Nvidia target the ultra high end users which you can measure in the single digit %'s (again as per the Steam Hardware survey July 2017, Titan, 1080Ti, 1080 is less than 4%) People on here seem to think every gamer is a high end enthusiast when in fact we're just a very small minority in the overall scheme of things.
Last edited:
Soldato
£650 for 1080 matching variant according to this.
Associate
Associate
If that is indeed true then Vega is dead in the water and Raja and RTG needs shooting.
However, when politicians have some bad news to tell the country they usually leak out something far worse and then when the real news comes out which isnt quite as bad...everyone feels a bit better. This may be what is happening right now. RX Vega is only matching 1080 so Crazy prices being leaked. Then on launch day those prices are lower than the leaks and everyone feels that is isnt so bad after all. Classic Ploy that may or may not be being used here.
Soldato
- Joined
- 28 May 2007
- Posts
- 10,144
The top 3 may be right but until AMD release the RRP then you can't go with inferior price/performance.
Associate
- Joined
- 19 Jun 2017
- Posts
- 1,029
yes waiting for radeon software saruman edition
on an unrelated note... amd expects its employees to work on sundays.. that reeks of an investment bank not a technology company.. they should have launched those cards on friday
Thank you for this writeup. Does explain why AMD must produce many different GPUs, doesn't reduce the fact that their planning is way flawed.
Could you remind what is missing in Vega whats prersent in Hawaii that allows to do 1:2 64 bit float calculation?
Nvidia releases higher end GPUs first because there is less volume and hgiher profit margins allows for lower yields. Once all the kinks are worked out then it is easier and cheaper to work on mass volume low-end parts like the 1050ti etc.
Double precision FP64 support requires dedicated hardware, this costs transistors which will increase die space and consume power. For gaming there is minimal use for Fp64 computation and thus gaming parts can have this stripped out. Nvidia reduced FP64 support in Maxwell just like AMD did with Fiji because they were stuck on the same old 28nm process and needed the extra transistor budget. With Pascal Nvidia moved to having a dedicated computer card with all the FP64 cores (GP100) while maintained a more optimal configuration for gaming. AMD haven't yet released a proper contender for HPC use, and even VEega 10 wont support fast FP64. At some point next year there shouldbe a Vega20 with 1:2 double precision.
That site is using a placeholder price as found on a Swedish hardware site. It is not the actual price according to AMD.
That was interesting to re-watch knowing what we know now.
However I came away with the following observations:
- AMD had pulled money and focus away from discrete graphics, due to bad business decisions about the future, and Raja had to convince them to re-focus when he came back on
- Vega is the culmination of him getting money and engineers back, and laying a foundation of important 'fundamentals' in the hardware/software
- This implies Vega is a bit of a swiss-army-knife card, and a proof of concept of approaching many different workloads
- So it's a bit of a jack of all trades, master of none, and has a lot of transistors used for things which won't benefit certain workloads (like gaming)
- Its main focus appears to be deep learning and large-dataset workloads (like editing 8K video and real-time raytracing), since these are important new growth areas
- So gaming, CAD, and other traditional professional work are secondary in the design
- Not necessarily on purpose per se, but they had a transistor budget and Vega is all about 1. Showing they can tackle any/every workload type in some form 2. Plugging gaps in capability they didn't have before, and also adding things even Nvidia can't do, like the SSG technology. Therefore 'new' stuff got the main focus.
- When discussing the "multi-generation" part, he was just referring to the community's "FineWine" term about how they improve their products a lot over time through drivers. And also we know there's a 1/2 rate FP64 HPC Vega card coming next year sometime too. So this didn't say anything about Navi either way, it could be an iteration of Vega or something totally new
The conclusion I come to is that Vega is more like a blueprint for the future. It completes a puzzle (and adds a couple of extra pieces) so that they can move forward from a strong start point. But it in of itself is not necessarily a great GPU for specific use-cases (other than the 2 mentioned they seem to have focused on a lot).
So I feel like it may turn out disappointing for gamers, but part of the plan is that now they've sorted their gaps they can hit the ground running for the 7nm generation and make big gains. They no longer need to add anything new and/or plug holes, just focus on improvements.
EDIT: Also bear in mind the Titan Xp/1080 Ti die is stripped out of all its compute, and doesn't have 2xFP16 etc. so continuing from all my above observations, it basically means Vega is a smaller die than the 1080 Ti. In that it's not 484mm2 of gaming-focused die, whereas the 1080 Ti is. So now I'm not surprised at its performance (though still disappointed at its perf/w).
- Status