Nothing special and of course there will be throttling, but that's how every reference design works.
Throttling will always occur (even with aftermarket cooling), but I would think that Vega would still manage to stay in boost clock range more often than not. Its' NVidia competitors are practically always above the stock boost clocks. The 1080 Ti and Titan Xp reference coolers are the best reference coolers ever made till now, and come with some sophisticated vapor chamber technology.
Speaking of which, does anyone know whether the Vega or Frontier Edition reference cooler uses vapor chamber or heatpipes?
Raja is a veteran who headed many a successful projects in both ATi and Apple. He was not a "cheap" choice.
"Bundles are the best chance to get Vega as a way to "poison the well" and deter miner hoarding (Vega + games, hardware, etc). The official word though is that AMD wants you to buy an entire ecosystem and really get into Freesync"
So... you're saying that for me to buy you're gpu I also need to buy a CPU/Mobo and Monitor....
Well guess my chances of getting this card are minimal at best. It's like AMD is trying their best every year to make sure I can't get their GPU.
"Bundles are the best chance to get Vega as a way to "poison the well" and deter miner hoarding (Vega + games, hardware, etc). The official word though is that AMD wants you to buy an entire ecosystem and really get into Freesync"
So... you're saying that for me to buy you're gpu I also need to buy a CPU/Mobo and Monitor....
Well guess my chances of getting this card are minimal at best. It's like AMD is trying their best every year to make sure I can't get their GPU.
Why? Fury Nano had the same number of CUs, ~15% lower clocks and 40% lower TDP compared to Fury XI find it hard to believe that a small decrease in stock clock and disabling 12.5% of the CUs leads to a ~29% decrease in board power from Vega 64 to Vega 56, for comparison, GTX1070 had a larger clock decrease and 25% of its CUs disabled, but only decreased board power by 16.7% compared to GTX1080..
I find it hard to believe that a small decrease in stock clock and disabling 12.5% of the CUs leads to a ~29% decrease in board power from Vega 64 to Vega 56, for comparison, GTX1070 had a larger clock decrease and 25% of its CUs disabled, but only decreased board power by 16.7% compared to GTX1080..
The problem is two fold - architecture and process. AMD have a vastly inferior architecture fir gaming and the GF 14LPP process cannot compete with TSMC 16FF+ . The most disturbing aspect is regressing over Fiji in perf/watt and perf/ sq mm and perf/flop with respect to gaming. Vega is an unmitigated disaster 10 years after the HD2900XT.They've upped the clock speeds of the 'basic' Vega's way past the point of sane efficiency to try and squeeze out a tiny bit more performance. For some reason they seem to be addicted to doing this sort of thing. Wish they'd stop doing it myself - efficient reference editions + let the AIB's go crazy is rather more sensible.
I bet you don't own a business then.
You seem to be missing the facts that:
1. Only one team can have the best idea/design.
2. You can't tell a team, "Go make the best design!" They either have the best invention or they don't, and you find out long after choices have been made for the primary design worked on.
3. Money plays a role in the development as you can hire the best design talent if you have more cash.
There are 32 teams in the NFL, the 31 who don't win the Super Bowl don't clean house on staffing. Companies make the moves they with the money they have.
Designing CPUs and GPUs is not stocking shelves.
If a mom and pop store has two stockers and one puts 20% more boxes on the shelf, mom or pop might say, "Sorry Willy, Jeff is doing 20% more work than you consistently, we're going to hire a new stocker.". There's not a lot of risk.
At Raja's level there aren't a lot of candidates and the ones that exist already work for your richer competitors.
I'm still in awe of how 30% more transistors plus 30-50% more clockspeed plus a process shrink from 28nm to 14nm, double the memory albeit with limited memory bandwidth AND supposed architectural improvements since FURY X has resulted in such a minuscule performance increase over Fury X. It takes some special kind of excrement sauce to achieve that so they should at least be proud of themselves in some perverse 'less is more' kind of way.
Once you get past the "sweet spot" of clocks for a given architecture (whether CPU or GPU), adding a bit more clock speed requires a lot more power.
AMD has been shipping most of their consumer GPUs in "factory overclocked" configuration for years. This makes efficiency look even worse than it really is. Dialing back a bit on clocks (and corresponding voltage) provides a large percentage of the performance with a dramatic reduction in power usage.
There are several tests of the Vega FE edition clock for clock with Fury X and the Vega FE is actually slower than the 3 years old architecture clock for clock. So they've basically regressed in terms of performance, rather than advanced.
I just don't get it how they could have spent 3+ years and developed an inferior gaming product.
At this point they should basically focus on the Polaris design and upscale that, optimize it where its possible, and I will tell them right now, add in more pixel performance, that is their issue.
Also their cards are not properly fed, they need to add in better scheduler and make the whole process simpler so its easier for developers.
So add in more pixel performance, improve the scheduler, use some of the unused vram to store data to feed back to the cores.
And I should care for that why? I'm a customer, I couldn't care less how much money or whatever it takes or doesn't take, I expect a great product, else I'm going to go with their competition.
They had over a year after the 1080, they are using a chip that is bigger die size wise than the 1080ti, has more transistors and yet it somehow 30% slower than the 1080ti!
I couldn't care less about their issues, fact of the matter is their GPU team failed and they should suffer the consequences and get fired! Simple as that.
Niiiice!Updated first post with all the latest AMD PR.
Also, anyone else notice that we're on page 69 ??
Then why do you care about chip size or transitors count? Shouldn't only perf/price matter? And it is the same in this case. Simple as that. Maybe if you care about perf/watt, in that case GTX 1080 is better choice, unless you downclock/undervolt Vega