Read more.Captain Jack's performance is revealed by ChipHell.
Read more.Captain Jack's performance is revealed by ChipHell.
But if this is on the new 20nm process, then it's worrying that the power consumption is comparable to the GTX980... When Nvidia's 20nm cards arrive, they could be much more efficient.
Depends on their size. I wouldn't expect the equivalent chip (large die) to be that much more efficient than this. The mid-range smaller die chips, yes, but then the smaller die ones from AMD will also be much more efficient too.
Obviously pinch of salt time, but either way AMD have done well here - if this is a 20nm chip then they've managed to beat nVidia to it - and impressively they've done so with an enthusiast card rather than a safe low-end one. On the other hand, if this isn't 20nm, then they've done impressively well to get such a large/performant chip in under such low power usage.
From reading elsewhere I thought the original leak identified this mysterious 'Captain Jack' card as the 380X not the 390X? Has this changed?
I read somewhere that Nvidia attributed a lot of the improvements they've made with their Maxwell power efficiency to their experience gained with their development of ARM cores like their Denver architecture. Although since AMD are also an ARM partner, presumably they should have similar resources?
They are but there is multiple levels of partnership with ARM I believe. Though that would explain the above power improvements.
However, I'll wait till I see a retail product on a test bench before I believe anything. The rumours are always the same. Company X will leapfrog Company Y and do it cheaper/cooler/quieter... let's just see.
(I want to believe though!)
Steam - ReapedYou - Feel free to add me!!
If that's true that the CJ card is the lesser 380X rather than the top-of-the-line 390X then those values are pretty impressive.
That said, like the Hexus writer, I'm pretty dubious that AMD would be able to pull out something this impressive that quickly, so I'm going to chalk these figures up as "spoof".
Of course, if AMD have managed to deliver a "midrange" card with 980-beating performance and get it to draw less power than the equivalent current-gen card, then I'll definitely raise a beer to them. And start looking to put in an order when they start shipping...
Pretty cool to get a little idea of what we might expect here! But I'll be waiting for Tom's Hardware benches before I make my mind up with these
A 4096, 1GHz + core is not going to be light on the electricity unless they've spun a lot of magic into it. 200W would be doing very good for such a big core.
The average benchmark score is a bit of blind ally. It only takes one of those games where it's super fast to skew the average by a significant margin and that could come from incorrect rendering / bug
i seriously hope they dont charge too much for this card because the specs look abit cool for wehat it is lol i actually might sell my gtx 780's for it if the reviews come across well
Doing a total upgrade in March next year so hope this will a big improvement on there last card
JABULANI NONKE
I agree, if this is a 28nm card then they've done really well.
If it is a 20nm chip, then I think AMD will be in trouble compared to NVidia.
The things is, moving from 28nm to 20nm should give approximately double the number of transistors within the same die size. (Very roughly this should give double performance for the same power consumption). Looking at the charts, the rumoured 390X is around 30% faster than the 290X, whilst consuming ~30% less power). This is more or less in-line with what you'd expect from a die-shrink of the 290X with around 30% more cores.
If NVidia do the same with the GTX980 (die shrink with 30% more cores), then you'd expect performance comparable if not higher than the 390X, and power consumption less than 150W...
Point to note!. Previous articles on Hexus, Toms Hardware and WCCFTech stated that the 380X would be competing against the GTX980 and that the 390X would be going up against the upcoming (not yet released!) Titan2 and 980Ti. Looking at the above benchmark table (which btw is very poorly worded and doesn't make much sense), the 'captain jack' test/preview sample scored 65.6 whereas the GTX980 scored 56.6, so I am inclined to believe this is a test sample for the 380X and not the 390X. Also, about 6 weeks ago there was an article on WCCFTECH titled 'GTX980 may lose performance crown to 380X'. Furthermore, Hexus et al reported that AMD confirmed it is on schedule for release of 380X in February 2015.
Based on previously disclosed information, I do think this leaked result is for the 380X and not 390X (also, the Titan2 and 980Ti are not even on this table).
If it really is for the 390X, then that is worrying because what are AMD going to do when the 980Ti and Titan2 are released?.
So imo, this leaked result is for the 380X and not 390X.
See -
http://wccftech.com/gtx-980-lose-performance-crown-amds-r9-380x-febuary-2015/
Even if it were 20NM,AMD is months ahead of Nvidia,so ultimate trying to say they are in trouble is coming across as excessive negativity. You should make the argument Nvidia is in trouble to the same level if they need to use 28NM parts to compete with 20NM parts for like six months at least.
That is if any of these rumours are true.
This is what slightly annoys me - AMD is frequently 3 to 9 months ahead of Nvidia,especially for midrange parts,on process node shrinks and no one ever says Nvidia is in trouble during those time periods ever.
There are currently 1 users browsing this thread. (0 members and 1 guests)