Ok so HD 7000 for the next months as the only available option.

AMD CGN2 needs to improve Boost feature, Lower Power Consumption,
Drivers latency issues, Heat and Noise combination, Increase number of
Stream Processors +2048, but keep 384-bit bus at High-End segment, also
increase ROPs +32, if there anything more I will add.

The consoles combining AMD GPU and GPU, also can contribute to keep AMD busy until the end of the year.

I found odd AMD Graphics Roadmap, they really need to check current Nvidia cards prices…

Now 3Dmark is a fair comparison between Nvidia and AMD cards? In 3Dmark
11 some people claimed that 3Dmark was irrelevant for measure real
world performance.

It is understood that there are benchmark software and actually gaming are two different things entirely. The reason why Tri & Quad SLI and AMD equivalent exsists is for benchmarking!! Now GNC 2.0 will obviously be more powerful the GNC 1.0 that is obvious. I do think that power requirements will go up a bit it certainly wont be going down not until the next fabrication shrink. Heat and Noise issues will be delt with how it has always been delt with 3rd party Asus, MSI etc

Nvidia has priced their parts more competitively now, however AMD still hold the price/performance at this time and now with the gaming bundles their positioning is excellent. The console business will not interfere with AMD in anyway, their designs are ready so its all up to TSMC or whichever fab to start the ramp up.

Fire Strike eats GPU’s for breakfast!

They don’t just slap a PC part onto a console, it has to be custom designed specifically for the console and run through rigorous testing and many revisions until it is perfect and ready to release. Could you imagine a recall of all PS4 a few days after launch after finding a fault with say the chosen VRM during specific usage? It would cost Sony hundreds of millions if not billions and this is why they put so much time and energy into the hardware. The software can always be updated, but if you ship hardware with even a minor fault, you are screwed as a console designer. AMD didn’t just hand Sony a load of Tahiti chips and walk away.. Now take this and multiply it by 2 for both MS and Sony. You are crazy if you think consoles played zero role in AMD’s workload and in turn the release of their next gen GCN 2.0

3DMark11 was a completely worthless benchmark. Just compare GTX580 vs. GTX680 or GTX680 vs. HD7970GE scores and check out how those games perform on average. When 3DMark11 came out, people didn’t just dismiss it. It took time to review performance in many DX11 games to start to see the picture that 3DMark11′s scores for 580/680/7970GE never added up to what gamers saw in the real world. Now I am not suggesting in any way that the new 3dMark should replace real world gaming benchmarks but at least the scores in it much more in line with the average performance in games between generations.

Regarding prices, it looks like they are using MSRP prices, not market prices. However, if you looked at market prices in US/Canada, things actually look worse for NV, not better. AMD currently has superior price/performance and performance at every price level above $150. The only card that’s hands down superior to purchase is the GTX690 over HD7990.

Possibilities of delays could also include:

1) Delays of 20nm node to end of 2014 (say Q4 2014). Then there is not much point in launching HD8000 series soon or HD8000 series will last nearly 2 years. Might as well work on improving HD8000 series or waiting until 28nm node matures so that they could get a bit more performance. The opposite could be AMD realized to make HD8000 series worthwhile, they need 20nm node. Perhaps 20nm node has been pushed forward to Q1 2014.

2) AMD is keeping busy rewriting the entire memory management subsystem for HD7000 cards. That could mean they are busy testing various games and getting this finished as soon as possible. This takes man-hours and financial resources. The other thing is they might want to wait to launch HD8000 series after the memory management is re-written and they release the rumored driver that fixed latency issues (because it’s inter-related to the memory management based on the comments by Dave Baumman).

3) AMD is pulling a Sony (Sony said they are not launching PS4 or announcing before MS does, but Feb 20th there is an event scheduled for a possible unveiling of PS4….?). Could AMD just be bluffing to see where Titan drops and then they can respond in 2-3 months and adjust prices of HD8000 series accordingly?

4) AMD has ran into serious issues trying to get much more performance out of HD8000 parts due to already near 250W power consumption of HD7970 GE reference card. How do you fit 2560 SPs, 48 ROPs, etc. Either HD8000 series would be a very minor refresh like HD6970 was or they’d need way more time to redesign the chip from scratch (Perhaps start with Pitcairn XT instead and make a larger version of that).

Here’s also a response from AMD rep at OCN’s thread about this ‘delay’:

“Guys, AMD never confirmed anything. I wrote the tweet, we never spoke to WCCFTech, and we’ve never so much as hinted that a next-gen product would be coming this year. That part where WCCFTech claimed we’ve released reports on a next-gen was completely fabricated (¶2, first sentence).

We even announced at CES that “Sea Islands” was our mobile products, and reinforced that today’s assortment of HD 7000 products at your local etailers are the products you can expect for months to come.

When I said that the HD 7000 Series is our focus for the foreseeable future, I meant it.

You can’t delay a product that hasn’t been scheduled, planned or announced.”

Just thank me for the tip, and also put the print screen :)

http://i.imgur.com/91rIU9c.png

Source: http://itoo.ws/OCNThracksquote

“When I said that the HD 7000 Series is our focus for the foreseeable future, I meant it.”

I wonder if the drivers that fixes the memory management issues of GCN not only solves stuttering in some DX10/11 games but also brings some “hidden” 10-15% performance boost like Catalyst 12.11s did. That would be nice.

1. Last I heard, risk was H213 (probably meaning early q4 given tsmc’s past) for risk, with mass production scheduled one year before 14nm (planned for February 2015). Given that they are ready for 20nm designs (9/12), and that’s usually a year-long process of fabrication tweaks and validation before production…that gives you a rough time frame on 20nm. I don’t think the time-line has changed much in a long while. That said, space will probably be extremely tight for a while given Apple’s shift to TSMC, nvidia’s Maxwell/Denever, etc.

2. It’s possible they would want to put their best foot forward, sure. I also wonder about the availability/price/performance of 7ghz ram, which 8900 and 8800 would both need (if 1792/2560). Anyone notice 4Gb (256MB chips) from both samsung/hynix, 8Gbps from Samsung, and 1.5v 7gbps from Hynix (1.6v is available now, 1.5v now from Samsung) were announced/delayed to Q313? Hmm…

3. I’m going to put my foot in my mouth. I called bs on this delay for a looong time because of such reasoning (as well as a lot of other factors leading to Q2 making the most sense), but now they’ve all but confirmed it. You can only hope you can take a company at their word, but also realize marketing has a job to do. After the trouble they’ve gone through now though, people are going to be pissed if they are being deceitful.

4. I look at it like this:

High-end needs to be within 300w. A shader core is ~ between 6-7mm2 on 28nm. That puts a 2560 part somewhere around R600 size (420mm2 or so), an increase of 15% or so over Tahiti. Logic scales power better than voltage/clocks. 7970 GE is 1.162v, volted to clock ~1150mhz-1175 at 250w. In theory, a refresh design would be aimed at a similar voltage for best scaling within 300w. As I’ve said before, 7ghz would mate well with 2560 at 1166mhz…a good spot to aim over-all (ie stock at ~1050/6400 to leave 10% for AIBs). Do I think that could be done within 300w? Conceivably, yes, especially with dynamic clocking tech.

Same goes down the scale. 2304 at ~1.05v within 250w seems reasonable (one would think an aim of 925/5000-1111/6000). 1792 obviously could be a host of things up to 200-225w. I imagine another concern is 1536sp at less than 150w, but think in terms of 925/5000 (1.05v and 1.35v 5ghz gddr5), not to mention 7870 draws around 115w stock and less than 150w overclocked (stock voltage of 1.218). This would be 20% more shaders, but probably 15% or so more logic using 16% less voltage aiming for a lower clockspeed (again maybe up to 1111/6000 or less). Seems very doable.

Like I said, I believe them when they say no new products, explicitly because they already have items positioned where they want versus nvidia (as detailed in the slide).

Tahiti LE is same performance as 660ti in a similar power envelope (<225w)…7970 faster than 680…would 8800 be the same? Probably not, just use less power than the former (essentially killing 7870) and be similar to the later for cheaper. What sucks is that there seems to be efficient designs laying on the table straight up waiting for them, and consumers want them. In pretty much all cases though, they'd probably make less or a similar amount off each product. (ie 8850 vs 7870 vs 660ti, or 8870 vs 7970/7950 vs gk104, perhaps 8770 vs 7850 vs 650ti).

Because 3Dmark score is what really matters.

384 bit vs 256 bit that’s the only difference still gtx 680 beats 7970 ghz edition in many tiles and is 2-5% slower in games that supports amd….let GTX TITAN arrive will see wat amd has to say 250 watt 7970ghz edition{flop} vs 235 watt GTX TITAN….than will see…..previously GTX 580 384 bit was 30% faster than hd 6970 256 bit …..hats off to kepler ……

All he can is just to deny all the leaks and say that sites like wccf is out there only for pageviews. The truth is always in the middle :)

Hidden performance boost, with frame delay.. TechReport showed this, AMD fixed it in later drivers and the performance boost was almost gone again. Im using a 7970 GHz btw.

12.11 was a big joke tbh. Loved by benchers, hated by gamers. I want stutterfree gameplay, not highest possible FPS..

Hi,

The main issue with our article was that we were treating rumors as official statements by AMD. We were approached by AMD PR representatives who asked us to edit the article hence what you are looking at is an updated version.

Their point was that how could there possibly be a delay when there was no HD 8000 product to begin with? As you know that all details we have about the HD 8000 GPUs are based on reports/rumors/speculation from various tech blogs and forums.

There is however a bit of confusion caused around the codenames themselves, Sea Islands is basically an internal codename for the HD 7000 series and Solar System is the internal codename for HD 8000 series, that is also mentioned on PCWorld’s article.

Furthermore, the rep confirmed that Sweclockers was wrong on the HD 7990 bit, there’s no dual GPU reference card planned by AMD themselves. They are relying upon AIB built HD 7990 products. For now we can wait until official words from AMD arrive next week as they will likely be holding an GPU Strategy conference with selected attendees from tech blogs.

Ya ur not biased or anything..