Details are almost nonexistent, but a new report claims that AMD has reached tape out for an upcoming Radeon 400 series of graphics cards, which could be the true successor to the R9 200-series after the rebranded 3xx cards.
Image credit: WCCFtech
According to the report:
"AMD has reportedly taped out two of its next-gen GPUs, with "Ellesmere" and "Baffin" both taping out – and both part of the upcoming Radeon 400 series of video cards."
I wish there was more here to report, but if this is accurate we should start to hear some details about these new cards fairly soon. The important thing is that AMD is working on the new performance mainstream cards so soon after releasing what was largely a simple rebrand accross much of the 300-series GPUs this year.
GPU side is progressing very
GPU side is progressing very well, the Fury was a very nice improvement.
Hope to hear similar news on Zen, being tapped out, and more imporantly more leaks on performance.
We need compention, all of us (Intel/AMD/Nvidia fans).
Being selfish, I need an upgrade soon so get to it AMD 🙂
Zen will probably not beat
Zen will probably not beat Intel in the IPC metric, but it will get close enough, and remember that Arctic Islands will have a new GPU ISA with even better asynchronous compute ability on the ACE units. So Intel will be at a disadvantage for GPU accelerated compute. AMD’s Zen Based APUs will have not only better CPU IPC performance but also the ability to send more work over to the ACE units. Having HSA 1.0+ certification and Unified memory Addressing on those ACE units will allow for better hardware asynchronos compute for gaming acceleration as well as other GPGPU workloads in addition to graphics. The VR gaming industry is already choosing AMD’s HSA aware APUs for their systems.
Keeping fingers crossed, we
Keeping fingers crossed, we so need the competition. AMD doesn’t have to beat Intel, just match or even reach 90% of its performance at same power level.
Still we’ll know more about Zen next year, and have them in hand toward end of 2016..
AMD has quoted numbers:
AMD has quoted numbers: http://wccftech.com/amd-officially-reveals-2016-cpu-roadmap-zen-k12/
As always take with a grain of salt, or perhaps the entire salt shaker.
I wonder if the late tape out
I wonder if the late tape out of the 400 series means they will only release a high end Arctic Islands part in 2016? If NVIDIA does a top down release of Pascal, AMD is fucked.
How long have we heard this
How long have we heard this tale before?
Both NVIDIA and AMD have done bottom-up and top-down releases of new architectures. It really doesn’t matter, they are both at the mercy of yields/binning, which a die shrink will most certainly make even more selective. They will both do what is most profitable combined and what will reduce fab costs as quickly as possible.
The overwhelming majority of GPUs sold are to OEMs, and cost is the only thing that matters to them. NVIDIA and AMD will just repackage old GPUs with new SKUs for higher margin, like always.
None of that matters. He’s
None of that matters. He’s an Nvidia fanboy who wanted to say “AMD is fucked” in the comment thread. That was the ONLY purpose for that post.
“AMD has reportedly taped out
“AMD has reportedly taped out two of its next-gen GPUs, with ‘Ellesmere’ and ‘Baffin’ both taping out – and both part of the upcoming Radeon 400 series of video cards.”
I count two next-gen GPUs, not one.
That may only be one new GPU
That may only be one new GPU but with two different memory configurations. Say one with HBM2 and the other with GDDR5 since it probably won’t make sense to include HBM2 on low end or even mid-range cards in 2016. So they need to continue to support GDDR5 but they can’t just slap a different type of memory on each GPU depending what the sale price will be so they have to design the GPU around the type of memory it will be using.
Its like saying Fiji and
Its like saying Fiji and Hawaii are the same GPU, one with HBM, the other GDDR5.
The memory controller is a massive part of a GPU design.
HBM also changes chip layout to support the interposer and the massive ‘pin’ count.
They might share architectural design for compute units, but thats where it ends.
Yea like mid and low range
Yea like mid and low range cards will be GDDR5, since they won’t be powerful enough to use speed HBM2 will offer. Nvidia can get away with pushing stack down using Maxwell in new cards where as AMD is forced to make a new chip since using re brands for 4th line of chips in a row wouldn’t be pleasant PR wise as they seen when they did it with 300 series.
AMD is going to take over the
AMD is going to take over the World.)
I am very happy, my AMD shares will be part of my pension money .)
MUST READ: (Link deleted)
MUST READ: (Link deleted)
Do not spam your site.
Do not spam your site.
The longer AMD is in the
The longer AMD is in the game, the better hardware buying options i will have. So more power to them, hope their products will be good.
AMD is the bang for the buck
AMD is the bang for the buck king!!! unless you are building something insanely high end, AMD is the way to go.
amd is bang for buck king?
amd is bang for buck king? Last I checked last year or so it was Nvidia that has forced prices down first with 970/980 and then again with 980ti.
i was kinda talking all about
i was kinda talking all about (processors too) since that convo was being tossed around too.
Who forced down prices before
Who forced down prices before that? Titan? gtx780ti? , it’s a loop get over it.
This is a joke ?
Mid range
This is a joke ?
Mid range card at 330 and 550$ under high end names is “forcing prices down” ?
Rly ?
RLY ?
New budget smartphone at
New budget smartphone at ₹9000 with Finger print Sensor- must read it (Link deleted)
You are a moron.
You are a moron.
I was just going to delete
I was just going to delete the spam post – but these replies are hilarious to me
Fine, I’ll leave them alone
Fine, I'll leave them alone too. 😉
About the 3xx
About the 3xx series…
Refreshed, not rebranded. Refreshed means that there’s small differences to the way they’re made, improvements to the manufacturing process, etc, which is what they did. As an example, the 390 and 390X, while the difference at first seems to be a RAM increase, there is also the high clocks that can be achieved due to manufacturing process improvement, and also the change in RAM manufacturer.
I agree on your definition
I agree on your definition that there should be small differences in silicone, but in this case that would require engineering work from AMD to change the silicone design. However what happened is that the 28nm process became more mature from TSMC, without any work from AMD. Just like the FX8370 was possible as Global Foundry process improved, and again no redesign from AMD was needed. In both cases giving a slight increase in clocks for the same power.
There were also 290x with 8GB from Sapphire. Just on the pricier size, and 8GB framebuffer was part of the R9 290x design. But again, no need for 8GB during that time until the 4k displays started to show up.
For AMD as they spent all the resources on the Fiji, they simply couldn’t do ANY engineering on the 290x. Intel did a good job during the Pentium 4 days to ensure AMD didn’t have the financial standing.. but I digress.
From the technical aspect this is just a rebrand.
This conjuration of reality
This conjuration of reality by fanboys starts to be fun.
Silicone is generally used to
Silicone is generally used to make large differences…
Anyway, even small process changes can require quite a bit of work to ensure that the design will still function properly. One thing to realize is that there are a lot of different jobs involved in the tape out of these chips. If you are going to have people on staff, then you want to make use of them all of the time. It is probably a completely separate team working on porting existing designs to a new process with other teams working on new designs. These jobs do not necessarily require the same skill set at all. Enthusiast say all manner of things, but they really have no idea what is going on behind the scenes. Scheduling of products to make good use of your engineering resources and budget is not a simple problem.
I think you should replace
I think you should replace enthusiasts with just people in general. People are just ignorant of their own ignorance.
I hope this is indeed brand
I hope this is indeed brand new stuff from AMD. Wasn’t the “200 series” just a “refresh” of the 7000 series????
I’m curious to see what 2016
I’m curious to see what 2016 brings in terms of new GPUs. I probably won’t buy one next year as I have a 980ti sli setup, but very curious to see if they get the nm down finally.
AMD is expected to using 14nm
AMD is expected to using 14nm and Nvidia is using 16nm.
It would make a lot of sense
It would make a lot of sense to get the smaller chips on the new process as soon as possible. The yeilds may not be that great, so a smaller die would be preferred at first. There are a lot more sales in mobile these days, and a smaller process obviously helps with power consumption. I could see AMD going after mobile much more aggressively, especially if Apple is interested in using their chips across their product line. It would be interesting to get a mid-range part with one design made to work with GDDR5 and one made to work on an interposer.
Hopefully, they have their big, high-end die ready in the same time frame or before Nvidia. The super high end parts are very low volume so they may still be able to manufacture them, even if the yields are low. Perhaps the large die 14 nm part will be priced above the the fury, and the fury will stay around at 28 nm for a while.
Any word if they’re going to
Any word if they’re going to re-write CCC? Nvidia’s software for managing driver updates is so much more reliable and friendly than AMD’s.