SoylentNews is people
SoylentNews
JEDEC has finalized the GDDR5X SGRAM specification:
The new technology is designed to improve bandwidth available to high-performance graphics processing units without fundamentally changing the memory architecture of graphics cards or memory technology itself, similar to other generations of GDDR, although these new specifications are arguably pushing the phyiscal[sic] limits of the technology and hardware in its current form. The GDDR5X SGRAM (synchronous graphics random access memory) standard is based on the GDDR5 technology introduced in 2007 and first used in 2008. The GDDR5X standard brings three key improvements to the well-established GDDR5: it increases data-rates by up to a factor of two, it improves energy efficiency of high-end memory, and it defines new capacities of memory chips to enable denser memory configurations of add-in graphics boards or other devices. What is very important for developers of chips and makers of graphics cards is that the GDDR5X should not require drastic changes to designs of graphics cards, and the general feature-set of GDDR5 remains unchanged (and hence why it is not being called GDDR6).
[...] The key improvement of the GDDR5X standard compared to the predecessor is its all-new 16n prefetch architecture, which enables up to 512 bit (64 Bytes) per array read or write access. By contrast, the GDDR5 technology features 8n prefetch architecture and can read or write up to 256 bit (32 Bytes) of data per cycle. Doubled prefetch and increased data transfer rates are expected to double effective memory bandwidth of GDDR5X sub-systems. However, actual performance of graphics cards will depend not just on DRAM architecture and frequencies, but also on memory controllers and applications. Therefore, we will need to test actual hardware to find out actual real-world benefits of the new memory.
What purpose does GDDR5X serve if superior 1st and 2nd generation High Bandwidth Memory (HBM) are around? GDDR5X memory will be cheaper than HBM and its use is more of an evolutionary than revolutionary change from existing GDDR5-based hardware.
(Score: 2) by takyon on Tuesday January 26 2016, @08:20AM
Computers may be "good enough" but AMD certainly has room to improve.
It has done well with the integrated graphics APU and we should see that continue. Skylake is supposedly a huge improvement, but the best Skylake "Iris Pro" 580 integrated graphics, GT4e, is nowhere to be found. The 'e' stands for eDRAM [wikipedia.org].
Oh, I guess they have gone on a publicity blitz for Iris Pro 580 / GT4e in the weeks since CES. They are saying it's a discrete graphics killer! And it might be for that (higher end?) budget segment. Luckily VR and 1440p/4K/5K/8K leaves open the "need" for better GPUs. Especially if any of that is desired on laptops or mobile devices.
http://wccftech.com/intel-skull-canyon-nuc-iris-pro-580-powerful-2016/ [wccftech.com]
http://wccftech.com/intel-iris-pro-graphics-gamers/ [wccftech.com]
They suggest that GT4e is better than what AMD has out on the market right now. I don't doubt that, but AMD's move to 14nm will give it plenty of room to improve their own integrated graphics. GT4e chips are also likely to be very expensive. While we have heard plenty about Zen's 40% increase in IPC (which could be misleading in some way), we don't know anything about Zen's integrated graphics. Theoretically, reducing transistors from 28nm to 14nm could mean 4 times the area (2x2)! Yeah, I know that's not actually true.
Back to improving chips. AMD needs to lower power consumption. Intel's recent Core-M chips are approaching the power consumption of its lower performance Atom chips, though they are much more expensive. Some of these lower-powered CPU brands may merge in the future. Like Atom and Celeron. Fanless laptops are great. I'm using an Intel Celeron N2830 based Chromebook. That has a TDP of 7.5 W, probably within reach of AMD. Let's see a 14nm AMD chip that can be put in a fanless tablet, laptop, or ultrabook. Ok so they do have [wikipedia.org] 4.5 Watt "Mullins" but they seem to be smoked by Intel Core-M. Compare E1 Micro-6200T to Core M-5Y10.
While you're right that core counts don't need to increase much, 8 core is a good milestone to reach. It future proofs the system and allows many applications/processes/tabs to run simultaneously. Also, both the Xbox One and PS4 are using 8-core x86 AMD "Jaguar" chips. While 1 of those cores aren't available to game developers, on the PS4 side, Sony has recently unlocked [wired.co.uk] the 7th core [extremetech.com]. So game developers have 6-7 cores to work with on millions of consoles. They may not be beefy cores, but it means we will see more parallelism in gaming. With the x86 architecture shared between consoles and PCs, you will see multi-platform games that can readily (hopefully?) take advantage of more than 4 cores. Productivity applications, video editing/encoding, servers, etc. can usually take advantage of additional cores.
Since Intel has frozen the core count at 4 for mainstream chips (6 and 8 for Sandy Bridge-E and Haswell-E), AMD could try and pull ahead in some benchmarks with new hex- and octo-core chips. They claimed to have made 6 and 8 core chips with the introduction of Bulldozer, but the dual-core "modules" were such a lie that they are being sued over it [theregister.co.uk]. Now that they are switching from Clustered Multithreading to Simultaneous Multithreading, Zen will probably include what can be considered "real" cores. They will probably make 6 and 8 core mainstream Zen models, although their mobile chips have been limited to quad-core (two modules).
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by Hairyfeet on Tuesday January 26 2016, @09:00AM
In what way, other than benches where I can produce article after article where Intel or the benchmarking company was caught rigging? Trusting benches anymore is about like the old "Quack.exe" on the bullshit scale.
To me a MUCH better metric is real life...can you name a single game that won't play on an FX-6300, a chip commonly found for sub $99? Or even the Athlon FM2 quad which goes for $79 and is recommended in many articles as the basis for budget gaming boxes? Lets face it, software, even games, just hasn't kept up with CPUs in a loooong time, and with the "next gen" consoles both using AMD netbook chips? I really do not see that changing. I used to have a buddy at the state capital that had a blind A/B test set up on a desk just to deal with those "AMD Suxors lulz" gaming tards where he had a $300 i5 on one side and a $99 FX-6300 on the other with both having $150 mainstream GPUs...know what anybody that tried it found? You literally could not tell the difference when it came to real world applications and games, the difference was so slight it was below the margin for error. Hell he could have easily swapped the FX for a Phenom II and I still doubt anybody would have known as for years I had a customer running high end GPU accelerated CAD designing robots and rockets to help out the local college on its NASA entries and it was all being done on a fricking Phenom I quad!
This is why I worry that AMD will end up coming out with the superior chip and we will see the Pentium 4 all over again, remember that? Intel had a clear dog on its hands yet it dominated the market because it rigged benches and paid off OEMs. The "2 billion" AMD settlement wasn't even 5% of what they made during that period so that obviously won't be a deterrent and the DOJ has been toothless for over a decade. This is why I've said for years AMD is doing it wrong because the first rule of business is you do NOT play the other guy's game, you make him play yours! I learned that lesson the hard way, by picking through the remains of other shops that all went broke trying to compete with Dell on price, did I do that shit? Fuck no! I instead pointed out that service from Dell was shit, it was filled with crapware, and would THEY come out to your house and set the system up for you at no extra charge? Nope.
For AMD to get ahead they need to stop trying to play the numbers game with Intel and instead show off what their chips can DO and how much you can save while doing itg. Show their laptops playing killer games while the Intel shit GPUs struggle while costing 40%+ more for the same hardware, show 2 systems built for $700 and how much nicer you get if you went AMD, with more RAM, SSD, dedicated GPU, etc. this strategy WORKS because we have proof, look at Nintendo SNES versus Genesis where in pure processing power the SNES had better hardware but Sega cleaned up because they refused to play the numbers game Nintendo's way, instead pushing how many cool games they could offer and how cool they looked.
So whether Zen is the second coming or not its really gonna depend on AMD not playing Intel's game. They have a pretty big advantage when it comes to GPUs and HBM should make it that much bigger of a gap but if they refuse to do anything but play the game by Intel's rules? They are gonna lose.
ACs are never seen so don't bother. Always ready to show SJWs for the racists they are.
(Score: 2) by meisterister on Wednesday January 27 2016, @01:52AM
They are saying it's a discrete graphics killer!
Aaand they said that for Broadwell, Haswell, Ivy Bridge, Sandy Bridge, Westmere, and, while not really fair because it is itself a discrete graphics solution with some weird integrated graphics-like features, the i740.
(May or may not have been) Posted from my K6-2, Athlon XP, or Pentium I/II/III.