Information & Technology

POLARIS


In a seemingly atypical move, AMD released the mid-range part first countering the usual halo launch followed by the volume parts. There have been a lot of unfounded rumors lately about AMD’s schedule and pending division sales, before you believe that nonsense look for ulterior motives from the ‘source’. Anyway here are the details.

First the technical side, >5 TF puts the RX480 at a higher performance level than both the PS4 and way way higher than the XBox One. It has 36 CUs or 2304 shaders in the old way of counting, mostly made possible by the 14nm Globalfoundries process. 4 or 8GB of GDDR5 on a 256b bus puts the RX580 on par with high-end cards of a generation or two ago and that isn’t counting some interesting low-level changes that we can’t talk about yet. Do note the DP1.3 and DP1.4HDR line though.

When you look at the power consumed, 150W, things get a lot more interesting. Not only is this a solid mid-range part, it should go nicely into notebooks downclocks or not. A binned part with different power management settings could effectively be dropped right into a gaming notebook with no other changes but a clock reduction would probably be smarter. Do note that the clocks aren’t directly listed but knowing that one shader can do two FLOPs/clock gets you ~1.1GHz (5000/2304*2) or about the range that you would expect. This is backed up by other data from Computex but don’t take any numbers solid. Unfortunately we can’t explain that part yet but stay tuned, we will, just don’t take numbers as hard or definitive this round yet.


Then there is the price. If you saw the AMD Computex webcast you know that AMD is aiming for volume and marketshare right now. If the high-end of their new introductions, that would be the RX480, is priced at $199 for the 4GB version, this should grab wads of marketshare for the company. We think it will, the price/performance ratio here is really solid, Nvidia will have a hard time competing with this for reasons we will talk about in a few days. This card and presumably others to come should hit the market sweet spot this round.


And the reveal: ZEN


What specs were revealed about Zen? 40% more IPC, presumably over Caririzo, 8 cores with 16 threads, and all for the new AM4 platform. The performance number should put Zen on par with Haswell, not bleeding edge but more than in the game, and the core count is borderline pointless for consumers but chopped dies will probably be really solid. That said it is obviously a ground-up new architecture and the first threaded core AMD is doing. Expect sea changes when the numbers come out. AMD also said the very different Zen server part will sample to a few OEMs in Q3 with mass samples in Q4, that should give you a nice window for launch.

Both of the new AMD offerings have a theme, vastly greater efficiency than their predecessors. Both are the first really new architectures we have seen from the company in a long, long time. Polaris 10/RX480 is a bit more iterative but a big step in functionality which should translate into marketshare as intended. Zen brings AMD back into the CPU core game, it is the first new architecture since Bulldozer. I can’t wait to see how they line up in real world testing, it has been too long since there was real competition in the enthusiast market.





































Nvidia GeForce GTX Titan 

"The Titan — the world's fastest supercomputer — has 18,688 Nvidia GK110 GPUs inside. Each and every one is more powerful than the graphics accelerator in most any gaming PC you'll find. That changes next week: Nvidia's announcing the $999 GeForce GTX Titan, a graphics card that harnesses the very same power to drive PC games. Wrapped in the same attractive aluminum chassis as the GeForce GTX 690, with the same specially designed cooling system and a polycarbonate window to show it off, the graphics card's a sight to behold, and the company claims it'll play some seriously strenuous games as well. Nvidia says it'll be on sale as soon as next week.

Before we tell you what the GeForce GTX Titan truly is, we should probably tell you what it is not. The GeForce GTX Titan might have the fastest GPU, but it's not necessarily the world's fastest graphics card. For the same $999 price as the Titan, you can pick up the aforementioned GeForce GTX 690, which combines two lesser GPUs on a single card, with 3,072 CUDA cores running at 915MHz by default. By contrast, a single GeForce GTX Titan is a little shy with 2,688 CUDA cores at 837MHz, and an Nvidia rep suggested that if you've got a beefy gaming PC with a single monitor attached, the GTX 690 would deliver better performance.

hat's only part of the story, though: with 6GB of GDDR5 memory and 384 bits of memory bandwidth (compare to 2GB and 256-bit per GPU in the GTX 690), the Titan could be better suited to deliver higher resolution games across multiple monitors. With a comparatively low 250W TDP and the custom cooling solution, Nvidia says it's also cool and quiet enough to fit into small form factor PCs like the Digital Storm Bolt and Falcon Northwest Tiki — unlike the comparatively hot and heavy GTX 690 or AMD Radeon HD 7970. A variety of OEMs (including both of them) will offer the Titan in living room sized computers soon. And yet where the Titan really shines, Nvidia says, is when you have three of them grouped together in an SLI configuration, where it should beat a pair of GTX 690s handily.

For $999, you'd no doubt expect incredible things out of a graphics card, much less three of them in a single ludicrously powerful gaming PC — it wasn't that long ago we found that a single $500 GeForce GTX 680 was nearly enough power to play the latest games across three 1600 x 900 monitors, let alone a single screen. $3,000 worth of GPUs should be overkill, no?

We'll see soon enough. Crysis 3 is out today, just in time to be (or escape being) Titan lunch."

SOURCE: theverge.com

RAM DDR4 is far!

First modules are done and functional. First corporations who will produce this babies are Samsung and Micron.

1.2 V unlike DDR3 1.5V

Capacity: starts from 4GB with chips made on 30nm technology.

Speeds: from 2400MT/s to 3200MT/s.

Production: 2013~2014
In stores: Unknown 

Ok, now the bad part!

You can't expect to upgrade. Why? PC's don't have the support yet. It seems Intel will make the support available for DDR4 in 2014 when it will release the new chipset Haswell EX for servers and in 2015 for PC's when it will release the chipset Broadwell on the 14nm technology.

High hopes for a faster adoption of the standard DDR4 comes from manufacturers like AMD, Nvidia, Qualcomm and TI, but no time is set yet.

   Its primary benefits compared to DDR3 include a higher range of clock frequencies and data transfer rates (2133–4266 MT/s compared to DDR3's 800 and higher) and lower voltage (1.05–1.2 V for DDR4, compared to 1.2–1.65 V for DDR3) with current remaining the same. DDR4 also anticipates a change in topology. It discards the multiple DIMMs per channel approach in favor of a point-to-point topology where each channel in the memory controller is connected to a single DIMM.

DDR4 memory comes in a 284-pin DIMM, similar to a 240-pin DDR-2/DDR-3 DIMM.[31]:11 The pins are spaced more closely (0.85 mm instead of 1.0) to fit more within the standard 51⁄4-inch (133.35 mm) DIMM width, the height is increased slightly (31.25 mm/1.23 in instead of 30.35 mm/1.2 in) to make signal routing easier, and the thickness is increased (to 1.2 mm from 1.0) to accommodate more signal layers.

DDR4 SO-DIMMs have 256 pins (rather than 204), spaced closer (0.5 mm rather than 0.6), and are 1.0 mm wider (68.6 mm rather than 67.6), but retain the same 30 mm height.

Internal banks are increased to 16 (4 bank select bits), with up to 8 ranks per DIMM.[31]:16

Protocol changes include:

Parity on the command/address bus
Data bus inversion (like GDDR4)
CRC on the data bus
Independent programming of individual DRAMs on a DIMM, to allow better control of on-die termination.