Subject: Processors | January 18, 2015 - 05:16 PM | Sebastian Peak
Tagged: SoC, rumor, processor, leak, iris pro, Intel, graphics, cpu, carrizo, APU, amd
A new report of leaked benchmarks paints a very interesting picture of the upcoming AMD Carrizo mobile APU.
Image credit: SiSoftware
Announced as strictly mobile parts, Carrizo is based on the next generation Excavator core and features what AMD is calling one of their biggest ever jumps in efficiency. Now alleged leaked benchmarks are showing significant performance gains as well, with numbers that should elevate the IGP dominance of AMD's APUs.
Image credit: WCCFtech
"The A10 7850K scores around 270 Mpix/s while Intel’s HD5200 Iris Pro scores a more modest 200 Mpix/s. Carriso scores here over 600 Mpix/s which suggests that Carrizo is more than twice as fast as Kaveri and three times faster than Iris Pro. To put this into perspective this is what an R7 265 graphics card scores, a card that offers the same graphics performance inside the Playstation 4."
While the idea of desktop APUs with greatly improved graphics and higher efficency is tantalizing, AMD has made it clear that these will be mobile-only parts at launch. When asked by Anandtech, AMD had this to say about the possibility of a desktop variant:
“With regards to your specific question, we expect Carrizo will be seen in BGA form factor desktops designs from our OEM partners. The Carrizo project was focused on thermally constrained form factors, which is where you'll see the big differences in performance and other experiences that consumers value.”
The new mobile APU will be manufactured with the same 28nm process as Kaveri, with power consumption up to 35W for the Carrizo down to a maximum of 15W for the ultra-mobile Carrizo-L parts.
Subject: Graphics Cards | January 13, 2015 - 12:22 PM | Ryan Shrout
Tagged: rumor, radeon, r9 380x, 380x
Spotted over at TechReport.com this morning and sourced from a post at 3dcenter.org, it appears that some additional information about the future Radeon R9 380X is starting to leak out through AMD employee LinkedIn pages.
Ilana Shternshain is a ASIC physical design engineer at AMD with more than 18 years of experience, 7-8 years of that with AMD. Under the background section is the line "Backend engineer and team leader at Intel and AMD, responsible for taping out state of the art products like Intel Pentium Processor with MMX technology and AMD R9 290X and 380X GPUs." A bit further down is an experience listing of the Playstation 4 APU as well as "AMD R9 380X GPUs (largest in “King of the hill” line of products)."
Interesting - though not entirely enlightening. More interesting were the details found on Linglan Zhang's LinkedIn page (since removed):
Developed the world’s first 300W 2.5D discrete GPU SOC using stacked die High Bandwidth Memory and silicon interposer.
Now we have something to work with! A 300 watt TDP would make the R9 380X more power hungry than the current R9 290X Hawaii GPU. High bandwidth memory likely implies memory located on the substrate of the GPU itself, similar to what exists on the Xbox One APU, though configurations could differ in considerable ways. A bit of research on the silicon interposer reveals it as an implementation method for 2.5D chips:
There are two classes of true 3D chips which are being developed today. The first is known as 2½D where a so-called silicon interposer is created. The interposer does not contain any active transistors, only interconnect (and perhaps decoupling capacitors), thus avoiding the issue of threshold shift mentioned above. The chips are attached to the interposer by flipping them so that the active chips do not require any TSVs to be created. True 3D chips have TSVs going through active chips and, in the future, have potential to be stacked several die high (first for low-power memories where the heat and power distribution issues are less critical).
An interposer would allow the GPU and stacked die memory to be built on different process technology, for example, but could also make the chips more fragile during final assembly. Obviously there a lot more questions than answers based on these rumors sourced from LinkedIn, but it's interesting to attempt to gauge where AMD is headed in its continued quest to take back market share from NVIDIA.
Subject: Graphics Cards | January 6, 2015 - 09:44 AM | Sebastian Peak
Tagged: rumor, nvidia, leak, gtx 960, GM206, geforce
VideoCardz.com is reporting that they not only know the upcoming GTX 960 core will be the GM206, but they reportedly have a photo of the unreleased chip.
Why are reported leaks always slightly out of focus? (Credit: VideoCardz.com)
The chip pictured appears to be a GM206-300, which the site claims will be the exact variant in the GTX 960 when it is released. The post speculates that based on the die size we can expect between 8 - 10 SMM's, or 1080 - 1280 CUDA cores. They further claim that the GTX 960 will have a 128-bit memory bus and reference cards will have a 2GB frame buffer (though naturally we can expect models with 4GB of memory after launch).
The post goes on to show what appears to be a search result for an ASUS GTX 960 on their site, but if this existed it has since been taken down. More than likely a GTX 960 is in fact close at hand, and the reported specs (and now multiple claimed listings for the card) are not hard to fathom.
We will keep you updated on this alleged new GPU if more details emerge.
Subject: Graphics Cards | January 21, 2014 - 03:49 PM | Ryan Shrout
Tagged: rumor, nvidia, kepler, gtx titan black, gtx titan, gtx 790
How about some fresh graphics card rumors for your Tuesday afternoon? The folks at VideoCardz.com have collected some information about two potential NVIDIA GeForce cards that are going to hit your pocketbook hard. If the mid-range GPU market was crowded wait until you see the changes NVIDIA might have for you soon on the high-end.
First up is the NVIDIA GeForce GTX TITAN Black Edition, a card that will actually have the same specifications as the GTX 780 Ti but with full performance double precision floating point and a move from 3GB to 6GB of memory. The all-black version of the GeForce GTX 700-series cooler is particularly awesome looking.
Image from VideoCardz.com
The new TITAN would sport the same GPU as GTX 780 Ti, only TITAN BLACK would have higher double precision computing performance, thus more FP64 CUDA cores. The GTX TITAN Black Edition is also said to feature 6GB memory buffer.This is twice as much as GTX 780 Ti, and it pretty much confirms we won’t be seeing any 6GB Ti’s.
The rest is pretty much well known, TITAN BLACK has 2880 CUDA cores, 240 TMUs and 48 ROPs.
VideoCardz.com says this will come in at $999. If true, this is a pure HPC play as the GTX 780 Ti would still offer the same gaming performance for enthusiasts.
Secondly, there looks to be an upcoming dual-GPU graphics card using a pair of GK110 GPUs that will be called the GeForce GTX 790. The specifications that VideoCardz.com says they have indicate that each GPU will have 2496 enabled CUDA cores and a smaller 320-bit memory interface with 5GB designated for each GPU. Cutting back on the memory interface, shader counts and even clocks speeds would allow NVIDIA to manage power consumption at the targeted 300 watt level.
Image from VideoCardz.com
Head over to VideoCardz.com for more information about these rumors but if all goes as they expect, you'll hear about these products quite a bit more in February and March.
What do you think? Are these new $1000 graphics cards something you are looking forward to?
Subject: Graphics Cards | April 15, 2013 - 03:34 PM | Tim Verry
Tagged: rumor, nvidia, kepler, gtx 700, geforce 700, computex
Recent rumors seem to suggest that NVIDIA will release its desktop-class GeForce 700 series of graphics cards later this year. The new card will reportedly be faster than the currently-available GTX 600 series, but will likely remain based on the company's Kepler architecture.
According to the information presented during NVIDIA's GTC keynote, its Kepler architecture will dominate 2012 and 2013. It will then follow up with Maxwell-based cards in 2014. Notably absent from the slides are product names, meaning the publicly-available information at least leaves the possibility of a refreshed Kepler GTX 700 lineup in 2013 open.
Fudzilla further reports that NVIDIA will release the cards as soon as May 2013, with an official launch as soon as Computex. Having actual cards available for sale by Computex is a bit unlikely, but a summer launch could be possible if the new 700 series is merely a tweaked Kepler-based design with higher clocks and/or lower power usage. The company is rumored to be accelerating the launch of the GTX 700 series in the desktop space in response to AMD's heavy game-bundle marketing, which seems to be working well at persuading gamers to choose the red team.
What do you make of this rumor? Do you think a refreshed Kepler is coming this year?
Subject: Graphics Cards | January 22, 2013 - 02:44 PM | Ryan Shrout
Tagged: nvidia, geforce, gk110, titan, rumor
A combination of rumors and news pieces found online and in some recent conversations with partners indicates that February will see the release of a new super-high-end graphics card from NVIDIA based on the GK110 GPU. Apparently using the name "Titan" based on a report from Sweclockers.com, this new single GPU card will feature 2688 CUDA cores, compared to the 1536 in the GeForce GTX 680.
If true, the name Titan likely refers to the Cray super computer of the same name built using GK110 Kepler Tesla cards. Sweclockers.com's sources are quoted with the clocks of this new super-GPU as well: 732 MHz core clock and 5.2 GHz GDDR5 memory clock. While those numbers are low compared to the 1000+ MHz speeds of the GK104 parts out today, this GPU would have 75% more compute units and presumably additional memory capacity as well. The memory bus width of 384-bits is a 50% increase as well which would indicate another big jump in performance over current cards. The CUDA core count of 2688 is actually indicative of a GK110 GPU with a single SMX disabled as well.
The NVIDIA Titan card will apparently be the replacement for the GeForce GTX 690, a dual-GK104 card launched in May of last year. The performance estimate for the Titan is approximately 85% of that GTX 690 and if the rumors are right it would see an $899 price tag.
Based on other conversations I have had recently you should only expect those same partners that were able to sell the GTX 690 to stock this new GK110-based part. There won't be any modifications and you will see very little differentiation between vendors branding on it. If dates are to be believed, we are hearing that a Feb 25th (or at least that week) launch is the current target.
Subject: Graphics Cards | July 7, 2012 - 05:07 AM | Tim Verry
Tagged: rumor, hd 7950 GHz edition, graphics card, gpu boost, gpu, amd
An interesting rumor floating around the internet concerns a new AMD 7950 graphics card. Specifically, a forum post over at EXPReview hinted at a new GHz Edition 7950 card. They reference a source and a photo of a graphics card with GHz Edition markings and two six-pin PCI-E power connectors. They further quoted a source that said AMD would be releasing a fourth card to its 7000-series lineup. In other words, there was not a whole lot to go on.
The alleged AMD Radeon HD 7950 GHz Edition graphics card.
I saw the rumor a couple of days ago, but decided to let it sink in for a bit before writing about it. After some thought and internal conversations, I think a 7950 GHz edition is plausible, and certainly a possibility. With the addition of the HD 7970 GHz Edition (which we recently reviewed), AMD now has a 7900 series part with its new GPU Boost (aka Powertune Boost) technology. Unfortunately, that tech can not simply be rolled out to the other existing 7900 series cards meaning that the original 7950 and 7970 are stuck without boost, leaving a gap in its lineup. Because of that, it seems logical that AMD would want to fill in that gap with a refreshed 7950 part that has higher clocks and GPU boost.
The 7970 GHz Edition helped AMD stay competitive with the NVIDIA GeForce GTX 680 and enabled AMD to (re)claim it had the fastest single GPU card. That card was enabled by AMD finding extra headroom and higher bins in its 28nm GPUs. The current Radeon HD 7950 runs at an 800 MHz GPU clock speed and a 5GHz GDDR5 effective clock speed. It is unclear if the company’s chips have enough headroom that they are able to overclock 7950 oriented silicon or if they are simply reworking the bins such that the 7950 GHz Edition is either using silicon that just barely can’t make the cut as a 7970 GHz Edition or AMD is artificially locking down 7970 GHz chips to make these proposed 7950 GHz GPUs. Apart from the increased clock speeds and GPU Boost, the rumored card is said to maintain the same hardware specifications as the existing 7950. Mainly, that means 1,792 stream processors, 112 texture units, and the same number of transistors – 4.31 billion – on a 28nm process. In comparison, the 7970 has 2,048 stream processors and 128 texture units.
As mentioned above, such a chip sounds like a logical step for AMD. By refreshing the 7950 with higher clock speeds and adding GPU Boost functionality it will be fight with the GeForce 670 for the performance crown, and may even come close to 680/7970 performance when overclocked further. I do have a couple of questions about this rumored card, however. Specifically, I am wondering if AMD will simply refresh the 7900 series cards and let the original releases fade out as the stock dwindles or if they will keep them around. I’m leaning towards the later but it will still be interesting to see, and it also makes me wonder how much a 7950 GHz Edition will cost me.
I’m curious what you think about the rumor though, do you think it has any merit? And assuming this is legit, where do you think it will stand performance-wise versus NVIDIA’s products?
If you are curious about how AMD’s new GPU Boost technology works, be sure to check out our Live Review of the Radeon HD 7970 GHz Edition with AMD’s Evan Groenke where we are given a live demonstration showing how it works and some background on the technology.
Subject: Graphics Cards | January 2, 2012 - 07:23 PM | Ryan Shrout
Tagged: amd, southern islands, rumor, leak
The review for the AMD Radeon HD 7970 3GB card based on the Tahiti GPU and the new Southern Islands architecture was released on December 22nd with expected availability on January 9th. In that review we show a diagram of the Tahiti GPU and its 32 Compute Units (CUs) that combine to form the total 2048 stream processors (SPs).
We asked and asked, but a die shot was never given to us for our review - a very non-standard practice for new launches. That started us wondering - was there something AMD was hiding from reviewers about the architecture? Were there some disabled CUs on the 28nm GPU that they had disabled for business, yield or clock speed reasons? Think of what Intel has done with Sandy Bridge-E or what NVIDIA did originally with the GTX 480 GPU.
AMD assured us that was not the case - Tahiti is the full die enabled, 32 CUs and 2048 SPs. And, based on some of our own internal information, that seems to be 100% the case.
But, an interesting image started floating around last week:
This image from the site ChipHell.com appears to show the development sheets for Sapphire's upcoming Radeon HD 7000 series products and their internal codenames. There are some really noteworthy things to look at though starting with the Atomic lineup.
While the Flex 6G is a 6GB card with 6 mini-DP ports on it running at the same clock speeds as our reference designs did initially, the Atomic RX card has a clock speed of 1335 MHz running on 2048 SPs and a pretty good memory overclock as well. If that is accurate, the performance difference between the Atomic RX and "Da Original" (likely the reference card) would be tremendous!
Here is what is more interesting - another card listed above the HD 7970s that seems to include 2304 SPs, or 36 CUs. Running at a reference speed of 1000 MHz, this card would have a noticeable advantage over the current HD 7970 cards. What's more...?
The Toxic ZX, if it exists, would run with 2304 stream processors at 1225 MHz! The performance of this card could easily beat out the Radeon HD 7970 3GB card by 35-45% with the shader and clock speed differences.
So, what does this all mean? Probably nothing, but it is fun to speculate on a few things. It seems possible that AMD either HAD or HAS another GPU waiting in the wings based on Southern Islands to compete with NVIDIA's Kepler when it finally gets released. Even though these documents seem to indicate that, I kind of find it hard to believe that AMD would have been able to keep this secret from the media and the competition for this long. It is also equally unlikely that AMD was able to quickly tape out another chip that we are unaware of as even a somewhat moderate change like adding in four very modular CUs takes many months.
And of course, we have to take in the possibility that these are all fake, or a decoy or were written up 18 months ago and plans have changed. Those are much less fun though.
Subject: Motherboards | June 14, 2011 - 06:35 PM | Tim Verry
Tagged: x79, rumor, lga 2011, lga 1366, Intel, cpu
Xbit Labs recently detailed a new rumor concerning Intel’s upcoming X79 chipset. According to a leaked document viewed by them, X79 will support both Intel’s current and upcoming high end processors sockets in the form of LGA 1366 and LGA 2011. What this means for the end user is that they will be able to purchase a x79 based motherboard that will support either Nehalem or Sandy Bridge-E processors, unless motherboard manufacturers decide to splurge and include both sockets on one board like the Asus’ concept board shown at Computex 2011. This means that while DIY enthusiasts and gamers are not likely to use these motherboards as an upgrade path to Sandy Bridge (as a CPU upgrade would likely still necessitate a motherboard upgrade due to both sockets not being physically present), IT departments will likely appreciate the continued support of the older 1366 processors on new motherboards as it will make replacement parts easy to find for high end 1366 based workstations.
On the other hand, manufacturers will benefit the most from the X79 chipset supporting multiple sockets, and thus reducing costs. This cost reduction may then allow for cheaper end-user costs.
Intel itself is planning to manufacture two X79 motherboards named the DX79SI and DX79TO, will each support LGA 1366 and LGA 2011 respectively. Xbit Labs reports that the DX79SI board is planned to be a feature packed LGA 2011, no-compromise affair, with support for up to 64GB of RAM (eight DIMM slots), three PCI-E 3.0 slots for multi-GPU configurations, 12 SATA (six SATA 3 6GB/s, six SATA 2 3GB/s) ports, four USB 3.0, 14 USB 2.0, 8-channel audio, Wifi and Bluetooth, and two Gigabit Ethernet connections.
In contrast, the DX79TO will feature a LGA 1366 socket, and brings two PCI-E 2.0 x16 slots, 8 SATA connectors (likely four SATA 3, four SATA 2), 2 USB 3.0, 6-channel audio, a single Gigabit Ethernet connection, and DDR3 memory support (there are no details on the exact DIMM configuration supported yet).
By lowering the cost of supporting two high-end CPU lines and platforms, Intel, motherboard manufacturers, and consumers likely have a win-win-win situation, providing that the rumor comes to fruition.