Battle of the IGPs
Our long journey with Frame Rating, a new capture-based analysis tool to measure graphics performance of PCs and GPUs, began almost two years ago as a way to properly evaluate the real-world experiences for gamers. What started as a project attempting to learn about multi-GPU complications has really become a new standard in graphics evaluation and I truly believe it will play a crucial role going forward in GPU and game testing.
Today we use these Frame Rating methods and tools, which are elaborately detailed in our Frame Rating Dissected article, and apply them to a completely new market: notebooks. Even though Frame Rating was meant for high performance discrete desktop GPUs, the theory and science behind the entire process is completely applicable to notebook graphics and even on the integrated graphics solutions on Haswell processors and Richland APUs. It also is able to measure performance of discrete/integrated graphics combos from NVIDIA and AMD in a unique way that has already found some interesting results.
Battle of the IGPs
Even though neither side wants us to call it this, we are testing integrated graphics today. With the release of Intel’s Haswell processor (the Core i7/i5/i3 4000) the company has upgraded the graphics noticeably on several of their mobile and desktop products. In my first review of the Core i7-4770K, a desktop LGA1150 part, the integrated graphics now known as the HD 4600 were only slightly faster than the graphics of the previous generation Ivy Bridge and Sandy Bridge. Even though we had all the technical details of the HD 5000 and Iris / Iris Pro graphics options, no desktop parts actually utilize them so we had to wait for some more hardware to show up.
When Apple held a press conference and announced new MacBook Air machines that used Intel’s Haswell architecture, I knew I could count on Ken to go and pick one up for himself. Of course, before I let him start using it for his own purposes, I made him sit through a few agonizing days of benchmarking and testing in both Windows and Mac OS X environments. Ken has already posted a review of the MacBook Air 11-in model ‘from a Windows perspective’ and in that we teased that we had done quite a bit more evaluation of the graphics performance to be shown later. Now is later.
So the first combatant in our integrated graphics showdown with Frame Rating is the 11-in MacBook Air. A small, but powerful Ultrabook that sports more than 11 hours of battery life (in OS X at least) but also includes the new HD 5000 integrated graphics options. Along with that battery life though is the GT3 variation of the new Intel processor graphics that doubles the number of compute units as compared to the GT2. The GT2 is the architecture behind the HD 4600 graphics that sits with nearly all of the desktop processors, and many of the notebook versions, so I am very curious how this comparison is going to stand.
Subject: General Tech, Processors, Mobile | July 5, 2013 - 03:50 PM | Scott Michaud
Tagged: qualcomm, Intel, Bay Trail-T, Bay Trail
Bay Trail is still seasons away but engineering samples are, and this should be no surprise, already in use at least for research and development purposes. Someone, somewhere down the line, decided to run a benchmark which was posted online. AnTuTu, the benchmark utilized, measures a spread of factors including memory, integer performance, floating point performance, 3D performance, and so forth. Unfortunately it does also include some non-CPU/GPU factors in its score, albeit barely, so best take it with a grain of salt.
Image Credit: The Droid Guy
The Silvermont-based chip, clocked at an... actually quite modest 1101 MHz, received a synthetic score of 43416. To put that in comparison: arguably the fastest ARM processor on the market, the Qualcomm's Snapdragon 800, tends to find itself with a score around the 30,000-32,000 range which is about 27-31% slower than Intel. The very popular albeit soon deprecated Nexus 7, powered by the Tegra 3, scores 12726.
Personally, I am getting a little flashback of the Intel vs. AMD battle about 8 years ago. We seem to be close to a Conroe (Core 2 Duo) vs. AMD Athlon 64 FX point between Intel and ARM. Intel eclipsed the AMD Athlon 64
FX-57 (update: I meant FX-62) and kept throwing more money at research than AMD could possibly afford. Unless ARM can severely undercut Bay Trail, Intel could follow past trends and simply bury their competitors with tens of billions in capital investment until their products are so far ahead that consumers default to Intel products.
If history repeats itself, this leaves Qualcomm and others in a difficult position. The solution seems to be either to tread water in a price point that Intel ignores or to collectively dump money into ARM and run the "out-research Intel" treadmill. Remember, this is a company who will dump twice AMD's revenue into their Research and Development year-over-year to keep ahead. Unlike Intel's GPU efforts, which did not seem like a problem that cash could solve alone, they know how to make processors.
I would not make business decisions under the assumption x86 will keep Intel hobbled indefinitely.
Subject: General Tech, Processors | July 3, 2013 - 03:12 AM | Scott Michaud
Tagged: Richland, APU, amd
Accidents happen. AMD has been rolling out their Richland APUs for the last month and partners have been keeping up with supporting products. While common, the problem with rolling releases is the potential confusion over what has and what has not been released. Unfortunately for MSI, their support chart for FM2 CPUs includes a couple of products which are news to us.
AMD will be able to hit the 45W TDP with the, apparently, upcoming A8-6500T and A10-6700T APUs. Tom's Hardware seemed to have slightly different information, their chart does not exactly jive with the one posted by MSI; for instance, they claimed the T suffix implied a low power variant when MSI's chart confirmed a 45W TDP... fairly loud and clear. As such, our table will be my best attempt at combining both charts along with a bit more leaked GPU information from TechPowerUP.
|Base Clock||2.1 GHz||2.5 GHz|
|L2 Cache||4 MB||4 MB|
|L3 Cache||0 MB (N/A)||0 MB (N/A)|
Radeon HD 8550D
(Not HD 8650D)
|Radeon HD 8650D|
|GPU Clock||720 MHz||720 MHz|
|GPU Boost Clock||844 MHz (???)||844 MHz (???)|
|GPU Shader Count||256||384|
It is impossible to know expected price, release window, or even whether the product still exists. For that, we will need to wait for an official unveiling... or at least another unofficial one.
Subject: General Tech, Processors, Systems | June 26, 2013 - 10:27 PM | Scott Michaud
Tagged: supercomputing, supercomputer, titan, Xeon Phi
The National Supercomputer Center in Guangzho, China, will host the the world's fastest supercomputer by the end of the year. The Tianhe-2, English: "Milky Way-2", is capable of nearly double the floating-point performance of Titan albeit with slightly less performance per watt. The Tianhe-2 was developed by China's National University of Defense Technology.
Photo Credit: Top500.org
Comparing new fastest computer with the former, China's Milky Way-2 is able to achieve 33.8627 PetaFLOPs of calculations from 17.808 MW of electricity. The Titan, on the other hand, is able to crunch 17.590 PetaFLOPs with a draw of just 8.209 MW. As such, the new Milky Way-2 uses 12.7% more power per FLOP than Titan.
Titan is famously based on the Kepler GPU architecture from NVIDIA, coupled with several 16-core AMD Opteron server processors clocked at 2.2 GHz. This concept of using accelerated hardware carried over into the design of Tianhe-2, which is based around Intel's Xeon Phi coprocessor. If you include the simplified co-processor cores of the Xeon Phi, the new champion is the sum of 3.12 million x86 cores and 1024 terabytes of memory.
... but will it run Crysis?
... if someone gets around to emulating DirectX in software, it very well could.
Subject: Processors | June 21, 2013 - 09:39 AM | Tim Verry
Tagged: Intel, haswell, cpu, Broadwell, 14nm
Alongside the good news of 8-core Haswell-E parts, VR-Zone revealed an updated Intel road map that makes no mention of the 14nm Haswell architecture die shrink Broadwell. Broadwell was originally intended to be the next "tick" in Intel's yearly "tick-tock" chip release schedule set to release next year. If recent reports are true, this will no longer be the case. Instead, 2014 will be dominated (at least on the Intel side of things) by consumer Haswell and enthusiast-grade Haswell-E chips.
What is going on with Broadwell?
Broadwell is essentially supposed to be a CPU using the Haswell micro-architecture that is built on a (impressively) smaller 14nm manufacturing process. There may be a few minor tweaks to the architecture or updates to the instruction set extensions, but the big difference between Broadwell and Haswell is the die shrink from 22nm to 14nm. The die shrink will allow for better low-power performance and will be beneficial in battery-powered mobile devices first and foremost. Likely as a result of the main benefits being mobile parts, Intel has previously announced that Broadwell chips would be BGA only, which means that there would not be a traditional LGA socket-ed desktop part. Broadwell chips would only come soldered onto motherboards in bare-bones systems, laptops, and tablets for example.
Despite the small architectural differences, the die shrink alone is a monumental task. Intel needs to not only be able to shrink Haswell and its wealth of transistors to 14nm, but it has to do so in a way that allows them to get the yields and power efficiency characteristics that they want. This is extremely hard, and the move to manufacturing nodes below 22nm is going to get exceedingly difficult. Intel accomplished 22nm with its Tri-gate 3D transistors, but with 14nm they are going to have to push beyond that, and even with its huge money vault, physics is working against them in a big way here. As a result of the huge challenges of moving to 14nm, it seems at this point that Broadwell will not be ready in time for a 2014 launch after all. Instead, Intel is now shooting for a 2015 launch of the BGA Broadwell chips alongside the LGA (socket-ed) 14nm Sky Lake processors (the "tock" to Broadwell's "tick").
Some enthusiasts and media have painted the Broadwell delay to be, at least in part, due to less competition from AMD. That is possible, but I can't help but thinking that slowing down Broadwell is the last thing Intel would want to do. The sooner Intel is able to move its Haswell (and future) micro-architecture-based chips to 14nm and beyond, the sooner AMD is put all that much farther behind. If Intel had managed 14nm Broadwell in 2014, AMD would have been screwed out of a lot of SFF NUC-type systems as well as mobile devices as they would not really be able to compete on performance or power efficiency! (Then Intel could happily focus on trying to bring down ARM in the mobile space, which it seems to want to do heh.) In some internal discussion with PC Perspective's Josh Walrath, I think that Intel would have loved to bring 14nm chips next year but, because of manufacturing process woes, the chips are simply not ready.
The new plan: Refresh Haswell in 2014 with a new Z97 chipset
Now, with the launch of Broadwell moved back to at least 2015, consumers will now be presented with a refresh of 22nm Haswell chips on the consumer side around Q2 2014 and the upcoming launch of enthusiast-platform Haswell-E processors in the second half of 2014.
The Haswell (LGA 1150) refresh will include better binned chips with a lineup that is likely to see a slight speed bump in stock clockspeed across the board as well as an updated Z97 chipset. The new chipset will support 1000 MB/s SATA Express and boot-level malware protection technology in the form of Intel Device Protection and Boot Guard. Granted motherboards using the updated Z97 chipset are not going to be all that alluring to those users already running Z87 chipsets with their Haswell processors. However, users that have not yet upgraded might as well go with the newer chipset and enjoy the small tweaks and benefits that go along with it. In other words, if you were holding out waiting to upgrade to a Broadwell CPU plus motherboard combo, you are going to be waiting at least another year. You will be able to grab a refreshed Haswell CPU and a Z87 or Z97 chipset-based motherboard next year though (which should still be a healthy upgrade if you have a pre-Sandy Bridge system).
Also worth noting is that if you have already upgraded to Haswell, you can rest easy knowing that you have at least another year of your chip being the newest model--quite a feat considering how fast the tech world traditionally moves!
On the other hand, if Haswell just isn't fast enough, there is always Haswell-E to look forward to in 2014! Haswell-E will bring 8-core, 16-thread chips with 20MB of L3 cache (up to ~140W TDP) and the X99 chipset, which should keep the top-end enthusiast market happy no matter the state of Broadwell.
I'm looking forward to more details regarding the 14nm manufacturing process, and hoping that once the chips are on the way the company will be willing to talk about some of the challenges and issues they faced moving to such a small process node (perhaps at IDF? One can hope.) In the mean time, Haswell has another year to shine and make Intel money while AMD works on its HSA and APU strategies.
What do you think about the 14nm Broadwell delay? Does it impact you, or were you waiting for Haswell-E anyway?
Subject: General Tech, Processors | June 19, 2013 - 08:37 PM | Scott Michaud
Tagged: overclock, amd
Thankfully, they were not "firing" on all four cylinders; while Ryan does prefer thermite, overclockers tend to prefer liquid nitrogen. There are some distinct advantages of ice over fire, the main one for computer users is the potential for massive bumps in frequency and voltage. Of course, you cannot really get any effective use out of a machine that relies on a steady stream of fluid cold enough that it takes less digits to write out its temperature in Kelvin, but a large bump makes good bragging rights.
Finnish overclocker, "The Stilt", managed to push his four-core part to 8000.39 MHz just long enough to have CPU-Z validate his accomplishment. With a frequency multiplier of 63.0 atop a bus speed of 126.99, this gets within 800MHz of the AMD FX-8350 running on just one module (6 of 8 cores disabled) recorded by ASUS late last year.
But no, it will probably not run Crysis.
OpenCL Support in a Meaningful Way
Adobe had OpenCL support since last year. You would never benefit from its inclusion unless you ran one of two AMD mobility chips under Mac OSX Lion, but it was there. Creative Cloud, predictably, furthers this trend with additional GPGPU support for applications like Photoshop and Premiere Pro.
This leads to some interesting points:
- How OpenCL is changing the landscape between Intel and AMD
- What GPU support is curiously absent from Adobe CC for one reason or another
- Which GPUs are supported despite not... existing, officially.
This should be very big news for our readers who do production work whether professional or for a hobby. If not, how about a little information about certain GPUs that are designed to compete with the GeForce 700-series?
Subject: General Tech, Processors | June 17, 2013 - 08:11 PM | Scott Michaud
Tagged: haswell, Intel, Second Opinion
Ryan reviewed the Core i7 4770K earlier in the month and found it an impressive product. He was not able to properly test the CPU paired with a discrete GPU because of time restraints; we value results measured from direct monitor output, which takes longer than FRAPS and other software results. Still, Ryan believes that the boost in raw CPU performance justifies its existence in desktops without a funky "-E" tagged along for good luck.
For a second opinion, you could check NitroWare to see what a cynical Aussie thinks of Intel's latest offering. Of note, they compare software-measured frame rates between the on-chip GPU and those measured from a GTX 460 on Sandy Bridge, Ivy Bridge, and Haswell. He is nothing if not thorough, collecting his findings over 20 pages.
Ultimately he finds that if you are running Ivy Bridge, you will not benefit too much from the upgrade; Sandy Bridge users and earlier, on the other hand, might want to consider this platform... unless they are wanting to jump into the enthusiast-slot offerings coming up late this year and Haswell-E late the following year.
Also be sure to check back when we have our frametime measurements complete!
Subject: Editorial, General Tech, Processors | June 15, 2013 - 07:02 PM | Scott Michaud
Tagged: Intel, Ivy Bridge-E, Haswell-E
In my analysis of the recent Intel Computex keynote, I noted that the displayed confidence came across more as repressing self-doubt. It did not seem, to me, like Intel wants to abandon the high-end enthusiast but rather catch up with their low performance and high efficiency competitors; they just know they are secure in that market. Of course, we could see mid-range choices dwindle and prices stagnate, but I cast doubt that Intel wants to exit the enthusiast market despite their silence about Ivy Bridge-E.
All Images, Credit: VR-Zone
And Intel, now, wants to return some confidence to their high-end consumers comma they are not slowing down exclamation point exclamation point.
VR-Zone, the site which published Ivy Bridge-E's lazy release roadmap, are also the ones to suggest Haswell-E will come before mainstream Broadwell offerings. Once again, all is right with the world. Slated for release around holiday 2014, just a year after Ivy Bridge-E, Haswell-E will come alongside the X99 chipset. Instead of Broadwell, the back to school window of 2014 will by filled by a refresh of 22nm Haswell products with a new 9-series chipset.
Seriously, it's like watching the face of Intel's Tick-Tock while a repairman is tweaking the gears.
In terms of specifications, Haswell-E will come in 8 and 6-core offerings with up to 20MB of cache. Apart from the inclusion of DDR4 support, the main advantage of Haswell-E over the upcoming Ivy Bridge-E is supposed to be raw performance; VR-Zone estimates up to 33-50% better computational strength. A depressingly novel area of improvement as of recent...
Lastly, with recent discussion of the awkwardly hobbled K-series parts, our readers might be happy to know that all Haswell-E parts will be unlocked to overclocking. This, again, leads me to believe that Intel is not hoping to suffocate the enthusiast market but rather sort their users: mid-range consumers will take what they are given and, if they object, send them on the bus to Funk-E town.
Note, while the headlining slide definitively says "All Processors Unlocked"...
... this slide says "For K and Extreme series products." I will assume the latter is out of date?
Which begs the question: what does our readers think about that potential strategy? It could lead to mainstream performance products being pushed down into BGA-territory, but cements the existence of an enthusiast platform.
Intel Prevents Overclocking of non-K Haswell Processors, and Strips Virtualization and TSX Features From K Parts
Subject: Processors | June 13, 2013 - 09:59 AM | Tim Verry
Tagged: tsx, overclocking, Intel, i7-4770k, haswell
First revealed at IDF Beijing, Intel's latest generation 4th Generation Core "Haswell" processors enjoy a refined architecture, improved processor graphics, an integrated voltage regulator (FIVR), and for the enthusiast crowd, new methods for overclocking.
In truth, the methods for overclocking Haswell are very similar to those used to overclock Intel's Sandy Bridge and Ivy Bridge processors. However, Intel has further unlocked the new Haswell CPUs. Enthusiasts can set an overclocked Turbo clockspeed, use additional base clock (BCLK) values (100 MHz, 125 MHz, and 167 MHz), and overclock the unlocked processor graphics core clockspeed and memory clockspeed (memory in 200 MHz or 266 MHz steps). The additional BCLK values allow for easy overclocks without putting the other subsystems (such as the PCI-E bus, GPU, and memory) out of spec, which is important for the PCI-E bus which needs to be close to 100 MHz for a stable system.
The following PC Perspective articles have further information on overclocking unlocked "K" edition Haswell processors:
- Integrated Voltage Regulator and Overclocking Haswell - Ryan pushes a Core i7-4770K to 4.6GHz
- Intel Talks Haswell Overclocking at IDF Beijing - Intel outlines overclocking features of Haswell at IDF
Although Intel's overclocking reveal at IDF was fairly detailed, the company did not get into specifics on how overclocking would work on non-K chips.
On that note, the crew over at the Tech Report uncovered some rather disheartening facts such that the non-K edition Haswell processors will, essentially, be locked at stock speeds and not overclockable (they are slightly more locked down than previous generations).
While the K edition Haswell processors, such as the Core i7-4770K, will enjoy unlocked multipliers, unlocked GPU and memory clockspeeds, and additional BCLK options, the standard non-K chips (ie Core i7-4770, Core i5-4670, et al) will have locked multipliers, no Turbo Boost clockspeed overclocking, and will not be allowed to use the additional 125 MHz and 167 MHz BLCK options, which effectively makes overclocking these standard chips impossible. It may still be possible to push the BLCK up a few MHz, but without the extra
stepping and gearing ratio options, the other component clockspeeds based off that same base clock are going to go out of spec and will become unstable fairly quickly as you try to push that BLCK up.
There is one saving grace for enthusiasts considering a non-K part, however. The standard non-K CPUs will have Intel's latest TSX extensions and enterprise virtualization technologies enabled.
Although quite the head scratcher, Intel has decided to disable TSX, vPro, and VT-d on the unlocked K edition chips. The TSX extensions are not widely used yet, but will provide a noticeable performance boost to future programs that do take advantage of them by allowing developers to essentially mark off sections of code that can be run independently, and thus increase the multi-threaded-ness of the application by running as much code in parallel across multiple cores as possible. Further, the vPro and VT-d features are used by virtual machine applications (with VT-d being more relevant to the consumer side of things).
In short, Intel has continued to lock down and artificially limit its chips, as many enthusiasts suspected would happen. Standard non-K Haswell processors are more locked down than ever, and even the premium unlocked K CPUs suffer with the (odd) removal of TSX and virtualization support. As Mr. Gasior points out, enthusiasts are going to be faced with an odd choice where they can either spend extra money on a premium K part that will overclock but is limited in other ways, or go with the lower cost part that has all of the ISA extensions and virtualization support turned on... but is not overclockable.
In my opinion, locking down the standard chips is one thing-- Intel needs to incentivize enthusiasts to go with the more expensive (~$25 premium) unlocked K processors some how -- but if those same enthusiasts are spending extra money for a premium chip, they should get all the features the accompanying non-K SKU has as well as overclocking.
What do you think about the artificial limitations placed on the various Haswell SKUs?