Intel Broadwell Architecture Preview: A Glimpse into Core M
by Ryan Smith on August 11, 2014 12:01 PM ESTBroadwell CPU Architecture
We’ll kick off our look at Broadwell-Y with Broadwell’s CPU architecture. As this is a preview Intel isn’t telling us a great deal about the CPU at this time, but they have given us limited information about Broadwell’s architectural changes and what to expect for performance as a result.
With Broadwell Intel is at the beginning of the next cycle of their tick-tock cadence. Whereas tock products such as Haswell and Sandy Bridge designed to be the second generation of products to use a process node and as a result are focused on architectural changes, tick products such as Ivy Bridge and now Broadwell are the first generation of products on a new process node and derive much (but not all) of their advantage from manufacturing process improvements. Over the years Intel has wavered on just what a tick should contain – it’s always more than simply porting an architecture to a new process node – but at the end of the day Broadwell is clearly derived from Haswell and will be taking limited liberties in improving CPU performance as a result.
| Intel's Tick-Tock Cadence | |||||
| Microarchitecture | Process Node | Tick or Tock | Release Year | ||
| Conroe/Merom | 65nm | Tock | 2006 | ||
| Penryn | 45nm | Tick | 2007 | ||
| Nehalem | 45nm | Tock | 2008 | ||
| Westmere | 32nm | Tick | 2010 | ||
| Sandy Bridge | 32nm | Tock | 2011 | ||
| Ivy Bridge | 22nm | Tick | 2012 | ||
| Haswell | 22nm | Tock | 2013 | ||
| Broadwell | 14nm | Tick | 2014 | ||
| Skylake | 14nm | Tock | 2015 | ||
All told, Intel is shooting for a better than 5% IPC improvement over Haswell. This is similar to Ivy Bridge (4%-6%), though at this stage in the game Intel is not talking about expected clockspeeds or the resulting overall performance improvement. Intel has made it clear that they don’t regress on clockspeeds, but beyond that we’ll have to wait for further product details later this year to see how clockspeeds will compare.
To accomplish this IPC increase Intel will be relying on a number of architectural tweaks in Broadwell. Chief among these are bigger schedulers and buffers in order to better feed the CPU cores themselves. Broadwell’s out-of-order scheduling window for example is being increased to allow for more instructions to be reordered, thereby improving IPC. Meanwhile the L2 translation lookaside buffer (TLB) is being increased from 1K to 1.5K entries to reduce address translation misses.
The TLBs are also receiving some broader feature enhancements that should again improve performance. A second miss handler is being added for TLB pages, allowing Broadwell to utilize both handlers at once to walk memory pages in parallel. Otherwise the inclusion of a 1GB page mode should pay off particularly well for servers, granting Broadwell the ability to handle these very large pages on top of its existing 2MB and 4K pages.
Meanwhile, as is often the case Intel is once again iterating on their branch predictor to cut down on missed branches and unnecessary memory operations. Broadwell’s branch predictor will see its address prediction improved for both branches and returns, allowing for more accurate speculation of impending branching operations.
Of course efficiency increases can only take you so far, so along with the above changes Intel is also making some more fundamental improvements to Broadwell’s math performance. Both multiplication and division are receiving a performance boost thanks to performance improvements in their respective hardware. Floating point multiplication is seeing a sizable reduction in instruction latency from 5 cycles to 3 cycles, and meanwhile division performance is being improved by the use of an even larger Radix-1024 (10bit) divider. Even vector operations will see some improvements here, with Broadwell implementing a faster version of the vector Gather instruction.
Finally, while it’s not clear whether these will be part of AES-NI or another instruction subset entirely, Intel is once again targeting cryptography for further improvements. To that end Broadwell will bring with it improvements to multiple cryptography instructions.
Meanwhile it’s interesting to note that in keeping with Intel’s power goals for Broadwell, throughout all of this Intel put strict power efficiency requirements in place for any architecture changes. Whereas Haswell was roughly a 1:1 ratio of performance to power – a 1% increase in performance could cost no more than a 1% increase in power consumption – Broadwell’s architecture improvements were required to be at 2:1. While a 2:1 mandate is not new – Intel had one in place for Nehalem too – at the point even on the best of days meaningful IPC improvements are hard to come by at 1:1, never mind 2:1. The end result no doubt limited what performance optimizations Intel could integrate into Broadwell’s design, but it also functionally reduces power requirements for any given performance level, furthering Intel’s goals in getting Core performance in a mobile device. In the case of Broadwell this means Broadwell’s roughly 5% performance improvement comes at a cost of just a 2.5% increase in immediate power consumption.
With that said, Intel has also continued to make further power optimizations to the entire Broadwell architecture, many of which will be applicable not just to Core M but to all future Broadwell products. Broadwell will see further power gating improvements to better shut off parts of the CPU that are not in use, and more generalized design optimizations have been made to reduce power consumption of various blocks as is appropriate. These optimizations coupled with power efficiency gains from the 14nm process are a big part of the driving force in improving Intel’s power efficiency for Core M.
Facebook
Twitter
RSS
158 Comments
View All Comments
mapesdhs - Monday, August 11, 2014 - link
Yeah, sure, and that's exactly what everyone was saying back when we were waiting for
the followon to Phenom II; just wait, their next chip will be great! Intel killer! Hmph. I recall
even many diehard AMD fans were pretty angry when BD finally came out.
Benchmarks show again and again that AMD's CPUs hold back performance in numerous
scenarios. I'd rather get a used 2700K than an 8350; leaves the latter in the dust for all CPU
tasks and far better for gaming.
Btw, you've answered your own point: if an 8350 is overkill for a game, giving 120fps, then
surely one would be better off with an i3, G3258 or somesuch, more than enough for most
gaming if the game is such that one's GPU setup & sscreen res, etc. is giving that sort of
frame rate, in which case power consumption is less, etc.
I really hope AMD can get back in the game, but I don't see it happening any time soon.
They don't have the design talent or the resources to come up with something genuinely
new and better.
Ian.
wurizen - Monday, August 11, 2014 - link
an fx-8350 isn't holding anything back. come on, man. are you like a stat paper queen obsessor or something? oh, please. an fx-8350 and an amd r9290 gpu will give you "happy" frame rates. i say happy because it i know the frames rates will be high enough. more than good enough even. will it be lower than an i7-4770k and an r90? maybe. maybe the fx-8350 will avg 85 fps on so and so game while the i7-4770k will avg 90 fps. boohoo. who cares about 5 more frames.also, while you mention i3 as a sufficient viable alternative to an fx-8350. remember that the cost will probably be about the same. and fx-8350 is like $190. maybe the i3 is 20 dollars less. but, here's the big but, an i3 is not as good as an fx-8350 in video editing stuff and photo editing stuff if one would like to use their pc for more than just games. an fx-8350, while not as power efficient as an i3 (but who cares since we are talking about a desktop) literally has more bang for the back. it has more cores and is faster.
amd will get back in the game. it is just a question of when. an fx-8350 is already toe-to-toe with an i7-2600k, which is no slouch in todays standard. so, amd just needs to refine their cpu's.
as for talent? amd came up with x64, or amd64 before intel. intel developed their own x86-64 later.
the resource that intel has over amd is just die shrinking. that's it. architecturally, an fx chip or the phenom chip before it seems like a more elegant design to me than intel chips. but that's subjective. and i don't really know that much about cpu's. but, i have been around since the days of 286 so maybe i just see intel as those guys who made 286 which were ubiquitous and plain. i also remember cyrix. and i remember g4 chips. and to me, the fx chip is like a great chip. it's full of compromises and promises at the same time.
Drumsticks - Monday, August 11, 2014 - link
I think AMD might have a way back into the game, but the difference right now is way worse than you say.http://www.anandtech.com/bench/product/697?vs=287
FX-8350 trails the 2600k frequently by 10-20% or more (in gaming).
http://www.anandtech.com/bench/product/697?vs=288
i5-2500k beats it just as badly and actually sells for less than the 8350 used on ebay. Games love single threaded power and the 8350 just doesn't have it.
wurizen - Monday, August 11, 2014 - link
the games they have in that comparison are starcraft 2 and dragon age. 47 fps at 768 p for 8350 looks suspect on starcraft 2. what gpu did they use?it's not way worse as i say. omg.
i have an i7-3770k oc'd to 4.1Ghz and a stock FX-8320 at stock. both can run cod: ghost and bf3. haven't tested my other games. might do the starcraft 2 test tomorrow. i don't have the numbers nor care. what ppl need to realize is the actual game experience while playing games and not the number. is the game smooth? a cpu that can't handle a game will be very evident. this means it's time to upgrade. and there are no fx cpu's from amd that can't handle modern games. again, they will trail intel, but that is like a car going at 220mph so that car wins but the other car is going at 190mph and it will lose but realistically and the experience of going at 190mph will still be fast. the good thing is that amd or cpu don't race each other unless you care about benchmarks. but, if you look past the benchmarks and just focus on the experience itself, an fx series cpu by amd is plenty fast enuff.
omg.
silverblue - Tuesday, August 12, 2014 - link
We're well within the realms of diminishing returns as regards standard CPU IPC. AMD has the most to gain here, though with HSA, will they bother?kaix2 - Tuesday, August 12, 2014 - link
so your response to people who are disappointed that broadwell is focused more on TDP instead of performance is to buy an AMD cpu with even lower performance?wurizen - Tuesday, August 12, 2014 - link
well, they don't have to get the fx-9590, which has serverlike cpu of 2008 like tdp or a gpu like tdp of 220 watts. there is a more modest tdp of 125w with the fx8350. all overclockable. seems like a good cpu for tinkerers, pc/enthusiast, gamers and video editors. i don't even think it's a budget cpu. there is the 6-core and 4-core variants which are cheaper. i am also not saying that an fx-8350 is like the best cpu since it's not and falls way down in the benchmark charts. but, it's not a bad cpu at all. it gets the work done (video editing) and let's you play games (wit's a modern cpu after) even though it's sort of 2 yrs old already. the 990FX chipset is even an older chipset. there's something to be said about that and i think im trying to say it. in light of all the news about intel, which we are guaranteed to get every year with each tick and tock... there is that little AMD sitting in the corner with a chipset that hasn't been updated for yrs and an 8-core cpu that's remarkably affordable. the performance is not that low at all. i mean, video editing with it or playing games with it doesn't hamper one's experience. so, maybe one will have to wait a couple more minutes for a video to render in a video editing program versus say an i7-4790k. but, one can simply get up from one's chair and return. instead of staring at how fast their cpu renders a video on the screen.know what i'm saying?
so, yeah. an fx-8350 with an old 990fx mobo and now intel's upcoming broadwell cpu's with z97 chipsets and all the bells and whistles and productivity for either one will probably be similar. also, most video editing programs now will also leverage the gpu so an old fx-8350 w/ a compatible gpu will have help rendering those gpu's....
i guess it's like new doesn't mean anything now. or something. like m2 sata and pci 3.0, which intel chipsets have over amd is kinda superflous and doesn't really help or do much.
know what im saying?
rkrb79 - Tuesday, November 18, 2014 - link
Agreed!!name99 - Tuesday, August 12, 2014 - link
Oh yes, Skylake.Intel has given 5% IPC improvements for every generation since Nehalem, but now Skylake is going to change everything?
If you're one of the ten people on the planet who can actually get value out of AVX-512 then, sure, great leap forward. For everyone else, if you were pissed off at IB, HSW, BDW, you're going to be just as pissed off with Skylake.
DanNeely - Tuesday, August 12, 2014 - link
No, the interest in Skylake is for all the non-CPU speed things promised with it. PCIe 4.0 and a bump from 16 to 20 CPU lanes (for PCIe storage) are at the top of the list. Other expected, but AFAIK not confirmed, benefits include USB3.1 and more USB3.x on the chipset than the current generation. We should have consumer DDR4 with Skylake too; but that's not expected to be a big bump in the real world.