Is Merrifield Intel's Bulldozer?

[Arachnotronic]

Supposedly Intel's next gen "Merrifield" system-on-chip is slated to launch at MWC, but doesn't anybody else find it odd that we know next to nothing about it other than the fact that it's a dual core and has ImgTec Series 6 graphics?

Seems to me that Merrifield may be DOA. Thoughts?

 

[BallaTheFeared]

You mean the next Larrabee?

Cause we had info on Bulldozer, we just had talking heads in PR telling us everything it wasn't while stating the leaks weren't true. Overhyping it into oblivion before AMD dropped the bomb.

 

[Khato]

Eh, what all do we know about the competition? NVIDIA's the only one that has tipped their hand early, no? I don't believe that either Qualcomm or Samsung have said a thing about their plans for high-end SoCs, correct? Though Samsung did confirm at CES that the Galaxy S5 will be released in the March/April time frame, so if it's using an ARM chip other than the mid-range Snapdragon 410 then there will have to be another 64 bit SoC announced sometime before then.

Which is my way of saying that Intel's simply behaving more like the competition is of late. It didn't used to be this way. Why, Samsung announced their Exynos 5250 pretty much a year before its release. Similarly, Qualcomm's Krait and its initial SoCs were announced at MWC of 2011 pretty much a year before release... But with respect to the next generation of SoCs that should be landing this year? Silence.

 

[NTMBK]

Well we also know that it's two Silvermont cores, so it should have solid CPU performance. Whether it is a success or not probably depends on the GPU side- if Intel cheaped out on GPU they way they did on Medfield and Bay Trail, it may struggle to support higher resolution displays.

Too early to say, frankly, but we'll just have to wait and see.

 

[Arkadrel]

Isnt it too early to tell?

I mean doesnt it depend alot on what the competition releases this year,
to compete with intel?

I read that Qualcomm wont have a new product until about 2H?

So Intel *could* have 1H of year to itself competeing against the older products from competitors.

 

[meloz]

But with respect to the next generation of SoCs that should be landing this year? Silence.

Mayhaps everyone has realized that they have picked the low hanging performance/watt fruits, and now future progress for SoCs will be very incremental and survivial will depend on being able to compete on cost against Mediatek.

That harsh reality will drain the enthusiasm to shout about future products.

People talk about "good enough" CPUs and how they have dulled the PC marketplace. The mobile segment is reaching the "good enough" plateau faster than anyone expected. In fact -by volume- they have already reached that state. Open any Android device and there's bound to be some mediatek or other cheap SoC in there, whereas the number of devices with cutting-edge nvidia or qualcomm SoC can almost be counted on one hand (figure of speech).

 

[jdubs03]

I don't see how two silvermont cores are even going to come close to something like the snapdragon 800. if intel is trying to make inroads into the high-end space it certainly won't be with merrifield, its too little too late for that. maybe the cores will be a silvermont+ or something, but even then it wouldn't be enough. we going to have to wait until moorefield (i think thats the platform name) in 2015 for anything substantial in that regard.

 

[StrangerGuy]

Mayhaps everyone has realized that they have picked the low hanging performance/watt fruits, and now future progress for SoCs will be very incremental and survivial will depend on being able to compete on cost against Mediatek.

That harsh reality will drain the enthusiasm to shout about future products.

People talk about "good enough" CPUs and how they have dulled the PC marketplace. The mobile segment is reaching the "good enough" plateau faster than anyone expected. In fact -by volume- they have already reached that state. Open any Android device and there's bound to be some mediatek or other cheap SoC in there, whereas the number of devices with cutting-edge nvidia or qualcomm SoC can almost be counted on one hand (figure of speech).

In China I can already find for $100 USD phones with Mediatek Octa-A7s, 1080p IPS screens, 2GB RAM. Even a now dominating Samsung would have problems fighting THAT let alone Intel. Just wait till everyone and their mom will have quad A53 paired with PowerVR Rogue or Mali 6xx/7xx this year...

The "good enough" performance is all too true. I was actually excited for the new Tegra until I realized there was nothing on Android outside of benchmarks that can even bring Adreno 320 to its knees.

 

[CHADBOGA]

Whilst both products will be less than hoped for, the comparison is a bit strange because in Merrifield, the issue will be lack of integrated comms, whereas with Bulldozer it was lack of performance that was the main problem.

 

[witeken]

I don't see how two silvermont cores are even going to come close to something like the snapdragon 800. if intel is trying to make inroads into the high-end space it certainly won't be with merrifield, its too little too late for that. maybe the cores will be a silvermont+ or something, but even then it wouldn't be enough. we going to have to wait until moorefield (i think thats the platform name) in 2015 for anything substantial in that regard.

Silvermont will have superior single threaded performance and performance/watt than S800. But I guess Merrifield won't compete with S800/805. Depending on the price, it will probably compete with Mediatek and Qualcomm quadcore A7s in the low and mid-end. Moorefield will compete with S805 and K1 in 2H14.

Broxton will succeed Moorfield in 2H15 with the 14nm Goldmont architecture and gen9 graphics.

I am not sure why Intel's smartphone roadmap is almost a year behind the tablet roadmap with Bay Trail, Cherry Trail and Willow Trail before Broxton. It would have been much better for Intel if Moorefield would come out in Q1 and a 14nm Airmont shrink in 2H14 before Broxton comes out in 2H15.

 

[FwFred]

Even if it doesn't win any major designs, Bulldozer is a failed analogy.

 

[jhu]

Intel had their Bulldozer. It was called "Pentium 4". Or maybe it was the other way around…?

 

[SlimFan]

Two Silvermont cores at over 2GHz would be a better user experience than anything else currently available from Qualcomm or Samsung or any other ARM SOC vendor. 8 Cortex A7's? When you're software uses 1 thread, and if you're lucky a second, and the other 6-7 cores are sitting there power gated, you'll really really be glad that you have an "octa" phone.

It's pretty common knowledge now that anything over two cores is very much diminishing returns, outside the ridiculous state of Android benchmarking.

Really, I see it the other way. Silvermont is the only thing that competes with Apple's A7. Apple has a dual core Cyclone-A7 64 bit chip with Series 6 graphics. That seems like exactly like what is in Merrifield. Merrifield should take the "real world" CPU performance crown, have best in class graphics, have the best battery life (perf/watt), and be 64 bit to boot. Who else has something that can compete with that? It should be an Apple A7 that's a little faster and has better battery life.

Adding two extra, slower Krait cores (or heaven forbid Cortex A7) won't lead to a better device, marketing bullet points aside. Of course, that seems to be all the OEMs care about. Personally I'd probably prefer a single core A15 to a quad core A7 because I have some basic knowledge about how software operates, but nobody would be willing to sell that in the Android ecosystem.

IW haven't heard anything about the rest of the SOC... display? imaging? etc. That might be just as likely to determine market acceptance.

 

[NTMBK]

Intel had their Bulldozer. It was called "Pentium 4". Or maybe it was the other way around…?

Northwood was a solid product. Prescott is where the wheels fell off.

 

[erunion]

Well we also know that it's two Silvermont cores, so it should have solid CPU performance. Whether it is a success or not probably depends on the GPU side- if Intel cheaped out on GPU they way they did on Medfield and Bay Trail, it may struggle to support higher resolution displays.

Too early to say, frankly, but we'll just have to wait and see.

It will have Imagination graphics, so shouldn't be a problem there.

But it won't have integrated baseband, so won't really be a direct competitor to snapdragon.

 

[Homeles]

I don't see how two silvermont cores are even going to come close to something like the snapdragon 800. if intel is trying to make inroads into the high-end space it certainly won't be with merrifield, its too little too late for that. maybe the cores will be a silvermont+ or something, but even then it wouldn't be enough. we going to have to wait until moorefield (i think thats the platform name) in 2015 for anything substantial in that regard.

Yeah, you're right. It's a bit unfair. The Snapdragons wouldn't stand a chance.

Supposedly Intel's next gen "Merrifield" system-on-chip is slated to launch at MWC, but doesn't anybody else find it odd that we know next to nothing about it other than the fact that it's a dual core and has ImgTec Series 6 graphics?

Seems to me that Merrifield may be DOA. Thoughts?

You of all people should know how quiet Intel is being in general since BK took over, not just with Merrifield.

 

[Arkaign]

Northwood was a solid product. Prescott is where the wheels fell off.

Yep, though I'd put it this way :

Willamette : Way too expensive, tied to pricey RDRAM, stupid short lived socket, and the low FSB and tiny cache made performance .. poor for the price.

Northwood : Fantastic. Great OC, and from the 1.6A to the end at 3+Ghz was equal or better than anything on the market, combined with awesome chipsets that were rock stable.

Prescott : An utterly failed promise. No real gains over Northwood with a pretty fair number of negatives.

Cedar Mill : The final P4 was actually quite good, and what Prescott was supposed to be, but by this point in time they made no sense compared to good single core AMD64 products, or even Intel's own Pentium D 805 for the OC budget nuts (X2-3800 was still $300+ for the CPU alone, so not really a value proposition at the time).

But yeah, those ~2 years of Northwood were pretty impressive. Even the first AMD64 chips split the benches (Athlon 64 3000+/3200+ vs. P4 3.2C). It just couldn't keep up when AMD ramped to 3500+, 3800+, and beyond when Prescott fell on it's face.

 

[jdubs03]

I want merrifield to be successful, but the timeframe in which its coming out is way late. i don't think single-thread performance will be meaningfully higher than bay trail, z3770 specifically. if that does play out it still will only be on par with tegra 4 and snapdragon 800 in single-thread. so either way its too little too late on that end. moorefield if it ships this year is a better shot, but yea broxton is the real foray into mobile.

from the first time i saw that digtimes report on cherry trail and willow trail i thought it was suspect. why would intel release willow trail (goldmont) only one quarter after airmont. the only way i could see that is if each of those are targeted to different price points, otherwise it just didn't make sense.

 

[Mondozei]

It's pretty common knowledge now that anything over two cores is very much diminishing returns, outside the ridiculous state of Android benchmarking.


Adding two extra, slower Krait cores (or heaven forbid Cortex A7) won't lead to a better device, marketing bullet points aside. Of course, that seems to be all the OEMs care about. Personally I'd probably prefer a single core A15 to a quad core A7 because I have some basic knowledge about how software operates, but nobody would be willing to sell that in the Android ecosystem.

Yes but this is because developers are lazy and/or under immense time pressure and stick to what works.
Properly done, quad core does have immense benefits over dual core. But until the software catches up, your point stands. But it's important to point this out as some uninformed people think that dual core is better than quad core just by somekind of natural law, which of course, it wouldn't be if the software ecosystem caught up. And at some point it will, because you can't just wing it in the mobile space like you can on a desktop where power supply generally speaking isn't a major concern.

In China I can already find for $100 USD phones with Mediatek Octa-A7s, 1080p IPS screens, 2GB RAM. Even a now dominating Samsung would have problems fighting THAT let alone Intel. Just wait till everyone and their mom will have quad A53 paired with PowerVR Rogue or Mali 6xx/7xx this year...

Mobile technology can get a long way ahead. There are still very major constraints. Once we get strong replacement and/or integration of MS Word optimized for mobile, then we will need much more raw power. As technology progresses, so does the power of the programs running on the tech. Loading times in of itself will be seen as an artifact of a past era.

The "good enough" performance is all too true. I was actually excited for the new Tegra until I realized there was nothing on Android outside of benchmarks that can even bring Adreno 320 to its knees.

Right, but that's because as a game developer you need to hit the mainstream, which means the de facto tech you're targeting is from 2011~ ish. Remember that many of us have a Western-centric perspective plus we're more interested in tech than most people. A lot of mass market smartphones sold today are phones like the Lumia 520 or their Android-equivalents(basically S225 and there about).

When the 30th procentile is at the level of S4 Pro, that's when you'll see most games doing a lot more. At the same time, some people are targeting the high-end. The premium segment is falling in terms of relative market share due to the explosion of the A-7/A-9 SoC's made by Mediatek and their equivalents for the 2nd and 3rd world, but in terms of absolute numbers it's growing very fast. A game developer today can get away targeting the mid/high echelon and going forward, this means great stuff.

Really, I see it the other way. Silvermont is the only thing that competes with Apple's A7. Apple has a dual core Cyclone-A7 64 bit chip with Series 6 graphics. That seems like exactly like what is in Merrifield. Merrifield should take the "real world" CPU performance crown, have best in class graphics, have the best battery life (perf/watt), and be 64 bit to boot. Who else has something that can compete with that? It should be an Apple A7 that's a little faster and has better battery life.

We won't know when Merrifield is out. It could be a paper launch with phones coming out in H2 this year and that means that it has to compete with Apple A8 chip.

I agree with your implicit assumption, that Apple is now the maker of the most elegant SoC solutions out there. A7 is a wonder of effiency clock-for-clock. But A8 is hardly going to get worse. I like BK's style of low-key performance at Intel. But those favourable to Intel still has to reckon with the fact that Bay Trail wasn't the smash hit a lot of people predicted. I remember the hype when the benchmarks were relased. And it got beaten by A7, even if most of the units shipping with A7 were iPhones and Baytrail is a tablet-centric SoC.

To assume that the story will be radically different this time merits strong arguments, certainly stronger than the one's I've seen so far in this thread.

 

[Mondozei]

I want merrifield to be successful, but the timeframe in which its coming out is way late. i don't think single-thread performance will be meaningfully higher than bay trail, z3770 specifically. if that does play out it still will only be on par with tegra 4 and snapdragon 800 in single-thread. so either way its too little too late on that end. moorefield if it ships this year is a better shot, but yea broxton is the real foray into mobile.

from the first time i saw that digtimes report on cherry trail and willow trail i thought it was suspect. why would intel release willow trail (goldmont) only one quarter after airmont. the only way i could see that is if each of those are targeted to different price points, otherwise it just didn't make sense.

While I suspect a paper launch at MWC, if Intel does indeed do a hard launch is very far away from "too late". As for performance, if it is merely 20% higher than Bay Trail's best, then it is already besting A7 and doing so 6 months before A8 comes out. So I don't share your pessimism at all in terms of raw performance. I question if they will launch at MWC or if it will be H2 and if it will be H2 will it be priced to premium? Because as Qualcomm likes to point out: the GPU and CPU are only 2/12 of a SoC. And what is Intel's solution on the 10/12 of the spaces? (This is even more true of Nvidia who only goes to bat on their GPU expertise and little else).

But is Merrifield "doomed"? We don't know but I'm skeptical of any prophets of doom suggesting so. I'm equally skeptical of those raving for Merrifield after falling for the same hype for Bay Trail. I think Intel's more careful now.

It wasn't too long ago when the Anandtech hive mind pronouned Intel ultimate victors simply because of their process node advantage and their technical expertise. They would crush all, these people pronounced. I never bought that. Now we see the reverse situation, whereas Intel could do no wrong before now it cannot do anything right by the same people. Maybe these people need to re-examine their prophecies and ask themselves some hard questions.

 

[jdubs03]

In terms of performance/release date, it is sub-par when compared against its competitors. Against the S800, if there is a hard launch this quarter, it still lags by half a year, same with Tegra 4 (more or less). I'm not considering the A7, as it doesn't run Android, but even then there is a considerable lag.

My point is, Intel has performance of last-generation (single-thread), and its competitors are not far away from new releases, for instance Tegra K1v1. Samsung and their Exynos 6 is supposedly not far away in an S5, so Intel will have to push a higher performing part out, unless their targeting mid-price segments.

 

[witeken]

I question if they will launch at MWC or if it will be H2 and if it will be H2 will it be priced to premium?

Intel basically has 0% market share. You can't come up with a dualcore and ask premium prices. Merrifield should launch at MWC (I think hard launch), Moorefield in Q3.

 

[witeken]

In terms of performance/release date, it is sub-par when compared against its competitors. Against the S800, if there is a hard launch this quarter, it still lags by half a year, same with Tegra 4 (more or less). I'm not considering the A7, as it doesn't run Android, but even then there is a considerable lag.

My point is, Intel has performance of last-generation (single-thread), and its competitors are not far away from new releases, for instance Tegra K1v1. Samsung and their Exynos 6 is supposedly not far away in an S5, so Intel will have to push a higher performing part out, unless their targeting mid-price segments.

I don't see how Merrifield lags by half a year or more. When I look at the benchmarks, I see next-generation (single thread) performance with 2015 performance/watt. I haven't seen any announcements for 2014 products that will threat Intel. S805 is just an S800 with higher clocks speeds that will be bottlenecked by the thermal limits of smartphones, I don't think there will be K1s in smartphones (again, because of power limits; K1 is a tablet soc), and lastly, Samsung hasn't announced an Exynos 6 soc yet, so I'm not sure if there will be one inside SGS5.

But will Merrifield actually compete with those products? It's a mid-end part. It will kill every quad or octa A7 that it competes with, both in efficiency and performance, certainly in single thread performance.

 

[Exophase]

As for performance, if it is merely 20% higher than Bay Trail's best, then it is already besting A7 and doing so 6 months before A8 comes out. So I don't share your pessimism at all in terms of raw performance.

A mere 20% higher than Bay Trail's best would mean a roughly 2.9GHz clock speed. We're talking the same uarch here. Max clock will more likely be lower than higher. This is only regarding single thread performance, since the core count will also be halved. This isn't a successor to Bay Trail, this is the same generation fit into a lower power profile.

It'll remain to be seen exactly what Intel does with the GPU. Until now they've been fairly conservative. As a total preliminary jab I'd expect them to go with a narrower design than what Apple has but with higher clock speeds. Similar to if you compare CloverTrail+ going with SGX544MP2 vs Apple's A6 going with SGX543MP3.

I definitely wouldn't relate this to Bulldozer, since Bulldozer was overhyped and underdelivered, while Merrifield looks like a pretty known quantity that will deliver exactly what's expected of it. It should be a lot better than CloverTrail+, and CloverTrail+ managed at least some wins. The CPU performance will be very competitive and I expect the perf/W to be best in class still.

What's really going to make a big difference to me is if Cortex-A15 can really have its perf/W improved as much as nVidia advertizes, while still on TSMC 28HPM. And if Samsung can pull off anything similar. 2GHz at 1W is much better than what we've seen with Samsung and nVidia's SoCs thus far and goes a long way to narrow the gap with Intel. Cortex-A57 and 20nm should help further. Will also be interested to see what process node Exynos 6 is being made on.

 

[Torn Mind]

Wilamette was Intel's Bulldozer. New architecture, some hype, yet the PIII won in some benchmarks....