AMD Zen Rumours Point to Earlier Than Expected Release

considering the IPC right now is only a little better than something that was produced 5 years ago.
Click to expand...

I say its even longer than that if you compare it to Intel however bulldozer was designed to have a lower IPC with a higher clock. Zen will not have this lower IPC design feature..
 
Got the new Wraith cooler in, so I will do a mini-review on it, change out some fans etc. just to see what is shaking with that. Actually I just ordered a new Koolance water block in to test with. I have not upgraded the AMD side of my cooler on the test bench in 4 or 5 years? Wow.

cpu-380a_p3-700x700.jpg
 
Last edited:
NaroonGTX said:
Even if AMD does meet the 40% or greater IPC jump, there's still gonna be people bitching about it for contrived reasons as usual.
Click to expand...

The issue here is that while AMD is increasing IPC (note: It's unclear how their SMT implementation factors into this claim), it's almost certain clocks will end up being lower. Adding 40% to IPC is nice, but if you have to reduce clocks 15% in the process, your actual single-core performance gain reduces down to 25%, which is nice, but still only IB level performance. Even AMD supporters will have issues purchasing an IB level processor at $400.
 
Even AMD supporters will have issues purchasing an IB level processor at $400.
Click to expand...

I expect the 4 core / 8 threaded processor to be priced to compete with the i5s and the 6 core / 12 threaded processor to be priced to compete with i7s. The 8 core / 16 threaded processor (hopefully) will be a > $400 part (I want AMD to compete with Intels enthusiast platform) but will not be the CPU for most users.
 
Last edited:
I believe they will do like years previous where the price is related to comparable Intel parts.
 
JustReason said:
I believe they will do like years previous where the price is related to comparable Intel parts.
Click to expand...

I would expect this policy to remain in place. AMD may get the price wrong at the start however they will adjust it to make the price a discount to what processor they believe they are competing with.
 
gamerk2 said:
The issue here is that while AMD is increasing IPC (note: It's unclear how their SMT implementation factors into this claim), it's almost certain clocks will end up being lower. Adding 40% to IPC is nice, but if you have to reduce clocks 15% in the process, your actual single-core performance gain reduces down to 25%, which is nice, but still only IB level performance. Even AMD supporters will have issues purchasing an IB level processor at $400.
Click to expand...

I'm not sure why people are thinking that Summit Ridge will ship out the box with 4.0ghz and higher speeds. I don't expect that at all, not for the octocore SKU's anyway, especially given the 95W TDP the processors are supposedly targeting. I think they're doing exactly what they said they are, focusing on per-core performance and energy efficiency.

I don't think the SMT factors into the IPC at all because AMD said the 40% figure was independent of process node and was per-core, and SMT is more about multi-core/multi-thread than anything else, i.e. irrelevant for actual single-threaded workloads. Besides that, it's 40% over Excavator which itself was a good leap over Steamroller, so it'll still be significantly faster than AMD's current "high-end" AM3+ lineup which is already four years old.

How it will stack up to Intel's current stuff, who knows as usual. But the clocks will definitely be lower and from what I've heard, assuming the yields are good and whatnot, you might be able to see 4.0~4.5ghz ranges with overclocks on Summit Ridge. I don't think people will be clocking them to 5ghz.
 
If it's bulldozer +40% IPC it will be really amazing.

If it's bulldzoer +40% IPC -10% clock speed and half the cores it will be bulldozer all over again.
 
-Dragon- said:
If it's bulldozer +40% IPC it will be really amazing.

If it's bulldzoer +40% IPC -10% clock speed and half the cores it will be bulldozer all over again.
Click to expand...

That's not how it works. Besides, you are completely forgetting the fact that Zen will have SMT to help with having less cores.
 
Just like Intel having hyperthreading.

IPC is only part of the puzzle, its how that increased IPC can be utilized to its full potential pretty much everything else in the chip that would help with throughput.
 
CaptNumbNutz said:
That's not how it works. Besides, you are completely forgetting the fact that Zen will have SMT to help with having less cores.
Click to expand...

AMD has been mum on the SMT implementation. This matters, especially considering how CMT ended up working.

Secondly, adding more cores is just a multiplayer to overall CPU performance. Take the performance of a single core [including any loss/gain from SMT cores], multiply by the number of cores, and you roughly get a CPUs maximum performance. Single-core performance is dominated by two factors: IPC and Clockspeed. Increasing one while decreasing the other results in lower then anticipated performance.

This is VERY important to note, given Globalfoundries 14LPP node is tuned for sub-3GHz performance, given it's a mobile optimized node. Maybe its no surprise, but I've been hearing a LOT of rumors about Zen being clocked in the 2.8GHz range.

In any case, it's a near certainty clock speeds will be reduced from BD/PD, so the actual gain in Single-core performance will be less then 40%. And if AMD is continuing it's tradition of given "best case" performance numbers, I wouldn't be terribly shocked if actual per-core improvement ends up in the 15-20% range, rather then the 40% a lot of people are hoping for.
 
-Dragon- said:
If it's bulldozer +40% IPC it will be really amazing.

If it's bulldzoer +40% IPC -10% clock speed and half the cores it will be bulldozer all over again.
Click to expand...

I'll be surprised if the numbers are closed to that outside of one or two benchmarks. Historically AMD has presented performance numbers of the best case scenarios rather than averages or based off any real world testing. I also seriously doubt you'll be seeing clock speeds within 10% of Bulldozer's. We'll see, but I'm not overly optimistic.

Voxata said:
I really hope the above post doesn't end up being true.
Click to expand...

I think it's probably pretty close to the mark. Unfortunately.
 
To be honest supposedly it should be Samsung process that GloFo is using, because GloFo was getting nowhere on their own.

The design for Zen supposedly is from scratch.
 
The design isn't from scratch its based off of other IP's they have had in the past, if AMD didn't diverge from their Athlon 2 design, they would have been in a better position than going to the opposite direction with BD.

Just like Intel went back to its Pentium III architecture to create their core line.
 
As long as they are dirt cheap and can push around 60+ FPS with a decent GPU...it'll do.
 
razor1 said:
The design isn't from scratch its based off of other IP's they have had in the past, if AMD didn't diverge from their Athlon 2 design, they would have been in a better position than going to the opposite direction with BD.

Just like Intel went back to its Pentium III architecture to create their core line.
Click to expand...

The problem is Intel does a LOT under the hood to improve performance that AMD simply can't. Intel spent a lot of time and money to optimize their branch algorithms for example, which AMD simply can't replicate to the same degree. AMD simply can't keep pace on IPC, which is why they went in a different direction. The problem is, for reasons now obvious, they ran into the same problems at > 4GHz clocks that Intel ran into with the Pentium 4.

Voxata said:
I really hope the above post doesn't end up being true.
Click to expand...

Over at Toms Hardware, I'm still milking the fact I was one of the handful that correctly estimated BD's performance at launch.

I expect higher tier chips to reach the 3.4 GHz range, but the days of 4GHz clocks are gone. But no matter how you slice it, that's still a good 20% clock speed reduction. So if IPC goes up by 40%, but clocks reduce 20%, guess what? You're single core performance is up 20%. And if those IPC numbers are "best case", well, suddenly Zen doesn't look so hot.

Which maybe explains why AMD is staying mum on clocks, which should be set in stone at this point.
 
gamerk2 said:
The problem is Intel does a LOT under the hood to improve performance that AMD simply can't. Intel spent a lot of time and money to optimize their branch algorithms for example, which AMD simply can't replicate to the same degree. AMD simply can't keep pace on IPC, which is why they went in a different direction. The problem is, for reasons now obvious, they ran into the same problems at > 4GHz clocks that Intel ran into with the Pentium 4.



Over at Toms Hardware, I'm still milking the fact I was one of the handful that correctly estimated BD's performance at launch.

I expect higher tier chips to reach the 3.4 GHz range, but the days of 4GHz clocks are gone. But no matter how you slice it, that's still a good 20% clock speed reduction. So if IPC goes up by 40%, but clocks reduce 20%, guess what? You're single core performance is up 20%. And if those IPC numbers are "best case", well, suddenly Zen doesn't look so hot.

Which maybe explains why AMD is staying mum on clocks, which should be set in stone at this point.
Click to expand...

Please show me evidence of a 40% IPC increase only at the same clock speeds. Also, can you show me evidence of the newest architecture being unable to hit a 4GHz stock speed? I honestly think we are doing nothing but conjecture here with zero proof.

Also, I could have said that the Bulldozer architecture would have less IPC and had been correct. (They did say that before anyways.) However, that does not make me a great Nostradamus. They have been mum on pretty much the whole kit and caboodle so we really do not have any idea what to truly expect.

I keep seeing these numbers being thrown around and finally just got tired of the incorrect conjecture forming. Yes, I am an AMD fan but that does not make me uninformed or have blinders on. (You may not have blinders on either but I still think you are reaching when it comes to what you are saying.)
 
I think the idea that the core count will stay the same too is exceedingly optimistic. People talk about about a high end 8C/16T part in the $400 range but I'd put money on most parts being 4C/8T just like intel with maybe a high end 6C/12T part depending on thermals. Honestly perf/W is really what AMD needs to compete since the mobile world is booming and a slow CPU with high power has a snowball effect on the quality of the rest of the system regardless of how cheap it is.

So I stand by my statement.

BD + 40% IPC + 20% SMT = super neat
BD / 2 (cores cut in half) + 40% IPC + 20% SMT - 10% (or more) clock speed = zendozer
 
ManofGod said:
Please show me evidence of a 40% IPC increase only at the same clock speeds. Also, can you show me evidence of the newest architecture being unable to hit a 4GHz stock speed? I honestly think we are doing nothing but conjecture here with zero proof.
Click to expand...

While we have no actual proof that this will happen we do have some historical context from which to make an educated guess from.

Remember the Pentium 4? That deeply pipelined pile of shit? How we had 3.8GHz processors back in 2004 and when the much higher IPC Core CPUs dropped high end clock speed dropped down to the 2.6GHz range? Hell intel stock baseline CPU speed only recently caught back up with what those old P4's did as far as pure clocks.

So it's reasonable to expect that when AMD switches from their own deeply pipelined pile of shit version of the P4 that they won't be able to maintain the same clock speed as they did before either. I don't think BD is as deeply pipelined as the P4 was and the IPC improvement probably isn't as big going from BD to zen so I don't think it's reasonable to expect the >30% clock drop that intel had, but if you're expecting a 4GHz zen out of the gate you haven't been paying attention to history.
 
gamerk2 said:
The problem is Intel does a LOT under the hood to improve performance that AMD simply can't. Intel spent a lot of time and money to optimize their branch algorithms for example, which AMD simply can't replicate to the same degree. AMD simply can't keep pace on IPC, which is why they went in a different direction. The problem is, for reasons now obvious, they ran into the same problems at > 4GHz clocks that Intel ran into with the Pentium 4.
Click to expand...

I agree, I should have stated, better position as in relative to what they are now but still would have been behind Intel, it will be difficult for them, but if Zen turns out to be decent enough to start gaining marketshare back and start balancing out their books, future iterations of Zen should get better as they will be able to "afford" more resources.
 
-Dragon- said:
While we have no actual proof that this will happen we do have some historical context from which to make an educated guess from.

Remember the Pentium 4? That deeply pipelined pile of shit? How we had 3.8GHz processors back in 2004 and when the much higher IPC Core CPUs dropped high end clock speed dropped down to the 2.6GHz range? Hell intel stock baseline CPU speed only recently caught back up with what those old P4's did as far as pure clocks.

So it's reasonable to expect that when AMD switches from their own deeply pipelined pile of shit version of the P4 that they won't be able to maintain the same clock speed as they did before either. I don't think BD is as deeply pipelined as the P4 was and the IPC improvement probably isn't as big going from BD to zen so I don't think it's reasonable to expect the >30% clock drop that intel had, but if you're expecting a 4GHz zen out of the gate you haven't been paying attention to history.
Click to expand...

Except that the history your referring to is nearly 12 years old. (Might as well be 100 years in computer terms.) With the node shrink from 32nm to 14nm and what is probably an entirely new architecture, 4GHz will be entirely in the realm of possibility. Also, nothing we have seen so far indicates a 40% ipc increase at the same clock speeds.

Regardless, I have to see evidence and not just simple conjecture from 12 year old history. On a different note, if AMD can get Zen in OEM systems in any significant quantity, that will be a big boost for them. Oh, and you keep claiming that essentially, Piledriver IPC is shit yet, at 4k resolutions, the games do not play really any better on my new I7 6700 k system at stock than on the FX 8350 system I had at stock as well. I just wanted to upgrade and not invest in older tech. Chances are, if I had realized the results early on, I probably would have waited and just did the 980 Ti only.

Oh well, I am still pleased with my upgraded system.
 
-Dragon- said:
While we have no actual proof that this will happen we do have some historical context from which to make an educated guess from.

Remember the Pentium 4? That deeply pipelined pile of shit? How we had 3.8GHz processors back in 2004 and when the much higher IPC Core CPUs dropped high end clock speed dropped down to the 2.6GHz range? Hell intel stock baseline CPU speed only recently caught back up with what those old P4's did as far as pure clocks.

So it's reasonable to expect that when AMD switches from their own deeply pipelined pile of shit version of the P4 that they won't be able to maintain the same clock speed as they did before either. I don't think BD is as deeply pipelined as the P4 was and the IPC improvement probably isn't as big going from BD to zen so I don't think it's reasonable to expect the >30% clock drop that intel had, but if you're expecting a 4GHz zen out of the gate you haven't been paying attention to history.
Click to expand...

It was only the Prescott that had the super deep pipe right?
 
gamerk2 said:
I expect higher tier chips to reach the 3.4 GHz range, but the days of 4GHz clocks are gone. But no matter how you slice it, that's still a good 20% clock speed reduction. So if IPC goes up by 40%, but clocks reduce 20%, guess what? You're single core performance is up 20%. And if those IPC numbers are "best case", well, suddenly Zen doesn't look so hot.
Click to expand...

The problem is due to math the loss of clock speed actually negates the IPC increase more than simple addition like that might show.

So 4GHz to 3.6 = 85% of the clock speed a loss of 15%, 40% more IPC = 1.4 times the work per clock, so 0.85 x 1.4 is actually 1.19, so in your example zen only gains 19% over BD.

alxlwson said:
It was only the Prescott that had the super deep pipe right?
Click to expand...

All P4's were NetBurst which was the super deep pipeline
 
ManofGod said:
Except that the history your referring to is nearly 12 years old. (Might as well be 100 years in computer terms.) With the node shrink from 32nm to 14nm and what is probably an entirely new architecture, 4GHz will be entirely in the realm of possibility. Also, nothing we have seen so far indicates a 40% ipc increase at the same clock speeds.
Click to expand...
So the fact that the BD architecture scales to high clocks at 32nm just like the NetBurst architecture did at 90nm which is why until BD that NB based chips held world records for clock speeds, it's pure conjecture to think that we might see a similar loss of high clock ability going the other direction just like what happened to intel. Got it. I'm sue AMD has much better chip designers who can work closely with their in house fab to ensure maximum performance. Oh... wait...

Seriously if you're not expecting some drop in clocks you're expecting way too much. I'll be happy if I'm wrong, but lets face it, between the historical similarities to BD and NB and AMDs track record with everything, my 10% is more likely to turn out to be an underestimation than overestimation.

ManofGod said:
Regardless, I have to see evidence and not just simple conjecture from 12 year old history. On a different note, if AMD can get Zen in OEM systems in any significant quantity, that will be a big boost for them. Oh, and you keep claiming that essentially, Piledriver IPC is shit yet, at 4k resolutions, the games do not play really any better on my new I7 6700 k system at stock than on the FX 8350 system I had at stock as well. I just wanted to upgrade and not invest in older tech. Chances are, if I had realized the results early on, I probably would have waited and just did the 980 Ti only.

Oh well, I am still pleased with my upgraded system.
Click to expand...
Congrats on being GPU limited.
 
I expect stock clocks to be similar to Intel's current stock clocks. The 4 core / 8 threaded processor should have a ~4GHz or so stock clock. The 6 core / 12 threaded processor should have a ~3.5 GHz clock. The 8 core / 16 threaded processor should have a ~3 GHz stock clock if it is going to be a 95W TDP processor. No guesses on turbo or overclocking ability.
 
-Dragon- said:
Congrats on being GPU limited.
Click to expand...

So, what you are saying is the the 6700K system was a really good investment? :rolleyes: Look, I am happy with my system but, I am not a fool to think that this was a good investment. Part of me wishes I would have waited until a cpu that can make a difference at 4k hit the streets. (The 8350 is quite fast considering it is what I am using at work right now.)

Better question would be, why would anyone bother with 1080p anymore except on lower end systems?
 
I think your biggest upgrade on that system came from the platform more than the CPU such as PCIe 3.0, native USB3, NVMe, etc... I only upgraded from a 2600k to haswell because my old motherboard shit itself, and I'm pretty sure my next system upgrade will probably be due to component failure as well since the high end PC market is fairly stagnant due to all development and silicon improvements being focused on the mobile markets. Which is also why I think the perf/W metric of zen will be way more important than any IPC or clock speed changes in the long run.

Regardless AMD has a long and sordid history of touting some aspect of an upcoming product that, in a vacuum, would lead to amazing performance increases all other things being equal, such as BD's "moar coars" and Zen's "moar eye pee cees" yet once the products are actually released and the media NDA lifts a few days after (I am remembering that right aren't I? When BD came out I seem to recall places selling them for days before NDAs lifted) we learn all things aren't equal and there's some other factor they neglected to mention that nullifies some to most of the trumpeted aspect that improved, such as wretched IPC in BD. A clock speed hit is a fairly reasonable guess here especially since clock speeds haven't really gone up since the 32nm days.
 
-Dragon- said:
I think your biggest upgrade on that system came from the platform more than the CPU such as PCIe 3.0, native USB3, NVMe, etc... I only upgraded from a 2600k to haswell because my old motherboard shit itself, and I'm pretty sure my next system upgrade will probably be due to component failure as well since the high end PC market is fairly stagnant due to all development and silicon improvements being focused on the mobile markets. Which is also why I think the perf/W metric of zen will be way more important than any IPC or clock speed changes in the long run.

Regardless AMD has a long and sordid history of touting some aspect of an upcoming product that, in a vacuum, would lead to amazing performance increases all other things being equal, such as BD's "moar coars" and Zen's "moar eye pee cees" yet once the products are actually released and the media NDA lifts a few days after (I am remembering that right aren't I? When BD came out I seem to recall places selling them for days before NDAs lifted) we learn all things aren't equal and there's some other factor they neglected to mention that nullifies some to most of the trumpeted aspect that improved, such as wretched IPC in BD. A clock speed hit is a fairly reasonable guess here especially since clock speeds haven't really gone up since the 32nm days.
Click to expand...

Unfortunately, PCIe 3 does not make any real difference at the moment, native USB 3 does not seem relevant and I will not be using any NVMe drives for a long while. I think I am starting to realize my biggest reason for an upgrade last September was because I could. :D Oh, and I do not have any USB 3.1 devices and probably will not for a long time.

Once again, not complaining, just kind of wishing I had waited a little longer. (Oh well, at least my benchmarks look good. :D)
 
That is true, the platform is more important, since CPU performance hasn't been a driving factor as it did in the past. But this is also due to the fact software makers still have to factor in the lowest common denominator when programming. I think things would have been different if Intel went with more cores in their consumer line up, but at the end of the day Intel wasn't pushed to do it.
 
drescherjm said:
I expect stock clocks to be similar to Intel's current stock clocks. The 4 core / 8 threaded processor should have a ~4GHz or so stock clock. The 6 core / 12 threaded processor should have a ~3.5 GHz clock. The 8 core / 16 threaded processor should have a ~3 GHz stock clock if it is going to be a 95W TDP processor. No guesses on turbo or overclocking ability.
Click to expand...

If AMD were using Intel's process node, sure. But Glofo's node is explicitly designed to be optimal at lower speeds, which is why I think the arch will initially top out at 3.5GHz or so. I'm sure we'll eventually get a 4GHz stock part, but almost certainly not at launch.

In regards to IPC, that claim comes straight from AMD. Though I continue to note AMD has a history of "best case" marketing numbers, and they never stated if they are counting that added performance from their SMT design to that number or not. So even AMDs claim could be on the high side to start with, even before factoring in clock speed reductions.

I'm expecting IB/Haswell performance for the CPU as a whole. I'm expecting pretty much the same dynamic we've had: Top AMD chips compete against top i5s and budget i7s.
 
ManofGod said:
Also, I could have said that the Bulldozer architecture would have less IPC and had been correct. (They did say that before anyways.) However, that does not make me a great Nostradamus. They have been mum on pretty much the whole kit and caboodle so we really do not have any idea what to truly expect.
Click to expand...

I actually nailed most of BD's performance dynamics: I predicted in single threaded workloads, it would likely be slower the Phenom II (It was, and NO ONE else predicted that outcome). I predicted the CMT implementation would cause performance loss that would choke performance in single-threaded tasks (It did; MSFT had to patch Windows). I predicted to some extent the poor cache latencies (though even I dismissed the numbers that leaked about a week before launch; even i didn't think they would be THAT bad).

I didn't just predict BD would be worse then SB, I predicted by how much and why. I also argued (and continue to argue) that game's wouldn't become significantly more threaded, and that the ones that did still wouldn't benefit AMD in any way, so the argument that you should purchase BD because it would outperform Intel in the future was just silly and wrong (true).

So yeah, I've had a good recent history predicting these things.
 
If AMD were using Intel's process node, sure. But Glofo's node is explicitly designed to be optimal at lower speeds, which is why I think the arch will initially top out at 3.5GHz or so. I'm sure we'll eventually get a 4GHz stock part, but almost certainly not at launch.
Click to expand...

I figured a year later and with the lower IPC would amount to being able to match Intel's frequency with a 14 or 16nm process that is not as good as Intel's. In the case of the 8 core / 16 threaded I was matching Intel's 22nm process but instead of a 140W part a 95W part. I do admit 3GHz may be extremely difficult to achieve if the 95W is a true 95W limit.
 
Last edited:
razor1 said:
That is true, the platform is more important, since CPU performance hasn't been a driving factor as it did in the past. But this is also due to the fact software makers still have to factor in the lowest common denominator when programming. I think things would have been different if Intel went with more cores in their consumer line up, but at the end of the day Intel wasn't pushed to do it.
Click to expand...

CPUs dont matter as much in games because GPUs are the bottleneck. Pretty much that simple. You don't typically see a CPU side bottleneck until you get into SLI'd configurations.

As for threading: More threads != more performance. I've tried to explain this many, many times over the years. At the end of the day, most of the game logic is serial, and if 80% of your workload is serial, the best performance improvement you can get by adding more threads is 20%
 
gamerk2 said:
CPUs dont matter as much in games because GPUs are the bottleneck. Pretty much that simple. You don't typically see a CPU side bottleneck until you get into SLI'd configurations.
Click to expand...

You also typically have to run games in extremely high resolutions for the CPU differences to really shine even in SLI / Crossfire. 1920x1080 isn't going to make such a system work that hard in any game.

gamerk2 said:
As for threading: More threads != more performance. I've tried to explain this many, many times over the years. At the end of the day, most of the game logic is serial, and if 80% of your workload is serial, the best performance improvement you can get by adding more threads is 20%
Click to expand...

This is why we never saw 4x the performance in games going from supporting 1 core to 2, or 2x the performance going from 2 to 4. The workloads just aren't linear with respect to adding more cores.
 
We are in an era where i3s can do 60 fps in most titles, cores will never be as important as IPC in a gaming scenario. But 4 cores seems to be the sweet spot and if there is a product with a couple extra cores with 10% lower IPC competing against i5 i am sure a lot of people will love it (unlike 2x - 4x cores for sacrifice of ~50 - 60% IPC which is retarded)
 
-Dragon- said:
I think the idea that the core count will stay the same too is exceedingly optimistic. People talk about about a high end 8C/16T part in the $400 range but I'd put money on most parts being 4C/8T just like intel with maybe a high end 6C/12T part depending on thermals. Honestly perf/W is really what AMD needs to compete since the mobile world is booming and a slow CPU with high power has a snowball effect on the quality of the rest of the system regardless of how cheap it is.

So I stand by my statement.

BD + 40% IPC + 20% SMT = super neat
BD / 2 (cores cut in half) + 40% IPC + 20% SMT - 10% (or more) clock speed = zendozer
Click to expand...

From 32nm to 28nm and now the step to 14nm for AMD it is a big leap. Mobile business on x86 is all Intel AMD can't make any money there especially since ARM rules there the market percentage gained from Intel would be so small that it would not even warrant spending any money in that segment. Another sign is that last few chips that were "mobile" were used in low cost laptops rather then tablets (certainly not phones AMD does not want to go there).

If AMD can't make a 8 core cpu work and just goes for 4 cores them I'm not to sure if that whole AM4 platform would be worth it, however we have seen that AMD can produce limited TDP cpu and that is something closer to the solution that AMD might take.
 
Pieter3dnow said:
https://www.youtube.com/watch?v=SOTFE7sJY-Q

no
Click to expand...


As much as they say its new, its never new, just like anything in the manufacturing of silicon products they are built on previous architectures. No need to redo the entire wheel when you already have most of the parts that are already good.

Intel said they core architecture was new too, when you look at everything as a whole its new, but not its individual components.

check 2:50 to 6:00 minutes, they are basing it off of older tech they have and arm tech they have worked with. At 6:30 Keller talks about a clean design plan then goes into they already have good parts they can do uses to start off fast. Listen to the whole thing, they aren't starting form scratch for all the parts.
 
razor1 said:
As much as they say its new, its never new, just like anything in the manufacturing of silicon products they are built on previous architectures. No need to redo the entire wheel when you already have most of the parts that are already good.

Intel said they core architecture was new too, when you look at everything as a whole its new, but not its individual components.

check 2:50 to 6:00 minutes, they are basing it off of older tech they have and arm tech they have worked with. At 6:30 Keller talks about a clean design plan then goes into they already have good parts they can do uses to start off fast. Listen to the whole thing, they aren't starting form scratch for all the parts.
Click to expand...

Yeah and those points seem intentionally vague where Jim Keller says using the DNA of Jaguar and Bulldozer. And the cross talk about K12 and Zen also makes it harder to come to certain conclusions.

But he starts with saying about making a big leap , you can not make a big leap if you copy and paste things and just give it a new name .....
 
You must log in or register to reply here.
Back
Top