K7/K8 Inventor Back at AMD

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
While this certainly is good news for AMD, should the folks at Intel be worried?

Now, though, Jim Keller, the one who spearheaded the invention of the K7 and K8 architectures that were better than anything else in their time (though we're sure some disagree with that) is back at Advanced Micro Devices. He will work on the K12 ARMv8-based products, but will take a look at Bulldozer as well and, maybe, finally allow it to get over the main problems.
 
In fact, 16-core units could surface as early as this year (2014), once the Steamroller generation of Bulldozer comes forth. Sadly, software still isn't coded for so many different threads, so most of the potential computing power will go to waste, at least at first.

Ha, Most of the processing power in 4 core machines is not be utilized.

16 cores sounds like a dream come true for rendering HD video in H.264 and higher.
Video codec scale nicely across multiple cores.
 
I thought I travelled back in time. I am SURE this was news like 2 months ago. Now, where did I put my hoverboard...

16 cores sounds nice, unless you're talking about games. More cores, in the current paradigm, means more latency. No way around that.
 
Ha, Most of the processing power in 4 core machines is not be utilized.

16 cores sounds like a dream come true for rendering HD video in H.264 and higher.
Video codec scale nicely across multiple cores.

Does it scale that well though? I mean AMD still lags in IPC/core.
 
In fact, 16-core units could surface as early as this year (2014), once the Steamroller generation of Bulldozer comes forth. Sadly, software still isn't coded for so many different threads, so most of the potential computing power will go to waste, at least at first.

"At least at first". Still waiting, how many years later, and even 4 cores are barely used. 7 years since the Q6600...
 
This news means nothing. Nothing will happen, nothing new will come of it. Multi-Core as of now is merely, for 99% of users, a marketing gimmick.
 
Him working on an ARM processor is fairly good news for them.

The competition (outside of Apple) seems very all over the place.
 
Does it scale that well though? I mean AMD still lags in IPC/core.

Well, in a non-scientific, seat of the pants test; when I switch from a 4 core to a 8 core machine, my FPS performance doubled using HandBrake. Take that for what it's worth.
;)
 
16 cores sounds nice, unless you're talking about games. More cores, in the current paradigm, means more latency. No way around that.

More cores does not help most desktop applications especially if it requires lowering the frequency or reducing the processing power of each core to allow for more cores at the same TDP.
 
I thought I travelled back in time. I am SURE this was news like 2 months ago. Now, where did I put my hoverboard...

16 cores sounds nice, unless you're talking about games. More cores, in the current paradigm, means more latency. No way around that.

Yeah... Super old news. Check the date of article.
 
16 Core AMD server chips have been out for a bit, havn't used them but they are a lot better priced than intels alternative.
 
AMD has been buying up old engineer's lately. They hired back Jim Keller from Apple who used to work for AMD. Seems like AMD is finally taking it's CPU business serious.
 
Now, though, Jim Keller, the one who spearheaded the invention of the K7 and K8 architectures that were better than anything else in their time (though we're sure some disagree with that) is back at Advanced Micro Devices

No disagreement is possible. For well over a year, Intel was eaten as a snack by AMD cpu performance--you couldn't find a game site anywhere using Intel cpus in that period...;) It was during this period that Intel began paying companies not to buy AMD cpus and threatening OEMs with shortages and higher prices if they also sold AMD cpus, etc. Intel recently gave AMD one thousand two hundred and fifty million dollars ($1.25B) in cash to settle that lawsuit, and further agreed on paper that it would never commit those anti-competitive actions again. I always believed that AMD should have held out for at least $10B, but I imagine they were ready to end the thing and Intel was ready to pay up and move on. Intel's great success with Core 2 comes directly from the x86-64 license extended to it from AMD under a cross-licensing agreement between the two companies. As a result of AMD's competitive fire in the cpu performance race at that time, we are where we are today instead instead of where Intel wanted to take us, which was Itanium 64. Had that happened, Intel would have owned practically everything--bus patents, included--which would have made the likelihood of another company rising to compete with it like AMD almost an impossibility. The market wisely went with x86-64 and rejected Itanium completely.

Intel is always "worried" about competition. That's just good sense. AMD has soundly beaten Intel once (when all the pundits said it was impossible for anyone to beat Intel in performance); so I see no reason why the company could not do it again.
 
AMD was only ahead of the P4 architecture. Before that it was considered a less powerful budget chip. And since the Core 2 days, AMD has been playing catch up and remains a less powerful budget chip. I just don't see this changing. But competition is healthy for consumers so here is to hoping!
 
16 or 32 cores possibly sometime soon. At least somethings happening. Been stuck at 4-8 cores forever now.
 
I think AMD realized offering a stock licensed ARM core put it in the same category as Allwinner. :p There's no money there. Apple, Samsung, STM and Qualcomm are the leaders at the higher end of ARM architecture licensees (as opposed to the masses of unwashed core licensees). What this means is that AMD is deciding to forego a market with one competitor (Intel) which doesn't have a vertically integrated business in favor of going against 4 entrenched players which each have many years of ARM design experience over any n00bs. glwt, even if AMD manages to make a competitive product for the high end.

No disagreement is possible.
With the K8 at least, until Conroe came out. K7 is more complicated. It was competing against Northwood later in its release cycle, which was faster than Athlon XP at top speeds available for each. At the end of the P3's life, K7 was faster of course. But performance was mixed with Willamette P4 vs Athlon XP.

It's easy to make a throwaway claim, but only 1/2 his statement was unambiguously true.
 
Now if he can just invent K9!

K9 was cancelled after he left: http://en.wikipedia.org/wiki/AMD_K9



what about K10?!?

Phenom, Phenom II and Athlon II models are K10



Bulldozer, Piledriver, Steamroller and Excavator are K11 chips

The current Kaveri A-series APUs are steamroller chips

Carrizo A-series APU coming out later this year will be from the excavator series chips

Everything released after that will have had some sort of input from Jim's decisions.
 
I think AMD realized offering a stock licensed ARM core put it in the same category as Allwinner. :p There's no money there. Apple, Samsung, STM and Qualcomm are the leaders at the higher end of ARM architecture licensees (as opposed to the masses of unwashed core licensees). What this means is that AMD is deciding to forego a market with one competitor (Intel) which doesn't have a vertically integrated business in favor of going against 4 entrenched players which each have many years of ARM design experience over any n00bs. glwt, even if AMD manages to make a competitive product for the high end.

With the K8 at least, until Conroe came out. K7 is more complicated. It was competing against Northwood later in its release cycle, which was faster than Athlon XP at top speeds available for each. At the end of the P3's life, K7 was faster of course. But performance was mixed with Willamette P4 vs Athlon XP.

It's easy to make a throwaway claim, but only 1/2 his statement was unambiguously true.

Anything that was not SSE optimized was buried by the AMD chips at the time.

If AMD could go back in time they should have released another generation of K9 chips optimized for less leakage while getting Phenom debugged and released a better Phenom in the late spring of 2008.
 
16 or 32 cores possibly sometime soon. At least somethings happening. Been stuck at 4-8 cores forever now.

The reason we haven't seen CPU's jump onto 8+ cores is because of coding. Lots of x86 code that developers don't want to rewrite to take advantage of multi-cores. Most code is stuck with single threaded design, and that's why Intel wins. If everything was written multithreaded, then the AMD FX 8350 would be a clear winner.

This is also why GPU's can do certain tasks better then CPU's. They scale must better with code compared to CPU's. Also, we'll need must faster ram than DDR4. It's a wonder why memory sticks don't come standard with 128-bit per stick.
 
If everything was written multithreaded

You do realize this is impossible. I mean not every algorithm can benefit from more than 1 thread.

the AMD FX 8350 would be a clear winner.

It could be a clear winner over an i5 but I do not see it being a clear winner over an i7. They cut too much out of each core to make that a reality.
 
Originally the goal was to give superior performance.

An 8 core i7 w/hyperthreading (showing 16 cores in windows) would be marketed against a 16 core 22nm steamroller chip (also showing 16 cores in windows).

That was the original intention with AMD being the faster, better performer and charge more money per chip.

Technically they could then place a 12 core chip against intels 8 core one and still perform better.

But alas... bugs galore...no fab to produce the product, product not scaling properly...etc.

Jim's back.... I expect something better.

Technically excavator was supposed to have a better FPU and interger unit....25-30% better. But the short term roadmap is not what it once was anymore.
 
A lot of misleading things said about high core count and multi-threaded software from people with only 4-core. 16-core loves Frostbite multi-threaded engine in Battlefield 4 and with a budget R9 270X it nearly maxes out the 200 fps cap at ~190 fps on Test Range. HDD disk I/O seem to benefit too since I usually spawn early with the SSD equipped.
 
is this type of talent that rare?

Evidently yes.

High level CPU architect who has a history of making the right calls, inspires others around him...and has good project management skills.... it also helps that he brought back a near platoon of seasoned design engineers with him.
 
It could be a clear winner over an i5 but I do not see it being a clear winner over an i7. They cut too much out of each core to make that a reality.

The 8 core Bulldozer and Piledriver chips actually did well against the i7s in multithreaded tasks like video encoding and the newer Battlefield games. Intel has a significant IPC advantage, but with 8 physical integer cores the AMD chips can keep up with the performance of the hyperthreaded Intel quads while being priced to compete with the i5s. Of course, since most stuff is single threaded the Intel chips win out in benchmarks.

I have both an FX 8320 and an i7 4930k. I really dont see any difference between the two in real-world use. The 8320 was significantly cheaper and I prefer AMD's chipsets. Seriously, why did Intel wait until Haswell to roll out a chipset with native support for more than more than 2 SATA 6GB/s ports?
 
AMD was only ahead of the P4 architecture. Before that it was considered a less powerful budget chip. And since the Core 2 days, AMD has been playing catch up and remains a less powerful budget chip. I just don't see this changing. But competition is healthy for consumers so here is to hoping!

AMD was first to 1GHZ, that was in the Pentium 3 days.
 
why did Intel wait until Haswell to roll out a chipset with native support for more than more than 2 SATA 6GB/s ports?

Because there is very little advantage to SATA III with more than 2 SSDs.
 
Well, in a non-scientific, seat of the pants test; when I switch from a 4 core to a 8 core machine, my FPS performance doubled using HandBrake. Take that for what it's worth.
;)

Yeah, that's pretty accurate from my experience, too. Handbrake is nearly a perfect core-for-core performance boost scenario. Amusingly, video encoding is about the only thing on my desktop that fully maxes out all four cores at the same time... and this is a CPU I bought almost 6 years ago. Added Ram (16GB to 24GB) was more helpful than a newer processor.
 
Amusingly, video encoding is about the only thing on my desktop that fully maxes out all four cores at the same time

Video coding standards are designed with parallel processing in mind so this tends to be one of the cases where parallel algorithms scale well.
 
Evidently yes.

High level CPU architect who has a history of making the right calls, inspires others around him...and has good project management skills.... it also helps that he brought back a near platoon of seasoned design engineers with him.

It sounds like a promising hire but I have to wonder how much one guy can really do to turn around the sinking ship of AMD's desktop CPU division.
 
This news means nothing. Nothing will happen, nothing new will come of it. Multi-Core as of now is merely, for 99% of users, a marketing gimmick.

Not true... not true in the LEAST. UNLESS you are calling multi-core as anything more than 4 cores. In that case I agree.

I couldn't IMAGINE going back to a single core system now.... I could barely imagine going back to a DUAL core system actually. 3 cores would be a minimum that most users could get by with.
 
Not true... not true in the LEAST. UNLESS you are calling multi-core as anything more than 4 cores. In that case I agree.

I couldn't IMAGINE going back to a single core system now.... I could barely imagine going back to a DUAL core system actually. 3 cores would be a minimum that most users could get by with.

Meh. I disagree. My Chromebook is an i3/Sandy Bridge based Celeron. 1.1Ghz Dualcore and it does me more than well enough to be my daily driver. I only use my desktop for video editing, encoding and animation/rendering when not playing games. I have absolutely no problems using the Chromebook (has XUbuntu 14.04 installed to it instead of ChromeOS, mind you)

System was nigh unusable at 2GB Ram installed. Upped it to 10GB and everything runs great.

I believe we have long since passed the point in which the CPU is capable of being anything that would set back a system. Other factors, such as the GPU, disk throughput and amount of ram are far more important once you have any kind of dualcore CPU, IMHO.
 
The reason we haven't seen CPU's jump onto 8+ cores is because of coding. Lots of x86 code that developers don't want to rewrite to take advantage of multi-cores. Most code is stuck with single threaded design, and that's why Intel wins. If everything was written multithreaded, then the AMD FX 8350 would be a clear winner.

This is also why GPU's can do certain tasks better then CPU's. They scale must better with code compared to CPU's. Also, we'll need must faster ram than DDR4. It's a wonder why memory sticks don't come standard with 128-bit per stick.

Spoken by someone who knows nothing about software engineering. You can run as many threads as you want on any number of cores. Threads != cores. There is virtually no correlation between them besides parallel computing. How the threads get divided out to the cores is the compilers job. Nothing to do with the software being written. So if you want to complain about poor multicore optimization complain to the companies that design all of the mainstream compilers for all the various programming languages used across all of the different types of programs in the wild. GNU makes free compilers for many different programming languages, start there.

TL,DR: There can be any number of threads running on a single cpu let alone multicpus. Software engineers do not specify which thread goes to which CPU. That is the compilers job, complain to them. Also as was pointed out already, there's not always a need to have multiple threads in every application and sometimes you just run a separate thread to do something very basic like monitoring/polling certain routines or for other things like timers.
 
AMD was only ahead of the P4 architecture. Before that it was considered a less powerful budget chip. And since the Core 2 days, AMD has been playing catch up and remains a less powerful budget chip. I just don't see this changing. But competition is healthy for consumers so here is to hoping!

not true. The original Athlon ate up the P3 at the time of launch, especially in games, as it's FPU trounced Intel at the time. What held Athlon back was the motherboards were less than stellar, and not as mature as Intel platforms. P4 was a gift to AMD. Netburst was something AMD should have taken to heart when they designed these new multi core cpus that seem to follow the same pitfalls Intel had with P4. Lower IPC and heat issues. Intels "Core" cpus were tweaked versions of the P3 that developed into their laptop Centrino or w/e they called it at the time, which was pretty badass. Maybe AMD will revamp the good in K8/phenom and put that on steroids like Intel did the P3 to core development.
 
It sounds like a promising hire but I have to wonder how much one guy can really do to turn around the sinking ship of AMD's desktop CPU division.

similar things were said back in the AMD K6-2 and K6-3 days. My first custom PC was a K6-3 which was pretty decent but slower than equivalent P3's of the day. still not bad for the $ though.

All the hate against AMD is silly. Love your intel cpus all you want, heck I have some as well. But keep intel in check b/c I do remember Intel being almost the only game in town, and their arrogance now is approaching that level it was back then, when they stagnated and made the P4 and IA64. AMD may not execute, or out perform Intel, but they are some form of competition and we can hope they become more relevant in the performance arena again. I have some 7850K cpus and a 8320 cpu and they are not shabby cpus. Do everything I need just fine for a good price. My intels are good too, in different ways of course. To each their own, but all you dipshits that want AMD to die have no concept of history and what will happen if AMD does go under or abandons x86 entirely.
 
The 8 core Bulldozer and Piledriver chips actually did well against the i7s in multithreaded tasks like video encoding and the newer Battlefield games. Intel has a significant IPC advantage, but with 8 physical integer cores the AMD chips can keep up with the performance of the hyperthreaded Intel quads while being priced to compete with the i5s. Of course, since most stuff is single threaded the Intel chips win out in benchmarks.
From what I see they sort of trade blows with the quad i5/i7's in heavily multithreaded applications and lose miserably in anything that doesn't scale well past 4 cores. I mean, not bad for the price, but I've never seen a reason not to pay the $30 and get the i5 k-series over the 8350. It doesn't need to be single threaded to favor Intel, just to not really take full advantage of 8 vs 4.

As far as all the 2,4,6,8 core vs software talk goes: I'm not going to pretend I know jack about coding for multi-threading, but having spoken to several programmers talking about scientific programming (wind field calculations, computational fluid mechanics, molecular/orbital modeling) multi-threading and parallelization is not as easy as just smashing some crap into a compiler and having it split threads up. The working data set needs to actually benefit from parallel computations. You can split all the threads you want but if they're sitting waiting on data because there is a dependency it doesn't mean shit. Sure it helps if you've got lots of different programs running and crunching in the background to keep them all happy, but if you want to get benefits out of 1 single program they won't all scale, and thats not always the programmer's fault.
 
is this type of talent that rare?
The internet tends to look for celebrity engineers who are somehow primarily responsible for the success or failure of a product. In reality most people have no idea what goes on behind the scenes.
 
Back
Top