• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

AMD Zen CPUs?

well to be fair its not entirely AMD's fault, but what they were asking from a processing node was entirely to high.

its not the design which is limited, but the node of which the processor is built on. AMD thought they would be able to get a high performance node that would suit their needs, but well they didn't. Still they are the highest clocking cpus on the market, it does show they had a node that was somewhat suited for their needs.
 
Only, you can't reach 5+Ghz without top-end cooling, and 6GHz is out-of-the-question. They dreamed of 10GHz just like Intel did with the P4 (which had similarly slow L2 cache), and hit the same process/power wall. When you consider how far back this design was begun (2005), it's not hard to see why they thought the sky was the limit on frequency, and process shrinks would continue to give them massive power reductions for decades to come. But I still wonder why they didn't see the failure of the P4 as something they should try not to replicate!
Click to expand...

You hit the nail on the head, especially with the last sentence. From what I recall, they wanted to hit >5ghz speeds right out of the box with Bulldozer, but they didn't anticipate GloFo's process node not being up to snuff. But then again, I'm not sure why they expected otherwise in the first place. Intel's approach of a higher-IPC, lower-clock design with fatter cores was clearly a winning design ideology.
 
NaroonGTX said:
You hit the nail on the head, especially with the last sentence. From what I recall, they wanted to hit >5ghz speeds right out of the box with Bulldozer, but they didn't anticipate GloFo's process node not being up to snuff. But then again, I'm not sure why they expected otherwise in the first place. Intel's approach of a higher-IPC, lower-clock design with fatter cores was clearly a winning design ideology.
Click to expand...

See, it's not really Global Foundry's fault that the design cannot be manufactured as-conceived. Between the start of Bulldozer in 2005 and ther release in 2011, the industry went through lots of growing pains with the Deep Submicron transition, and we learned a lot AFTER Bulldozer was already being built:

Each process shrink before used to reduce the threshold voltage required to drive gates, and reduced the amount of charge required to activate those gates. You got more area to put things into the chip, and with the introduction of Copper interconnects the future looked bright in the early 2000s.

But since then, we've seen the end of that road. Shrinks are not giving you as much reduction in switching power because the wire interconnect delay is higher than the gate delay (unless you use careful design tricks). Also, they no-longer reduce the threshold voltage significantly because:

1. They're hitting the physical limitations of activation voltage for silicon, and can't go lower. Subthreshold is a pain-in-the-ass.

2. They're hitting the point where the threshold voltage is so low it's making leakage the dominant power drain, which is bad for mobile devices, but if we don't lower it the gates can't switch fast enough. This is why we have added the ability to power-down cores entirely (Vdd=0), as clock-gating alone does nothing for leakage. They've also added the ability to mix high and low threshold transistors (slow and fast) so you can only use high-leakage parts in critical paths.

With TrinityKaveri, it's likely that they increased the threshold voltage to reduce leakage (since it's targeted at laptops, and won't run over 4 GHz at 35w or less). But Bulldozer/Piledriver definitely are leaky designs (low vth = faster switching time). But in Deep Submicron you also have to account for the wire delays, which can really only be surmounted by raising the voltage. This means hitting target speeds requires a massive voltage increase, which means you have a leaky + high dynamic power core = POWER WALL!

It's unfortunate that the interconnect delay is likely what killed the grand plans of Bulldozer, and there's not a process node in existence that would make it work.
 
Last edited:
Interesting conversation. Its bringing me back to some of the original Talking Points that AMD would bring up when discussing the Bulldozer project. Some of the major points were:

1) More cores to give you more possibilities

and

2) High GHz, ideally something that would be very dynamic

Obviously when BD hit, most of us were expecting 4ghz+ and manageable thermal. I don't think anyone running a desktop would care about differences in power consumption, as long as we could handle it easily. Low and behold what we got were in opposite directions of each other, lower speed, terrible power consumption.

This kind of deals with Intel's plan to solve all their problems by adding more hz when they originally released the P4 with RDRAM. I liked AMD's plan but it was ultimately set to fail and the worst part was the writing was on the wall before they even started.

So I hope AMD does take the intel approach on this next rendition. I don't "need" intel like speed but I would like competition that will drive better product releases.
 
I can get to 5GHz on my A10-6800k, stable, with a little over 4.23v and an H80i...but it runs hot in stability testing. For that reason (70+ deg C) I never tested it longer than about 15-20 minutes, but it hasn't crashed in games (Portal, Civ5, KSP) or benchmarks (cinemark, ~44fps; unigine valley/heaven, ~10fps) yet.
 
Baseless speculation time!!!

My prediction will be that:

The 'zen' CPUs will still use a modular design philosophy, sharing certain elements as well as an L2 across multiple cores and sharing a small amount of L3 over the whole CPU. There may be two or four cores per module.

The CPUs will have an integrated dual-channel DDR4 memory controller, unlikely to see a tri or quad channel.

They will use a virtual threading method similar to HT for their integer operations, allowing each separate integer core to work on two threads per cycle.

The desktop components will launch in 2, 4 and 6 core variants with integrated radeon graphics. The integrated graphics will match a 260x at the top-end.


Baseless arguments begin!
 
maxius said:
#1 2016 and it is ment for replacement for the cmt construction cores and aimed at the top end of the market so yes.
#2 Zen is a new smt architecture from the masterminds that gave us x86-64, the k8, the imc and ddr1 (if it wasnt for amd we would be on ia64 cpu's fed by rambus memory)
#3 amd is playing this close to the vest gl finding info before they want it out in the public
#4 very valid questions no answers yet turn in after excavator is out the door
Click to expand...

x86-64 has a lot of legacy "luggage", I would hesitate to use the word "masterminds" in that context.
A "mastermind" would have changed ISA in an effective manner, not prolonged a ISA in such a manner.
x86 is +30 years old, it shows, sadly.

And the K8 were, as I recollect, nothing more than a K7 with x86-64 and a IMC, not the first CPU to use a IMC either?
Again I must disagree with your use of the term "masterminds" and the credit you apply, no offense.
 
Factum said:
x86-64 has a lot of legacy "luggage", I would hesitate to use the word "masterminds" in that context.
A "mastermind" would have changed ISA in an effective manner, not prolonged a ISA in such a manner.
x86 is +30 years old, it shows, sadly.

And the K8 were, as I recollect, nothing more than a K7 with x86-64 and a IMC, not the first CPU to use a IMC either?
Again I must disagree with your use of the term "masterminds" and the credit you apply, no offense.
Click to expand...

Please tell me which CPU you could have purchased and used at home/work that had an integrated IMC other then the A64?
 
Trimlock said:
Interesting conversation. Its bringing me back to some of the original Talking Points that AMD would bring up when discussing the Bulldozer project. Some of the major points were:

1) More cores to give you more possibilities

and

2) High GHz, ideally something that would be very dynamic

Obviously when BD hit, most of us were expecting 4ghz+ and manageable thermal. I don't think anyone running a desktop would care about differences in power consumption, as long as we could handle it easily. Low and behold what we got were in opposite directions of each other, lower speed, terrible power consumption.

This kind of deals with Intel's plan to solve all their problems by adding more hz when they originally released the P4 with RDRAM. I liked AMD's plan but it was ultimately set to fail and the worst part was the writing was on the wall before they even started.

So I hope AMD does take the intel approach on this next rendition. I don't "need" intel like speed but I would like competition that will drive better product releases.
Click to expand...

It's bit hard to remember as that was few years ago but I think general expectation was IPC bit better than Phenom, with higher frequency than Intel chips and second core being much stronger version of HT.
 
ccityinstaller said:
Please tell me which CPU you could have purchased and used at home/work that had an integrated IMC other then the A64?
Click to expand...

I am starting to feel a bit old now, so bear with me while I take you down memory lane.

I once had a 25MHz i386SL IBM laptop.
If you examine that CPU, you will find 1 answer.

But a IMC is not always the best solution, if memory serves me well Intel's Core2Duo with FSB had better memory performance than AMD's counterparts that used IMC.
One of the reasons AMD's K8 arch looked so "good", was Intel's failure with their NetBurst arch, more than any "mastermindery" on part of AMD's part.

But that is really a detour (the performance/availability) as the objection I made was to the term "mastermind" be used about the adding of a IMC and the prolonged lifesupport for an again ISA.

Again, a true "mastermind" would not have made a wheelchair for an ISA old.
But you can hardly blame AMD for doing so.
If Intel had launched a new ISA, I doubt AMD would have gotten a license (no IBM to dictate terms anymore for Intel's IP) so one could argue they had no choice.

But having no choice != mastermind in my opinion.
 
Factum said:
I am starting to feel a bit old now, so bear with me while I take you down memory lane.

I once had a 25MHz i386SL IBM laptop.
If you examine that CPU, you will find 1 answer.

But a IMC is not always the best solution, if memory serves me well Intel's Core2Duo with FSB had better memory performance than AMD's counterparts that used IMC.
One of the reasons AMD's K8 arch looked so "good", was Intel's failure with their NetBurst arch, more than any "mastermindery" on part of AMD's part.

But that is really a detour (the performance/availability) as the objection I made was to the term "mastermind" be used about the adding of a IMC and the prolonged lifesupport for an again ISA.

Again, a true "mastermind" would not have made a wheelchair for an ISA old.
But you can hardly blame AMD for doing so.
If Intel had launched a new ISA, I doubt AMD would have gotten a license (no IBM to dictate terms anymore for Intel's IP) so one could argue they had no choice.

But having no choice != mastermind in my opinion.
Click to expand...

And mastermind doesn't mean long term-better-than anything. It means the first responsible for its invention. And not exactly associated with extreme intelligence as well.
 
Michaelius said:
It's bit hard to remember as that was few years ago but I think general expectation was IPC bit better than Phenom, with higher frequency than Intel chips and second core being much stronger version of HT.
Click to expand...

Yea outsiders were but not AMD :D
 
^Yeah, lol. AMD engineers at the time outright said that the IPC per core would be a bit lower than K10.5 clock for clock.
 
durquavian said:
And mastermind doesn't mean long term-better-than anything. It means the first responsible for its invention. And not exactly associated with extreme intelligence as well.
Click to expand...

But it wasn't the first IMC?

Intel: i386SL
DEC: Alpha 21066
HP: PA-7300LC

All had IMC
All before K8
Skillfull yes...mastermind, no.
 
Factum said:
But it wasn't the first IMC?

Intel: i386SL
DEC: Alpha 21066
HP: PA-7300LC

All had IMC
All before K8
Skillfull yes...mastermind, no.
Click to expand...


yes those had IMC's. Although not be confused with the Athlon 64's 64 bit one.

Give them credit where credit is due, and try not to down play their achievements.
 
ebduncan said:
yes those had IMC's. Although not be confused with the Athlon 64's 64 bit one.

Give them credit where credit is due, and try not to down play their achievements.
Click to expand...

I do, I said "skillful".
I disagree with the term "mastermind" being used, look at my posts again.
 
The point is that AMD were the first to have a 64-bit IMC for any consumer platforms. Yes they didn't invent the concept of the IMC but that's not what it was ever about. The murmuring about whether or not they were "masterminds" for that is just fluffy semantics.
 
Factum said:
x86-64 has a lot of legacy "luggage", I would hesitate to use the word "masterminds" in that context.
A "mastermind" would have changed ISA in an effective manner, not prolonged a ISA in such a manner.
x86 is +30 years old, it shows, sadly.

And the K8 were, as I recollect, nothing more than a K7 with x86-64 and a IMC, not the first CPU to use a IMC either?
Again I must disagree with your use of the term "masterminds" and the credit you apply, no offense.
Click to expand...

listen here troll can you develop integrated circuits it takes 5+ years for an architecture to come to release. no the k8 was not a k7 with an imc and x86-64 dont discredit amd's past achievements. if it was not for amd we would be all on ia64 right now with rambus memory intel had zero plans to move forward with x86. yes the ic developers at amd are masterminds the k8 was a ground up architecture and the 1st x86-64 chip with an imc that supported ddr1 and at the time of the pentium 4 and rambus it took intel to the woodshed. the k8 put the hurt on intel if Intel hadn't cut special deals with oem's to keep amd out amd would be in a far better place than it is today.

BACK ON TOPIC PLEASE
 
maxius said:
listen here troll can you develop integrated circuits it takes 5+ years for an architecture to come to release. no the k8 was not a k7 with an imc and x86-64 dont discredit amd's past achievements. if it was not for amd we would be all on ia64 right now with rambus memory intel had zero plans to move forward with x86. yes the ic developers at amd are masterminds the k8 was a ground up architecture and the 1st x86-64 chip with an imc that supported ddr1 and at the time of the pentium 4 and rambus it took intel to the woodshed. the k8 put the hurt on intel if Intel hadn't cut special deals with oem's to keep amd out amd would be in a far better place than it is today.

BACK ON TOPIC PLEASE
Click to expand...

Could you post the major architechtual differences, I would love that.
It beats namecalling.
 
I expect all we can do is speculate for the next year+ that is unless AMD releases any official information. I suspect they will keep quiet this time.
 
ManofGod said:
Original topic, do you have anything to say about it?
Click to expand...

I did and I do, but some people seem to have very thin skind.

Here is why I say K7->K8 didn't alter the "core-design" very much:
http://www.anandtech.com/show/1098/2

It's not as major as K6->K7
It's not as major as K8->10h(K10)

Ontopic I will say, that I need to see Zen in action before making judgement, CPU's are very hard and AMD is not the AMD it was a few years ago (they cut away a lot of manpower).
 
Factum said:
I did and I do, but some people seem to have very thin skind.

Here is why I say K7->K8 didn't alter the "core-design" very much:
http://www.anandtech.com/show/1098/2

It's not as major as K6->K7
It's not as major as K8->10h(K10)

Ontopic I will say, that I need to see Zen in action before making judgement, CPU's are very hard and AMD is not the AMD it was a few years ago (they cut away a lot of manpower).
Click to expand...

Yeah, and they brought back a lot of manpower as well. That makes them not the AMD of even the last few years when Bulldozer was released. Lets make one thing clear, it is not thin skin around here but your attitude of, "hey, I was just joking" type stuff that is trolling.
 
Factum said:
I did and I do, but some people seem to have very thin skind.

Here is why I say K7->K8 didn't alter the "core-design" very much:
http://www.anandtech.com/show/1098/2

It's not as major as K6->K7
It's not as major as K8->10h(K10)

Ontopic I will say, that I need to see Zen in action before making judgement, CPU's are very hard and AMD is not the AMD it was a few years ago (they cut away a lot of manpower).
Click to expand...

Why don't you read the whole article? k7->K8 had many changes, not just x86-64 and IMC, the pipline changed significantly, branch prediction, hyper-transport, etc. obviously they would build on their existing technology, just like the current i7 is built off of the pentium 2 / 3 / core m.
 
The thing about Zen that I'm wondering about is AM4 this time around or will remain in the FMx segment?
 
Digital Viper-X- said:
Why don't you read the whole article? k7->K8 had many changes, not just x86-64 and IMC, the pipline changed significantly, branch prediction, hyper-transport, etc. obviously they would build on their existing technology, just like the current i7 is built off of the pentium 2 / 3 / core m.
Click to expand...

That is not the argument I was responding to, this is the one I was responding to:

maxius said:
listen here troll can you develop integrated circuits it takes 5+ years for an architecture to come to release. no the k8 was not a k7 with an imc and x86-64 dont discredit amd's past achievements. if it was not for amd we would be all on ia64 right now with rambus memory intel had zero plans to move forward with x86. yes the ic developers at amd are masterminds the k8 was a ground up architecture and the 1st x86-64 chip with an imc that supported ddr1 and at the time of the pentium 4 and rambus it took intel to the woodshed. the k8 put the hurt on intel if Intel hadn't cut special deals with oem's to keep amd out amd would be in a far better place than it is today.

BACK ON TOPIC PLEASE
Click to expand...

Shifting golaposts are boring, no offense(not that you did it, but the end result is the same).
And still the IMC did this:
http://www.anandtech.com/show/2045/5

But it seems people are far to touchy on this topic.
And I still don't see giving x86 extended lifesupport as the fineste hour of CPU "masterminds", no offence.
 
Factum said:
That is not the argument I was responding to, this is the one I was responding to:



Shifting golaposts are boring, no offense(not that you did it, but the end result is the same).
And still the IMC did this:
http://www.anandtech.com/show/2045/5

But it seems people are far to touchy on this topic.
And I still don't see giving x86 extended lifesupport as the fineste hour of CPU "masterminds", no offence.
Click to expand...

Your response is a standard troll tactic. Enjoy your short stay here.
 
Personally, I don't care if AMD's next high-performance desktop CPU is an APU, so long as I get >4 threads that can compete with an i5, I can overclock to make up for the rest... :)
 
Digital Viper-X- said:
why are people so anti-apu? the current high end i7s are all "APUS"
Click to expand...

Trimlock said:
I would prefer to have cpu's with both options honestly.
Click to expand...

While I agree that all current CPU's are "APUS" unless AMD/Intel decides to turn em off (for consumer parts, not talking about Xeon/Opteron SKUs), I can see the appeal of both...While I don't use the onboard GPU of my 3770K, I have used it in the past and it has come in handy for short periods of testing etc..I think the main issue is the current state of OpenCL..It is continuing to get better with more and more software supporting it, but it will be decades until everything uses it.

Take video encoding for example. If I want quick and X% quality, I can use Handbrake on the iGPU without issue. But many (myself included) prefer higher quality rips that traditional x86 offers. This is time consuming, but it scales lineally with every core you add, so the 'moar cores the better!
 
Trimlock said:
I would prefer to have cpu's with both options honestly.
Click to expand...

what difference does it make though? For the company to have to create a separate sku, saparte dies, etc, it would probably end up costing more for the few people that want it. (EG: SB-E / IB-E / Haswell-E)
 
Digital Viper-X- said:
why are people so anti-apu? the current high end i7s are all "APUS"
Click to expand...
Intel high end i7's are the LGA2011 chips, not LGA 115x. LGA2011 chips do not come with built in graphics.:p

That bit of misunderstanding aside, I do understand the point you are trying to make, and I do agree with you.

The problem I think people see is the fact that AMD's focus on APU's has primarily been improving the GPU portion, when it is the CPU portion that desperately needs attention. The other problem is the overwhelming portion of the TDP and silicon used in AMD's current APU's are dictated by the onboard graphics, which eat up any remaining silicon area and thermal headroom available for improving the CPU.

Don't get me wrong, the methodology AMD has used thus far on APU's is decent and they don't need to improve the CPU as much since these chips cater to the low end of the market segment. They are staying afloat and being smart by catering to the masses, rather than the niche. But AMD needs to refocus on the CPU if they are going to make chips that compete with Intel's high end.

I see no problem with a mid-level chip having an onboard GPU, but AMD still has a long ways to go before their mid-level chips even begin to compete with more than Intel's low end.
 
R3MF said:
fm3.

there will most likely be no am4, they aren't aiming above s1150.
Click to expand...


Amd said, they are specifically targeting the high end market. Not the main stream market although i'm sure they will have designs to target that market as well.

They don't scrap their entire current cpu advancement and go to work on a new architecture to be released in 2016 to shoot for a mainstream market.
 
Digital Viper-X- said:
what difference does it make though? For the company to have to create a separate sku, saparte dies, etc, it would probably end up costing more for the few people that want it. (EG: SB-E / IB-E / Haswell-E)
Click to expand...

For any reason I can think of, there are reasons which is why both Intel and AMD has these options, just not on the same sku. I think they should keep it.
 
AMD has never competed in the 'E' segment. They have two server-only sockets (C32 and G34) and none of them have consumer parts. AMD's AM3 is a relic that competed with 775/1156. FMx is what future chips will be released on.
 
You must log in or register to reply here.
Back
Top