Scali wrote:

I'm trying to have a mature conversation here

Not anymore, you're not. Not when your "mature" conversation includes stuff like this:

And all we get are stupid AMD fanboys and trolls.

So really, why exactly did you make this post, with these obvious factual errors?

It's incredibly annoying to have to keep explaining the same thing over and over again

(either that, or basically just clueless about what a modern CPU actually consists of, and can't look further than the overly dumbed-down view of [...]

Look, I love a good argument. I tend to learn something in the process, and that's always good. So I'll argue until the point where it becomes obvious that 1) we won't agree; 2) we're arguing different arguments; 3) someone gets butt-hurt, because that's when it stops being an interesting conversation, it's a pissing match, and that's not fun for anybody.

You're not the only one here slinging mud, Scali, but you do have a predictable pattern where you step in, dominate an argument (albeit usually because you legitimately have something to add, but then) beat the horse to death, use bullet points that are sometimes not exactly 1:1 with the argument at hand as a slippery way of being "right" while not quite having the same conversation, then you slowly get more and more defensive, and then offensive, when someone doesn't give in and accept your stance. Maybe you do need to just let it go. What are you hoping to accomplish?

I don't like piling on to someone, but at any given time there's at least one thread where you're the bully. It takes an army of opponents to keep up. Please, either be the bigger man and knock it off with the aggressive attitude, or if you just can't, then do yourself a favor and avoid debate. Just walk away from it. You're a valuable resource here, and I respect you, but sheesh bruh...

SirNickity wrote:

Not anymore, you're not. Not when your "mature" conversation includes stuff like this:

Oh yea, single me out, that's going to help!
You're part of the problem.

Other than that I think you have the situation backwards:
I say something, and I'm basically always right, because, let's face it, I'm a smart and experienced guy, who doesn't bother to open his mouth unless he actually knows what he's talking about, and thinks he has something valuable to contribute. That's just how I roll. I try never to be wrong, and I generally succeed in that quite well.
That is not 'bullying', and I can't help 'dominating' a subject just because I happen to be knowledgeable (it isn't 'dominating' unless someone actually tries to argue against me and is continuously proven wrong because they basically are not properly informed... but I can't help that. It's their own misjudgement of their own knowledge that leads them to that situation. That situation doesn't occur when someone starts doubting themselves and asks questions for clarification, or goes looking for more information by themselves. It only happens when people refuse to admit they were wrong).

Problem is, especially with 'touchy' subjects, such as AMD or linux/GPL/'free software', some people just will not accept alternative views, not even when they are properly supported by facts and logical argumentation.
These people (plural being the key here) then continue to attack me until the end of time, and derail every possible discussion. They are bullying me. They think they are entitled, because there's multiple of them, legitimating and reinforcing their view (you are the perfect example of that, you single me out and are bullying me, and are blissfully ignorant about the fact that you're doing it, thinking you're entitled, because you're on 'the other side', which in your view is 'the right side', moral superiority).
I may not be defending the most popular side, opinion, facts or whatnot, but that doesn't give people the right to act the way they do.

Read the thread again, VERY clearly, and see how often people jab at me, try to derail the discussion with strawmen, bringing issues back up that had already been decided etc, sling personal attacks around etc (like 'willfully obtuse' etc)... And then look at what point I finally decide not to act that nicely anymore, because clearly it isn't getting through these people's heads that their conduct is unacceptable, and they should just be more mature and admit when they've been proven wrong, or at least concede that not everyone is going to see things their way. I very clearly do NOT instigate such situations, EVER!

You calling me out as a bully doesn't exactly help. You are again reinforcing the problem: I'm the bully, so it's okay to attack me. That's what you're saying.
If people would just stop ganging up on me the way they have been doing over the years in various threads, the problem will simply go away, because I'm never the one that starts it.
NEVER.

And arguing that I need to "walk away" when I'm just pointing out that Westmere has its CPU logic split over two dies, that's just retarded to be honest.
Why would I need to walk away? I'm not spreading lies, misinformation, rewriting history or anything. I'm merely pointing out a simple fact. The problem is that some of these facts are apparently rather 'sensitive' to a certain group of people. It's these people who need to be dealt with.
I don't see any reason why I would accept that people rewrite history on Westmere just to make Zen 2 look good. Why would I have to "walk away"?
What is the value of a discussion forum when discussions can't possibly lead to fact-finding and valuable information, because all discussions are dominated by some ignorant bullies with an agenda, who will just scare everyone out of any discussion so they can have the last word, and they can rewrite history as they see fit?
Where everyone who dares to oppose this behaviour is singled out as a 'bully' and told to "walk away"?

What I hope to accomplish is exactly that: a discussion forum where we can actually openly discuss anything, in a friendly and constructive manner, and actually find facts and valuable information, not clouded by personal agendas or whatever other type of propaganda.
For that to happen, certain people need to either be chased away, or be made to understand that propaganda and other kinds of misinformation are not tolerated, so they have to clean up their act if they want to stay and contribute.

You think something like 8088 MPH could happen if we stuck to some romanticised version of hardware that some people with whatever agendas were trying to push? Or would it only be possible if we looked at the hardware and software as it is, warts and all, trying to find all factual information we can?
The same argument applies to many other software and hardware projects on this site, such as the various modern sound device replicas and their supporting software.

Amazing meltdown.

Scali wrote:

You are trying to make an economical argument (using some arbitrary metric like value/business segment) in what is a technical discussion (where the same generation of technology is generally scaled towards the various segments).

So, Clarkdale having an IGP but Lynnfield not having one is not a technical difference?

The bitch of it is, I agree with you in one sense. As I mentioned, very few technological innovations are truly that innovative (and chiplets are no exception). They're a good solution to modern process limitations now that key enablers (like high speed serial links) are more common place, but I definitely agree that chiplets are at best a new spin on an old idea. I still think that chiplets represent a reverse course from the prevailing trend which will probably be held to as Moore's Law begins to fail. Clarkdale, on the other hand, was a brief detour.

What I don't understand is why you seem to always get so mulish and proceed as if your argument is perfect when it often isn't. I also don't understand why you claim you aren't a big Intel fan (or a big AMD foe) when you so often go through contortions to fawn over them in discussion. Please try to be honest with yourself, give credit where credit is due, and try to challenge your own reasoning once in a while. You're clearly knowledgeable and I respect that, but I think your stubbornness too often gets in the way of quality discussion.

gdjacobs wrote:

So, Clarkdale having an IGP but Lynnfield not having one is not a technical difference?

That's not what you said, or at least, not how I interpreted it:
"Intel didn't have a true successor to Penryn and the G45 chipset in the value and business segments until Arrandale and Clarkdale came out, so I'm not sure the analogy holds."

I read that as "Westmere is a successor to Penryn/G45, so it's in a different segment than Nehalem/Lynnfield, so those "don't count", even though they did have memory controller and PCIe controller etc on-die"
So I read it as you saying the order is:
Penryn -> Westmere
Where technically it is:
Penryn -> Nehalem (+IMC) -> Lynnfield (+PCH) -> Westmere (-IMC -PCH)

Obviously, if Intel can do the integration on-die with the more high-end models, then they could have theoretically made a lower-end model out of it as a 'proper' successor to Penryn/G45. And since Westmere was also a die-shrink, putting the GPU on-die should not have been too much of a challenge.
The fact that these lower-end CPUs are 'missing' is merely the result of economic/marketing choices, not technical ones (I gather it made more economic sense to continue selling Penryn/G45 solutions for the time being, in that segment). The technology was available, Intel could have built them.

The situation in which Westmere was developed, may not be comparable to Zen 2, but that doesn't really matter for the point that was under discussion, does it?
The point was merely whether or not Zen 2 was being innovative with its chiplet architecture.
Where I argued that pretty much the exact same concepts were applied in Westmere (for whatever reasons): multiple dies, different manufacturing process, and high-speed links connecting them.
Was Westmere the right idea at the wrong time? Who knows. All I know is that Westmere's existence is in the way of seeing Zen 2 as a true innovation (and even Westmere is not something I would consider innovative, but that's another point altogether).

gdjacobs wrote:

What I don't understand is why you seem to always get so mulish and proceed as if your argument is perfect when it often isn't.

Where does this notion of my argument not being perfect come from?
Isn't it more because my argument is being pulled out of context by various people in order to try and construe it as something that it never was to begin with?

I mean, my argument was basically that Westmere consisted of a 32 nm die and 45 nm die in one package, I don't see what's imperfect about that.
Then people tried to pull that argument in all sorts of directions.

gdjacobs wrote:

Please try to be honest with yourself, give credit where credit is due

Did it ever occur to you that AMD gets WAY too much credit on average, and I do actually give credit where credit's due, but at a realistic/objective level, rather than the online/media fad?
Also, I came here to add some perspective on some parts of a discussion, I didn't come here to sing praise for any particular brand.
Which is always so strange. I clearly didn't give Intel a lot of credit for innovation in this thread either, did I? For example, I clearly did not give them the credit for HyperThreading.
Yet, people only focus on AMD.

And well, that's the rub, isn't it? I never actually gave my opinion on AMD's Zen 2 architecture or anything. The tone of the discussion simply didn't allow that.
Because if you would ask me what I think, then I would clearly say that AMD's approach is currently a very practical and successful one.
Intel's approach with stacked dies is ambitious, and that's part of why they're currently struggling to get any progress. They can't seem to get their 10 nm going properly.
Should they have gone for the chiplet approach as well? Who knows? It would have been a good short-term strategy, I would think. But what will their 10+ nm and stacked dies bring them in the future? Who knows at this point. Once they iron out the problems, it may just work better than AMD's solution, and then Intel may once again pull out a significant lead.

That sorta reminds me of the Pentium 4 era, where people would say that Intel's 65 nm process was horribly broken etc. It wasn't. The Pentium 4 architecture was 'broken'. Core2 ran like a total dream on that same 65 nm process.
Then people argued that they took out HyperThreading because it only worked on the long and inefficient pipelines of the P4. And it didn't. HT reappeared in Nehalem, and despite the short and efficient pipelines, HT still worked at least as well as it ever did on P4.
People often call things completely wrong, and think companies are ahead or behind in some areas, where they aren't. They may see some symptoms, but are wrong about the underlying causes, and therefore mispredict what happens when the issues are fixed.

But one thing I did say is that the two approaches aren't mutually exclusive. Which more or less implies that I think that both AMD and Intel may combine the two approaches in the future. So there is not One True Way(tm).

gdjacobs wrote:

and try to challenge your own reasoning once in a while.

What makes you think I don't?
I hate to point out the obvious, but you don't get to be as knowledgeable and experienced as I am if you don't always challenge yourself.

In all seriousness though... look... it's easy, instead of throwing insults at me you could just explain to me what the difference between a regular PCB and the FR4 substrate under the CPU die is for example (is it just that they're technically not "printed"?) cause to me PCB always meant a board with circuits connecting stuff, I'll gladly listen, but your writing style makes it really hard to not make you sound overly pretentious and hard to learn something valuable from.
I like learning new things, so if I'm wrong I'm glad when someone corrects me, but insults are not a good way to start.

also why do you care if anyone says something positive about AMD (though I don't really see how "they are not the same" has any value attached to it) even if they don't deserve it in your opinion, that is not a personal attack against you, also I think nobody is actually disputing anything proven of what you said and yet you keep repeating the same thing that we already agreed on over and over, you are taking this way to personal by several magnitudes.

if you want people to listen to you, stay calm and explain your point, if they don't agree, then they don't agree, if they don't want to learn, then they don't want to learn.
but pretentious outbursts of "I am always right" (even if that were the case) and throwing insults at people really does not help in a discussion about anything.

imi wrote:

In all seriousness though... look... it's easy, instead of throwing insults at me you could just explain to me what the difference between a regular PCB and the FR4 substrate under the CPU die is for example (is it just that they're technically not "printed"?) cause to me PCB always meant a board with circuits connecting stuff, I'll gladly listen, but your writing style makes it really hard to not make you sound overly pretentious and hard to learn something valuable from.
I like learning new things, so if I'm wrong I'm glad when someone corrects me, but insults are not a good way to start.

Funny that you demand such in-depth responses to your post, which was basically a single-line trolling fallacy.
You are not entitled to anything, and if you wanted to learn from me, obviously you need to review your approach, because it is not a very constructive one.

I'll ignore the rest of your pretentions blame-shifting post.
Funny how everyone is talking to me from a position of moral superiority, exactly as I said above.

Scali wrote:

That's not what you said, or at least, not how I interpreted it:
"Intel didn't have a true successor to Penryn and the G45 chipset in the value and business segments until Arrandale and Clarkdale came out, so I'm not sure the analogy holds."

Those segments (as a complete platform) pretty much require an IGP.

Scali wrote:

The situation in which Westmere was developed, may not be comparable to Zen 2, but that doesn't really matter for the point that was under discussion, does it?

Depends on whether you're interested in scoring points or a real discussion. To me, engineering decisions are interesting.

Scali wrote:

The point was merely whether or not Zen 2 was being innovative with its chiplet architecture.

Sure, but neither was Intel. So what?

Scali wrote:

Where does this notion of my argument not being perfect come from?
Isn't it more because my argument is being pulled out of context by various people in order to try and construe it as something that it never was to begin with?

It's a conspiracy!

Scali wrote:

gdjacobs wrote:

Please try to be honest with yourself, give credit where credit is due

Did it ever occur to you that AMD gets WAY too much credit on average

Sure, I believe I said so. Most innovations touted in the media are just rewarmed versions of stuff from before.

Scali wrote:

Also, I came here to add some perspective on some parts of a discussion, I didn't come here to sing praise for any particular brand.
Which is always so strange. I clearly didn't give Intel a lot of credit for innovation in this thread either, did I?

You trumpeted Intel as the prior art for package level integration. AMD has chiplets, but Intel got there first. Even if you don't believe so, it comes across as very defensive of Intel as the better engineering house.

Scali wrote:

Because if you would ask me what I think, then I would clearly say that AMD's approach is currently a very practical and successful one.

I agree. They've used the enabling technology (high speed serial links) to make effective use of modular package assemblies. As a result, they have a very economic, performant, and scalable design.

Scali wrote:

Intel's approach with stacked dies is ambitious, and that's part of why they're currently struggling to get any progress. They can't seem to get their 10 nm going properly.

Stacked dies has been quite successful since it debuted with mobile chipsets, but it has physics issues against it. Stacking won't be viable for functional areas with high thermal load, so GPUs and desktop CPU cores are probably out. They will be using some variant of chiplet with EMIB and some stacking, but I'm not sure how far along they are.

Scali wrote:

Should they have gone for the chiplet approach as well? Who knows? It would have been a good short-term strategy, I would think. But what will their 10+ nm and stacked dies bring them in the future?

As I said, they'll be using a mix of approaches going forward. They have made this public. Moore's Law is running out of steam, so single integrated dies just won't cut it.

Scali wrote:

Then people argued that they took out HyperThreading because it only worked on the long and inefficient pipelines of the P4. And it didn't. HT reappeared in Nehalem, and despite the short and efficient pipelines, HT still worked at least as well as it ever did on P4.

SMT yields the greatest benefit in architectures that suffer disproportionately from pipeline issues, so long pipelines with a large stall or miss penalty and in-order designs. It still increases the efficiency of other designs (although it's worth noting the Nehalem instruction pipe isn't exactly short compared to, say Penryn).

Scali wrote:

But one thing I did say is that the two approaches aren't mutually exclusive. Which more or less implies that I think that both AMD and Intel may combine the two approaches in the future. So there is not One True Way(tm).

Definitely.

Scali wrote:

What makes you think I don't?
I hate to point out the obvious, but you don't get to be as knowledgeable and experienced as I am if you don't always challenge yourself.

Okay, but that's not a reason for trying to weaponize the BB. If we're able to talk in a reasoned and complete way instead of always angling to destroy everyone else in the thread, I bet we can dump 90% of the drama, be much more productive, and have a way better time.

This very discussion is something you so rarely see on the Internet nowadays - political correctness and all that.

Please carry on. I love it.

-sent from my Threadripper 1950X

gdjacobs wrote:

Those segments (as a complete platform) pretty much require an IGP.

IGP as in integrated on the motherboard? Because those existed for Nehalem/Lynnfield (H57/H55/Q57 chipsets).
Or specifically integrated into the CPU package? I would say that is somewhat arbitrary. Penryn had the IGP in the chipset. Clearly in that segment you'd want to integrate it in the CPU package eventually, but Westmere is not necessarily the make-or-break point. They could have left the IGP in the chipset for another generation.

I think you are putting goalposts in arbitrary positions. I don't like that type of 'discussion'.
I think your view of how well you conduct yourself in a mature, technical discussion, does not quite match with reality.

gdjacobs wrote:

Depends on whether you're interested in scoring points or a real discussion. To me, engineering decisions are interesting.

Engineering decisions are interesting to me as well, but there's this thing about discussions and making statements in a certain context.
I made the statement that Westmere consisted of two dies, like Zen 2.
The engineering decisions behind that choice may have been slightly different, but that is not part of the context in which I made my statement.
I never claimed Intel worked from the same engineering decisions as AMD did. I merely said that Intel has taken the same multi-die approach to a CPU package in the past as well.

And that's where the problem is. You are trying to pull my statement into a different context, and then try to say my argument was 'imperfect'.
Clearly I would object to that, because it's just very unsportsmanlike to use a strawman fallacy in a discussion like that. Again, don't try to shift the blame on me. Look at your own contribution first. You know people are not going to like it if you twist and turn their words and use fallacies to try and 'win' a debate.

gdjacobs wrote:

Sure, but neither was Intel. So what?

I merely responded to someone who claimed that AMD *was* being innovative with their chiplet architecture.
Again, context, it is important.
The "so what" here is that someone made a comment that was incorrect, and I responded with a correction.
That would have been the end of that, but apparently some people, including you, just can't let it end there, and have to keep harassing me about it.

gdjacobs wrote:

Sure, I believe I said so. Most innovations touted in the media are just rewarmed versions of stuff from before.

Then we agree on that. And then you understand why I try to give a more realistic view of developments, by pointing out how things pushed as 'innovations' look eeriliy similar to what others have done in the past.

gdjacobs wrote:

You trumpeted Intel as the prior art for package level integration.

I object to your use of the word "trumpeted".
Why would you phrase it in a subjective way like that?
I gave Westmere as an example, because it was the first that came to mind.
I never claimed Intel was the first, nor that Intel was innovative in doing so.
I merely wanted to debunk the claim that Zen 2 was 'innovative', because Zen 2 clearly was NOT the first.
You understand the concept of falsification? To disprove something, you only need one counter-example. Westmere is one such counterexample. I never implied it was the first, or the only one. I merely disproved Zen 2 being the first. Nothing more. Anything you want to read into it on top of that, is your own bias.

gdjacobs wrote:

Even if you don't believe so, it comes across as very defensive of Intel as the better engineering house.

That is your subjective view.
I am just stating fact that Intel got there before AMD. Intel being "the better engineering house", well, that's your interpretation of it all. I think it says a lot more about you than about me. Think about it.

gdjacobs wrote:

Stacked dies has been quite successful since it debuted with mobile chipsets, but it has physics issues against it. Stacking won't be viable for functional areas with high thermal load, so GPUs and desktop CPU cores are probably out. They will be using some variant of chiplet with EMIB and some stacking, but I'm not sure how far along they are.

I think you are missing the bigger picture here.
"Functional areas with high thermal load" -> "GPUs and desktop CPUs cores"
See the mental leap you made there? You are still thinking of GPUs and CPUs as single dies. You forget to factor in chiplets: You can slice and dice the functional logic of a CPU or GPU into any number of sub-modules. This is a divide-and-conquer approach: You can divide the logic in such a way that there will be entire chiplets that do not contain functional areas with high thermal load. And you can stack those together.

That's also how HBM is implemented by the way: They are put on top of a GPU package. Which by your logic wouldn't work, because you say GPUs are 'functional areas with high thermal load'. Yes, they are... but you can create a 'low thermal load' area at the edges of the GPU package, and stack your HBM there.

gdjacobs wrote:

SMT yields the greatest benefit in architectures that suffer disproportionately from pipeline issues, so long pipelines with a large stall or miss penalty and in-order designs.

This is where your view is again limited. Long pipelines are just one case where SMT helps, they are not the only case.
In-order designs are a special case altogether, as their take on SMT is a very limited one, as they are in-order, which is more or less mutually exclusive with proper SMT.

As said, SMT originated from IBM, and they applied it to their POWER architecture, which is an entirely different beast from a Pentium 4. Sun also used SMT, again with entirely different types of CPUs.

I think what you are missing is that x86 is inherently inefficient. The instructionset dates from the 1970s, and only allows two operands per instruction, where one operand doubles as source and destination.
This inherently causes inefficiencies and requires all sorts of fancy register renaming and recombining logic to extract more ILP.
And that is why SMT works so well on any x86, no matter how efficient it is.

I mean, I hate to point out the obvious, but Nehalem was at its introduction the x86 CPU with the highest IPC ever. It couldn't possibly sustain that IPC if it suffered severely from stalls. There must be more to the story. Which is what I said above. Even the fastest x86 CPUs are inherently inefficient in their execution backend because the x86 instructionset has a lot of dependencies hardwired into the instruction format.

gdjacobs wrote:

(although it's worth noting the Nehalem instruction pipe isn't exactly short compared to, say Penryn)

It is short compared to Pentium 4. Again, context, goalposts... Look at yourself first. You really think you are conducting a good discussion?
Anyway:
Penryn: 12-14 stages
Nehalem: 20-24 stages
Pentium 4: 28-31 stages

What needs to be noted however, is that there's a thing known as a 'Loop Stream Detector', where part of the pipeline can be skipped (similar to Pentium 4's trace cache). Penryn has a simple one, the Nehalem one is more advanced.
So effectively not all 20-24 stages are used during execution, bringing it closer to Penryn. Which makes sense, because otherwise it would never have gotten better IPC than Penryn.

gdjacobs wrote:

Okay, but that's not a reason for trying to weaponize the BB.

Again, there's the moral superiority.
You are making assumptions, and then working from those, while your assumptions are flawed.

gdjacobs wrote:

If we're able to talk in a reasoned and complete way instead of always angling to destroy everyone else in the thread, I bet we can dump 90% of the drama, be much more productive, and have a way better time.

Again, I'm not the problem.
The way I see it, some AMD fanboys were convinced that Zen 2 was the best thing since sliced bread, and chiplets were so awesome and innovative.
Then I pointed out Westmere, and cognitive dissonance ensued. Of course I was a heretic, and needed to be stoned!
The 'best' argument I've found from the other side is "they are not the same"... Yea, that's a reasoned, productive discussion!
Either you have arguments why they're (fundamentally) different, or you accept my arguments why they're remarkably similar (two dies, I/O split off to a die on older manufacturing). There's no other option.

Again, I'm not the problem.

The first step to solve a problem is to admit it. 🤣

It's quite humorous to read this thread. Go on, please. 😁

GigAHerZ wrote:

The first step to solve a problem is to admit it. 🤣

Yes, so the other side has to admit.
As I stated very clearly, I have nothing to admit, the problem simply isn't with me. I have given a very clear description of what the problem is.
Prove otherwise first, and then perhaps you have an argument.
But currently there are simply some people (who as I explained above are part of the problem) who go by the assumption that the problem must be with me (and who are obviously quite biased, as pointed out before), and are then pressuring me into admitting a problem, which I have already stated doesn't exist.

Just because there's more of you, and I'm being singled out, doesn't mean you're right. Funny how you people keep pressuring me into "looking at yourself cricitally", while you clearly are not applying this to your own actions whatsoever.

Scali wrote:

There's no other option.

tbf... Has anyone actually offered up proof or documentation against the points that Scali has raised?.... can't see any?... just whining... obvious trolls are obvious

I like Scalis technical mumbo jumbo, can't follow most of it, but things that I know to be true through my own technical experience have agreed with things Scali has said... so I'm inclined to believe him... and he's being doing it a looong time (I followed some of his articles in flipcode like 18 yrs ago or what ever... if anyone here knows what that is 😀).. interwebs is full of enough BS these days and unlike many others here, Scali actually has experience and so tbh I value his opinion more than many. I mean, how many microprocessor designers or people with actual technical experience working in the microprocessor (or even other hardware/software) industry do we have in this thread?... me thinks not many, especially the loudest.

AMD give more bang for buck certainly, but they have had failings in other areas and while they are cheaper, I think this resonates with home users... aka gamers... i.e to price/performance ratio. Industry uses Intel/nV, and they have for many years (most of Intels market dominance in the entire CPU market can probably be attributed to this). In my professional experience (15 yrs CAD developer) I have had very little dealings with AMD hardware, corporate users just don't have it... be that due to other factors like certain maintenance contracts, or even stability etc... but upshot is Intel have dominance in certain corporate industries. tbh I've had to deal with many more issues/support tickers dealing with the little AMD hardware that I have been exposed to at work, than with Intel/nV hardware that most (if not all) our users have. Take from that what you will.

At home I would probably get AMD (cheap bastard I be), and probably be perfectly happy and get ooodles of performance that a tiny amount of the games I play would require. But at work I would be surprised if I saw some AMD hardware on a desk somewhere.

Since AMD appeals to gamers (what I would classify as home users, unless you are a professional gamer of course), I can understand the fan-boi-ism on a site like this. Intel/nV just always seem to be that one step ahead with a few things, and coupled with some uber issues that have happened with AMD, I can understand while business keep with Intel/nV... unless they are really short for cash... in which case they probably don't need to upgrade anyway.

Disclaimer: Had AMD hardware from '97 -> '06, from K6 -> Athlon64, from Rage128 Pro to... ok maybe I went nV GPU earlier on... but I liked my Rage128 and even some Radeons and never had major issues with my AMD CPU's... Would like to get a Ryzen if I could think of something to justify that amount of CPU power in the home.. we'll see.

If money was no object, I would get Intel/nV…. coz... slightly better benchies, and business use them… I presume for a reason. Is that such a crime, or do we have to lurve the underdog... coz underdog...

/2cents

spiroyster wrote:

I like Scalis technical mumbo jumbo, can't follow most of it, but things that I know to be true through my own technical experience have agreed with things Scali has said... so I'm inclined to believe him... and he's being doing it a looong time (I followed some of his articles in flipcode like 18 yrs ago or what ever... if anyone here knows what that is 😀).. interwebs is full of enough BS these days and unlike many others here, Scali actually has experience and so tbh I value his opinion more than many. I mean, how many microprocessor designers or people with actual technical experience working in the microprocessor (or even other hardware/software) industry do we have in this thread?... me thinks not many, especially the loudest.

Totally agree. I just can't stomach the constant barrage of attacks on anyone and everyone else. It's childish, as is the insistence that "if I have to stop then they have to stop." It's a strategy that is proven to lead to a downward spiral. (See: https://www.wnycstudios.org/podcasts/radiolab … eserves-another)

This isn't to single one person out, but over the course of 5 pages, several people have come in, got into a tangle, then left. One constant remains. At some point, you have to acknowledge that if "everyone else" is always doing something wrong, maybe you're the problem.

SirNickity wrote:

Totally agree. I just can't stomach the constant barrage of attacks on anyone and everyone else.

Again, you are BIASED!
You only see my reactions to other people, but completely fail to detect how these people (including yourself) are attacking me first.

SirNickity wrote:

as is the insistence that "if I have to stop then they have to stop." It's a strategy that is proven to lead to a downward spiral. (See: https://www.wnycstudios.org/podcasts/radiolab … eserves-another)

That is completely NOT what I said earlier.

SirNickity wrote:

This isn't to single one person out, but over the course of 5 pages, several people have come in, got into a tangle, then left.

If only they had left. They (which also means you) kept trolling and provoking.

SirNickity wrote:

At some point, you have to acknowledge that if "everyone else" is always doing something wrong, maybe you're the problem.

Are you deliberately gaslighting me? Sounds like it.
As I said before, it's not "everyone else". It's a few people, and they're all in the pro-AMD camp, apparently (and just look at that first line in spiroyster's post that you didn't quote... what does that say? Exactly. That was directed at you as well, in case you missed that. And at least to me it makes it obvious which side of the debate should "walk away", because that's generally what you do when you don't have any more points to argue).
I mean, just look at gdjacobs. He thinks he's pretty smart, trying to paint me as some pro-Intel guy with his subtle twisting and turning.
But you can read the bias through his words. He's not smart enough to actually pull off what he's trying to pull off. And people tend to get angry when they get confronted with that, their ego hurts.
And you, weren't you one of those idiots in the earlier secure-boot debate? You probably still have a chip on your shoulder from that. It's not easy looking at yourself and judging your own behaviour, actions, and consider the consequences.
I mean, I don't expect that you and I will ever get along. But unlike you, I'm not actively going to provoke and troll you in every thread. Again, you, and people like you, are the problem. People who can't let go.
You're an idiot. Your conduct has been completely obnoxious on various occasions, yet you feel surprised when someone (in this case me) tells you off about it. Then suddenly it's 'an attack' and 'gee where could that possibly come from' and 'gee you have a problem'. Really, grow up. If you act obnoxious, that's what you get.
If you don't have any more points to argue, you walk away. Instead, you, and various people like you, start piling down on me personally. Trying to beat me into submission because they can't admit they don't have any more points, and basically lost the debate.
I mean, how did this go from "Westmere has two dies in the package" to "Scali must have a problem because we keep attacking him"? It isn't even remotely on-topic, and is pure ad-hominem. And either you actually don't even see that you're doing it, or you're just a sad manipulative a-hole because you're a poor loser.

The irony of it all is that I'm not even pro-Intel to begin with. I started my computer life with Z80s, 6502s and 68000s. I have very little with Intel or x86 in general. I think it's a pretty horrible architecture, and Intel and others should have stopped flogging that dead horse decades ago.

Scali wrote:

Again, you are BIASED!
You only see my reactions to other people, but completely fail to detect how these people (including yourself) are attacking me first.

I don't just see your reactions. But I see that, over and over again, a conversation about something that doesn't matter at all ends up in a flame war, and the one dependable constant is that, at some point, you turn from facts and figures to aggressive commentary about the person you're debating. As soon as you get personal, the conversation dies and it devolves. There's no reason adults should act like that. You're not an adolescent anymore, knock it off with the personal attacks. That goes for anyone else here, too, not just you.

Scali wrote:

If only they had left. They (which also means you) kept trolling and provoking.

Why's it always have to be someone else, Scali? Why can't you be the bigger man? That's the only reason I'm making this about you. Because it always involves you. I could go after the other dozens of people you brawl with, but it takes two to tango, and you're always that other guy.

Scali wrote:

As I said before, it's not "everyone else". It's a few people, and they're all in the pro-AMD camp, apparently

This ceased to be about processors when people started making it personal. Who the --- cares about AMD vs. Intel?

Scali wrote:

And you, weren't you one of those idiots in the earlier secure-boot debate?

Yep.

Scali wrote:

You probably still have a chip on your shoulder from that.

Nope.

Scali wrote:

I mean, I don't expect that you and I will ever get along.

Why not? I like you just fine until you go slinging mud and turning a civil conversation into a brawl. I'm just asking for you to please calm the crap down. Aside from that, I would absolutely buy you a beer.

Scali wrote:

You're an idiot. Your conduct has been completely obnoxious on various occasions, yet you feel surprised when someone (in this case me) tells you off about it.

True. Probably true. Not surprised.

Scali wrote:

If you don't have any more points to argue, you walk away.

If I run out of points in a debate, either I've said my piece, or I'm wrong. The latter happens, and I'm 100% OK with that. I disagree with you about the Secure Boot thing, but that's not why I left. I walked away because you pissed in the Cheerios of that conversation, it wasn't fun or enlightening anymore, and the argument was inherently one of perception, so if no minds were changed by that point, no minds were going to be changed. There's no sense in dragging it on after that.

OK. That's all I have to say about this. I love ya man, but you need to grow up a little. Once again, please feel free to get that last word in, if it makes you feel better. Just please, when you cool down, look back at the way you handle arguments and .. regardless of all of our other faults and personality issues ... can YOU do better? I think you can, and that's really all YOU can control, right? Everyone else is their own problem.

There's no point playing chess with a pigeon - it knocks the pieces over, craps on the board, and flies back to its flock to claim victory. 😀
(Paraphrased quote from Scott D. Weitzenhoffer)

Scali wrote:

Because those existed for Nehalem/Lynnfield (H57/H55/Q57 chipsets).

There was no GMA in the H55+ chipsets so Lynnfield with H55 had no IGP.

Scali wrote:

And that's where the problem is. You are trying to pull my statement into a different context, and then try to say my argument was 'imperfect'.
Clearly I would object to that, because it's just very unsportsmanlike to use a strawman fallacy in a discussion like that. Again, don't try to shift the blame on me. Look at your own contribution first. You know people are not going to like it if you twist and turn their words and use fallacies to try and 'win' a debate.

Maybe that's a mistake on your part. I'm not trying to "win" a debate. I just want a stimulating discussion. If you don't like the direction of the conversation, just say you weren't really going there and move on.

Scali wrote:

I merely responded to someone who claimed that AMD *was* being innovative with their chiplet architecture. Again, context, it is […]
Show full quote

I merely responded to someone who claimed that AMD *was* being innovative with their chiplet architecture.
Again, context, it is important.
The "so what" here is that someone made a comment that was incorrect, and I responded with a correction.
That would have been the end of that, but apparently some people, including you, just can't let it end there, and have to keep harassing me about it.

It's less clear cut than you make out. You and I may agree that these clever engineering solutions aren't really that innovative, but others think they are. Instead of getting caught up in a fight about word choice, we can focus on the engineering that's behind the product and the context in which it's happening. I think it'll elevate the discussion.

Scali wrote:

Then we agree on that. And then you understand why I try to give a more realistic view of developments, by pointing out how things pushed as 'innovations' look eeriliy similar to what others have done in the past.

Sure, you and I think that way, but I could also see how it could be interpreted the other way given the immediate context. The previous few generations of chip have all been integrated at the die and AMD deliberately stepped back from that. It seems like a stupid hill to fight a battle over.

Scali wrote:

I object to your use of the word "trumpeted".

Perhaps not trumpeted, but deliberately or not, your response was definitely constructed in a very one sided fashion about an issue (the choice of a word in some marketing) that's about as one sided as it is important.

Scali wrote:

That is your subjective view.

Not really. You've voiced opinions about the AMD64 instruction set in the same way previously while extolling EPIC as somehow being better (despite failing to reach it's engineering or business objectives). I'm sure I could find other data points.

In my opinion, the EPIC design had superior ambition and that's about it. Other architectures and architectural evolutions were much more interesting (the Eclipse MV from Soul of a New Machine, for instance), but AMD64 was clearly the best path forward for x86.

Scali wrote:

See the mental leap you made there? You are still thinking of GPUs and CPUs as single dies. You forget to factor in chiplets: You can slice and dice the functional logic of a CPU or GPU into any number of sub-modules. This is a divide-and-conquer approach: You can divide the logic in such a way that there will be entire chiplets that do not contain functional areas with high thermal load. And you can stack those together.

No, I was talking about cores in the Intel core/uncore sense.

Scali wrote:

That's also how HBM is implemented by the way: They are put on top of a GPU package. Which by your logic wouldn't work, because you say GPUs are 'functional areas with high thermal load'. Yes, they are... but you can create a 'low thermal load' area at the edges of the GPU package, and stack your HBM there.

I can go out tomorrow and buy a desktop GPU card with HBM. None of it will be stacked on the GPU. The DRAM modules are stacked on each other, Captain Crunch (John Draper) style.
http://www.willegal.net/blog/?p=7246

Scali wrote:

This is where your view is again limited. Long pipelines are just one case where SMT helps, they are not the only case.
In-order designs are a special case altogether, as their take on SMT is a very limited one, as they are in-order, which is more or less mutually exclusive with proper SMT.

This is just plain wrong. How is the SMT used in the Power 6 or Fujitsu SPARC VII not proper?

Scali wrote:

Sun also used SMT, again with entirely different types of CPUs.

No they didn't. The T1 and T2 used multicore and round robin vertical thread dispatch. Oracle has used SMT on some of their later UltraSPARC designs, as has Fujitsu.

Scali wrote:

I think what you are missing is that x86 is inherently inefficient. The instructionset dates from the 1970s, and only allows two operands per instruction, where one operand doubles as source and destination.
This inherently causes inefficiencies and requires all sorts of fancy register renaming and recombining logic to extract more ILP.
And that is why SMT works so well on any x86, no matter how efficient it is.

It works pretty good in RISC designs where ILP was an explicit design objective.

Scali wrote:

I mean, I hate to point out the obvious, but Nehalem was at its introduction the x86 CPU with the highest IPC ever. It couldn't possibly sustain that IPC if it suffered severely from stalls. There must be more to the story. Which is what I said above. Even the fastest x86 CPUs are inherently inefficient in their execution backend because the x86 instructionset has a lot of dependencies hardwired into the instruction format.

This isn't a limitation of x86 so much as a limitation of multiple issue CPUs in general. All the register renaming, speculative execution, etc. still isn't enough to keep a modern CPU's execution units fully stuffed.

Scali wrote:

It is short compared to Pentium 4. Again, context, goalposts... Look at yourself first. You really think you are conducting a go […]
Show full quote

It is short compared to Pentium 4. Again, context, goalposts... Look at yourself first. You really think you are conducting a good discussion?
Anyway:
Penryn: 12-14 stages
Nehalem: 20-24 stages
Pentium 4: 28-31 stages

What needs to be noted however, is that there's a thing known as a 'Loop Stream Detector', where part of the pipeline can be skipped (similar to Pentium 4's trace cache). Penryn has a simple one, the Nehalem one is more advanced.
So effectively not all 20-24 stages are used during execution, bringing it closer to Penryn. Which makes sense, because otherwise it would never have gotten better IPC than Penryn.

You know as well as I that IPC does not necessarily scale with pipeline depth. Intel did a lot of work on BPU and speculative execution with Netburst and they were able to bring it forward into their newer designs and mitigate a lot of the penalties of a longer pipeline. Presumably they felt there were enough execution slots which would be idle (even with improved single thread ILP) to make the transistor cost of implementing SMT worthwhile. i.e. Given two options, they chose both.

Scali wrote:

Again, there's the moral superiority.
You are making assumptions, and then working from those, while your assumptions are flawed.

There's the righteous indignation.

Scali wrote:

Again, I'm not the problem. The way I see it, some AMD fanboys were convinced that Zen 2 was the best thing since sliced bread, […]
Show full quote

Again, I'm not the problem.
The way I see it, some AMD fanboys were convinced that Zen 2 was the best thing since sliced bread, and chiplets were so awesome and innovative.
Then I pointed out Westmere, and cognitive dissonance ensued. Of course I was a heretic, and needed to be stoned!
The 'best' argument I've found from the other side is "they are not the same"... Yea, that's a reasoned, productive discussion!
Either you have arguments why they're (fundamentally) different, or you accept my arguments why they're remarkably similar (two dies, I/O split off to a die on older manufacturing). There's no other option.

I'm willing to extend people the courtesy of hearing why they hold to these opinions. What's their logic? What's the context? Even if their logic falls apart, it's possible (and refreshing) to set things straight while not being a total dick.

Have you ever asked why threads have a tendency to go toxic when you're involved in them? I can assure you, it's not because a pack of AMD Fanbois are following you around.

Time to end this, because some people apparently don't know when to walk away, and keep pounding on, gaslighting, framing me etc. You keep throwing walls of text at me for no reason other than that you're butthurt AMD fanbois or whatever your exact problem is.
I don't waste any more time trying to respond to your lies.
To make it absolutely obvious: you can repeat these lies as often as you want, but that doesn't make them true. And the fact that I will stop responding to them now does not mean I agree with them. Nothing has changed.