Case Proven: People that think X2 > Core2 clock for clock

[imported_Crusader]

Originally posted by: zsdersw

Originally posted by: Crusader
I left the insults intact to shine poorly on you guys, not on myself.
When you grow up, you'll understand. I'm not offended.

In response to that cherry picked statement out of my post (while everythign else was left alone, which tells me you just want to argue and are annoyed someone sticks up for the merits of AMD tech)-

I'm sorry you're not offended. Better luck next time, I suppose.

I picked out that statement of yours because it's the only thing that I disagreed with.

I'm not annoyed that you stick up for "the merits of AMD tech". I'm annoyed that you simultaneously criticize, wrongly, Intel tech by making idiotic statements about "quality engineering", etc.

Hey, like I said if Netburst is your idea of quality engineering.. more power to you.
I'm glad I can annoy and offend you. Its like watching a child who pissed their diaper communicating with you. So yeah, either a child, or a shill.

Originally posted by: zsdersw

Originally posted by: CrusaderYes, its a small part of the market. But its the MAJORITY of the market looking at early adoption of C2D.
If we took a poll of who earns their living entirely off their desktop PC where shaving off a few seconds on some arcane benchmark could possibly matter, the results would be rather slim.
When the majority of people here use their PCs for games, forums and basic Windows apps..maybe use Wildfire + some programming apps.. the upgrade doesnt make sense.
In gaming C2D=A64 when you are GPU limited.

Are you seriously disputing this? Because you argued with me, calling me names.. you know its the fact of the matter.. but you still want to argue? :roll:

Since when does the "early adopters of C2D" market have anything to do with the broader discussion about how one CPU stacks up to another? It is but one element in the discussion.. and not a terribly conclusive one at that.

As far as the name-calling is concerned, when it looks like a duck and sounds like a duck, chances are it's a duck. I call 'em as I see 'em. I'm not going to apologize for that.. and if you or anyone else has a problem with it, T.S.

LOL.. what a childish tool. TS! :disgust: My god. Fvckin kids.
Erm.. because people buying C2D still TODAY are in reality of the market place, early adopters of C2D? Small market? In the grand scheme.. yes.
There are plenty of "terribly" conclusive dictations we can make from the C2D launch till this point. If you could wrap your peabrain around more than defending Intel you might be worthy of speaking to.

 

[MercenaryForHire]

Whee, multiple-thread pileup!

- M4H

 

[StrangerGuy]

Back to the original topic...

It doesn't matter how you guys argue about it, it is a cold hard FACT that the C2D is more efficient than A64 X2 clock-for-clock. Only the ignorant and AMD fanboys will say its not.

 

[zsdersw]

Originally posted by: Crusader
Hey, like I said if Netburst is your idea of quality engineering.. more power to you.
I'm glad I can annoy and offend you. Its like watching a child who pissed their diaper communicating with you. So yeah, either a child, or a shill.

LOL.. what a childish tool. TS! :disgust: My god. Fvckin kids.
Erm.. because people buying C2D still TODAY are in reality of the market place, early adopters of C2D? Small market? In the grand scheme.. yes.
There are plenty of "terribly" conclusive dictations we can make from the C2D launch till this point. If you could wrap your peabrain around more than defending Intel you might be worthy of speaking to.

Netburst was engineered perfectly to suit its design goals. It was geared for high clock speed.. and it did just that. Additionally, computing tasks were optimized to take advantage of these perfectly engineered design goals and, as a result, those tasks excelled on Netburst chips. The fact that Netburst was inferior to its competition in many other tasks does not mean that Netburst was "poor quality engineering".. by any serious standard of measuring "engineering".

For someone who earlier decried name-calling, you certainly do plenty of it yourself. Congratulations.. you're a hypocrite in addition to being an idiot.

 

[Munky]

Netburst was engineered perfectly to suit its design goals... Additionally, computing tasks were optimized to take advantage of these perfectly engineered design goals

Never have I seen the abomination known as "Netburst" and "perfection" mentioned so many times in relation to each other :roll:

The fact that Netburst was inferior to its competition in many other tasks does not mean that Netburst was "poor quality engineering"

The Netburst design goals were wrong. Same as the nv30 design goals were wrong. In the end, both products were inferior, and whatever measure of "engineering quality" you use is just rhetoric.

 

[zsdersw]

Oh please. Netburst was not an abomination. It's competition did better on most of the benchmarks, but it was hardly anything close to what you're characterizing it as. It's the same thing with C2D versus K8. Just because C2D is better on most of the benchmarks doesn't make K8 an "abomination".

 

[imported_Crusader]

Originally posted by: zsdersw

Originally posted by: Crusader
Hey, like I said if Netburst is your idea of quality engineering.. more power to you.
I'm glad I can annoy and offend you. Its like watching a child who pissed their diaper communicating with you. So yeah, either a child, or a shill.

LOL.. what a childish tool. TS! :disgust: My god. Fvckin kids.
Erm.. because people buying C2D still TODAY are in reality of the market place, early adopters of C2D? Small market? In the grand scheme.. yes.
There are plenty of "terribly" conclusive dictations we can make from the C2D launch till this point. If you could wrap your peabrain around more than defending Intel you might be worthy of speaking to.

Netburst was engineered perfectly to suit its design goals. It was geared for high clock speed.. and it did just that. Additionally, computing tasks were optimized to take advantage of these perfectly engineered design goals and, as a result, those tasks excelled on Netburst chips. The fact that Netburst was inferior to its competition in many other tasks does not mean that Netburst was "poor quality engineering".. by any serious standard of measuring "engineering".

Wow.. LOL! This tops it all, sure thing town fool! Guys like you crack me up, and you need your sh*t eating grin smacked off your pompous face.
You wouldnt know what the hell was going on if you were hit upside the face by a clue-by-4.
Netburst had "perfectly engineered design goals"?
I'll let you in on something noobcake- netburst was originally supposed to scale to 10ghz.
Netburst, 3.8ghz FTW! :thumbsup:

For someone who earlier decried name-calling, you certainly do plenty of it yourself. Congratulations.. you're a hypocrite in addition to being an idiot.

When dealing with an immature punk such as yourself, sometimes the only thing to get through to a ho is a bitch smack. Dont dish it out if you cant handle it, boy.

 

[Viditor]

Originally posted by: zsdersw

Originally posted by: Viditor
I have a feeling that you don't understand the difference between memory latency and bandwidth...

Memory Latency = The time between initiating a request for a character in memory until it is retrieved
Bandwidth = the maximum amount of information (bits/second) that can be transmitted along a channel

While it's true that even the K8 has more than enough Bandwidth, it's performance is very sensitive to Latency (for example DDR400 is about equivalent to DDR2-667 in performance because the latency of DDR2 is much higher).
Memory performance is a combination of bandwidth and latency...

So, are you saying that you don't believe that the latency is much lower with an ODMC?

No, that's not what I'm saying at all.. and yes, I know the difference between latency and bandwidth. I'm annoyed that you presumed I didn't know it.

I did not intend to annoy you...it's just that every time I state that the biggest advantages of HT and ODMC are the reduced latency, you say that bandwidth isn't important. I was pointing out that the two are quite different.

Why did the switch from DDR to DDR2 not benefit the K8 very much? Because of its higher latency. What matters most in memory latency? The timings. What do the timings depend on? How the memory was designed and what the memory standard is capable of.

The timings are only half of the story...length of the traces and signal path are the other half and are equally important...

In spite of the lower memory bandwidth and presumed higher latency of the C2D with its FSB, it soundly beats K8 in nearly every test. What does that say about an IMC? It says that CPU architectures can be designed to raise the performance bar without an IMC. Would Core 2 Duo be better off with an IMC? I doubt it. It's performance doesn't depend on memory latency or bandwidth as much as competing chips and the FSB is far from being an Achilles heel for it in most market segments.

Of COURSE C2D would be IMMENSELY better with an ODMC (I'm not sure why you think that fetching memory faster wouldn't help...)! What the C2D core shows is that there are OTHER things besides latency that matter as well...for example,

1. Out-of-order loads for memory (intel calls this "memory disambiguation")
2. Processing 128-bit SSE3 instructions without slowing down
3. More advanced branch prediction unit and larger command buffers
4. Micro-ops fusion technology

The thing is that K8L (unlike K8) will have a lot of this as well, so all other factors being equal the reduced latency in I/O and memory whould make it faster than C2D.
The one thing that K8L won't have is the ability to decode and execute 4 commands per clock cycle...however I have yet to find a normal situation where this is actually used (in fact 3 commands/clock cycle is fairly rare...).

All of that said, keep in mind that we only have theory for K8L for now...whether it actually has these features or how well they work we won't know till the end of Q1 07 (when samples start coming out).

 

[Markfw]

Originally posted by: StrangerGuy
Back to the original topic...

It doesn't matter how you guys argue about it, it is a cold hard FACT that the C2D is more efficient than A64 X2 clock-for-clock. Only the ignorant and AMD fanboys will say its not.

That may be true iun many cases, but a C2D in a cheap motherboard ($50) vs an X2 in a cheap ($50) motherboard ?? In my experience, they are the same speed at the same clock !. Its only one case, and bears further examination, but I find it interesting.....

 

[imported_Crusader]

Originally posted by: Viditor

Originally posted by: zsdersw

Originally posted by: Viditor
I have a feeling that you don't understand the difference between memory latency and bandwidth...

Memory Latency = The time between initiating a request for a character in memory until it is retrieved
Bandwidth = the maximum amount of information (bits/second) that can be transmitted along a channel

While it's true that even the K8 has more than enough Bandwidth, it's performance is very sensitive to Latency (for example DDR400 is about equivalent to DDR2-667 in performance because the latency of DDR2 is much higher).
Memory performance is a combination of bandwidth and latency...

So, are you saying that you don't believe that the latency is much lower with an ODMC?

No, that's not what I'm saying at all.. and yes, I know the difference between latency and bandwidth. I'm annoyed that you presumed I didn't know it.

I did not intend to annoy you...it's just that every time I state that the biggest advantages of HT and ODMC are the reduced latency, you say that bandwidth isn't important. I was pointing out that the two are quite different.

Why did the switch from DDR to DDR2 not benefit the K8 very much? Because of its higher latency. What matters most in memory latency? The timings. What do the timings depend on? How the memory was designed and what the memory standard is capable of.

The timings are only half of the story...length of the traces and signal path are the other half and are equally important...

In spite of the lower memory bandwidth and presumed higher latency of the C2D with its FSB, it soundly beats K8 in nearly every test. What does that say about an IMC? It says that CPU architectures can be designed to raise the performance bar without an IMC. Would Core 2 Duo be better off with an IMC? I doubt it. It's performance doesn't depend on memory latency or bandwidth as much as competing chips and the FSB is far from being an Achilles heel for it in most market segments.

Of COURSE C2D would be IMMENSELY better with an ODMC (I'm not sure why you think that fetching memory faster wouldn't help...)!

Yeah, thats whats amazing here is the opposition to my idea that an IMC is superior to Intels methods.
I'm guessing zsdersw and dmens are either paid shills.. or just kids that dont have absolutely any idea what they are talking about.
Hey, its from Intel. It must be superior! So was Netburst.. oh, wait

I'm guessing when my FX55 is to slow for me, K8L will be here and we'll get the real C2D chips that push their potential a bit more out to combat K8L

 

[dmens]

Of COURSE C2D would be IMMENSELY better with an ODMC

out of all people, you should understand the concept of return on investment. On 1P/2P, the negligible speedup is not worth the large amount of design and validation effort required. not to mention there's potential drawbacks to having a controller attached all the time... it's called real engineering decisions, something crusader would never understand!

oh, as for netburst and its design goal errors, none of you are remotely qualified to comment, so don't even try.

 

[harpoon84]

You all fail to raise the fact C2D was designed *without* an IMC in mind - hence the hefty L2 cache sizes. If C2D was designed from the ground up to work with an IMC, the L2 cache would be nowhere as big.

Intel had a decision to make regarding C2D - go for a big cache, or go for an IMC. Obviously they couldn't do both, either due to financial feasibility, technological reasons or a bit of both. Intel chose the large cache method, and so far it's paying fine dividends. C2D is a fine modern CPU and soundly beats AMD's current offerings.

So all this talk about IMC vs FSB is useless because it's all theoretical and doesn't take into account the vastly different design methods employed by AMD and Intel.

Sure, AMD's IMC is more efficient than Intel's FSB, but Intel CPUs have 4x the L2 cache of AMD CPUs.

In the end, CPU performance is determined by the sum of it's parts, and currently Intel has more weapons in it's arsenal than AMD's ageing K8 design.

 

[dexvx]

Originally posted by: Markfw900

Originally posted by: StrangerGuy
Back to the original topic...

It doesn't matter how you guys argue about it, it is a cold hard FACT that the C2D is more efficient than A64 X2 clock-for-clock. Only the ignorant and AMD fanboys will say its not.

That may be true iun many cases, but a C2D in a cheap motherboard ($50) vs an X2 in a cheap ($50) motherboard ?? In my experience, they are the same speed at the same clock !. Its only one case, and bears further examination, but I find it interesting.....

I find that hard to believe when AT has tested the cheap $60 Asrock VSTA board:

http://www.anandtech.com/mb/showdoc.aspx?i=2797&p=13

And found out that it still performs within 10% of the 975/965 boards at stock speeds.

 

[Markfw]

Originally posted by: dexvx

Originally posted by: Markfw900

Originally posted by: StrangerGuy
Back to the original topic...

It doesn't matter how you guys argue about it, it is a cold hard FACT that the C2D is more efficient than A64 X2 clock-for-clock. Only the ignorant and AMD fanboys will say its not.

That may be true iun many cases, but a C2D in a cheap motherboard ($50) vs an X2 in a cheap ($50) motherboard ?? In my experience, they are the same speed at the same clock !. Its only one case, and bears further examination, but I find it interesting.....

I find that hard to believe when AT has tested the cheap $60 Asrock VSTA board:

http://www.anandtech.com/mb/showdoc.aspx?i=2797&p=13

And found out that it still performs within 10% of the 975/965 boards at stock speeds.

Thats just gaming...., and I am comparing C2D to X2, not to another C2D.

 

[theteamaqua]

i cant believe after soo many benches peopel still believe X2 is faster than Conroe ... fine go live in ur dream world whgile i enjoy higher fps in games.

i think people who said that X2 is faster are either:
1. AMD workers
2. AMD fanboyz
3. or people who have connection with AMD and try to damage control

my E6400 at 3.37GHz, show me an AMD CPU thats faster than that after OC on air cooling.

 

[dmens]

i think people who said that X2 is faster are either:
1. AMD workers

please dont lump amd engineers with the retards.

 

[Viditor]

Originally posted by: dmens

Of COURSE C2D would be IMMENSELY better with an ODMC

out of all people, you should understand the concept of return on investment. On 1P/2P, the negligible speedup is not worth the large amount of design and validation effort required. not to mention there's potential drawbacks to having a controller attached all the time... it's called real engineering decisions, something crusader would never understand!

oh, as for netburst and its design goal errors, none of you are remotely qualified to comment, so don't even try.

ROI is a perfectly valid point, but I wouldn't characterize the latency benefits of ODMC and HT as "negligible"...
I certainly do understand why Intel didn't go that route initially...for example it would have produced some serious problems with RDRAM (which was their gameplan) and would also be more difficult to implement on the Netburst architecture.
However, to say that it would have little to no effect on a C2D design just isn't right.

In addition, just getting the C2D chips designed and out as quickly as they did was a modern miracle on Intel's part (one for which their engineers should be majorly congratulated)! I strongly doubt that even Intel had the resources to completely redesign the memory as well (IIRC, they pulled the Whitefield team in order to accomplish it).

So yes, I agree that Intel probably made right engineering decision based on ROI (remember that CSI was originally supposed to be out this year), but that doesn't mean that the returns are close to negligible...they just aren't as good as the other improvements they made.

As to Netburst, my own impression is that it was a decision made by the marketing division more than engineering. It was made at a time when RDRAM was the presumptive memory standard for Intel (and with which it would have done quite nicely), and before Intel knew how great a loss there would be from leakage as frequencies kept increasing.

The founder of AMD (J Sanders III) had a great line to cover this...

"The semiconductor industry is like a strange form of Russian Roulette, you pull the trigger on a design and 3 years later you find out if you've blown your brains out or not."

 

[zsdersw]

Originally posted by: Crusader
Wow.. LOL! This tops it all, sure thing town fool! Guys like you crack me up, and you need your sh*t eating grin smacked off your pompous face.
You wouldnt know what the hell was going on if you were hit upside the face by a clue-by-4.
Netburst had "perfectly engineered design goals"?
I'll let you in on something noobcake- netburst was originally supposed to scale to 10ghz.
Netburst, 3.8ghz FTW! :thumbsup:

Sorry, I can't be the town fool. OcHungry has that job on your days off.

Netburst scaled fine given the limitations of the manufacturing processes it was made with.

When dealing with an immature punk such as yourself, sometimes the only thing to get through to a ho is a bitch smack. Dont dish it out if you cant handle it, boy.

I can handle it just fine.. I just find it funny that someone who decries name-calling can be so hypocritical about it.

 

[zsdersw]

Originally posted by: Crusader
I'm guessing zsdersw and dmens are either paid shills.. or just kids that dont have absolutely any idea what they are talking about.

Neither.

 

[zsdersw]

Originally posted by: Viditor
Of COURSE C2D would be IMMENSELY better with an ODMC (I'm not sure why you think that fetching memory faster wouldn't help...)!

What would you say the speed bump for C2D would be with an IMC versus what it has now in a desktop or 1P/2P server? I'm guessing it would be similar to the bump you'd get using lower latency DDR2. The end result? Nothing close to "IMMENSELY better".

 

[Viditor]

Originally posted by: zsdersw

Originally posted by: Viditor
Of COURSE C2D would be IMMENSELY better with an ODMC (I'm not sure why you think that fetching memory faster wouldn't help...)!

What would you say the speed bump for C2D would be with an IMC versus what it has now in a desktop or 1P/2P server? I'm guessing it would be similar to the bump you'd get using lower latency DDR2. The end result? Nothing close to "IMMENSELY better".

I don't think you get it here Z...
The ODMC of the K8 decreases the # of clock cycles burned in memory read/write by ~35-40%.
Anandtech article
So, no...it's not similar to using lower latency ram.

 

[stevty2889]

Originally posted by: zsdersw


Netburst scaled fine given the limitations of the manufacturing processes it was made with.

It wasn't the process..Dothan was made on the same process and didn't have any of the netburst issues..Netburst is dead and for a good reason, it couldn't reach the potential it was intended to reach.

 

[Viditor]

Originally posted by: stevty2889

Originally posted by: zsdersw


Netburst scaled fine given the limitations of the manufacturing processes it was made with.

It wasn't the process..Dothan was made on the same process and didn't have any of the netburst issues..Netburst is dead and for a good reason, it couldn't reach the potential it was intended to reach.

I agree stevty...but to be fair, if Intel had found a way to decrease the leakage significantly (which nobobdy can do yet) then Netburst would still be a most viable architecture.
Dothan changed the architecture to match the limitations of our current abilities in manufacturing.

 

[zsdersw]

Originally posted by: Viditor
I don't think you get it here Z...
The ODMC of the K8 decreases the # of clock cycles burned in memory read/write by ~35-40%.
Anandtech article
So, no...it's not similar to using lower latency ram.

Perhaps it's you who doesn't get it. We're talking about the impact on C2D.. not K8.

 

[Zap]

In before the lock.

Gosh, a lot of name calling around here - people sure do get worked up and emotional over some computer chips.

Originally posted by: zsdersw
Oh please. Netburst was not an abomination. It's competition did better on most of the benchmarks, but it was hardly anything close to what you're characterizing it as. It's the same thing with C2D versus K8. Just because C2D is better on most of the benchmarks doesn't make K8 an "abomination".

Wow, a voice of reason in this Wild West thread.

Originally posted by: Viditor
The founder of AMD (J Sanders III) had a great line to cover this...

"The semiconductor industry is like a strange form of Russian Roulette, you pull the trigger on a design and 3 years later you find out if you've blown your brains out or not."

That's a cool quote.

Some random thoughts...

I do think it funny that people are getting so worked up. I'm probably wrong, but in my mind I imagine the C2D fanboys (ya know, the ones starting threads with titles such as "OMFG AMD IS SO DEAD SUXXORS") having parties and rejoicing after being on the receiving end when A64 was "better" than Netburst. Just as people now still buy AMD chips which overlap the C2D in price (not talking about cheap single core which IMO AMD is still ahead - where's the $80 Core 2 Solo for desktop, Intel?) even though C2D is perhaps overall "better," there were people purchasing 90nm Netburst chips when A64 was perhaps overall "better."

Heh, I guess people just have to find something in their life to be passionate over. Safer than drugs, and cheaper than women. :laugh:

As for myself, still looking forward to a $250 or less overclockable mATX motherboard and Core 2 Duo 4300 combo as a candidate for my gaming rig. Until I can "bring balance to the force," I'll continue to be happy with my 5 AMD and 3 Intel setups.