Intels Core New Low

[TraumaRN]

http://www.firingsquad.com/features/intel_core_new_low/


Posted on FS today, good read about what kind of company Intel has turned into, and while I'm not going to discuss major points, I do believe the author is correct in saying that Intel is just reacting to AMD, they arent innovating as much as in the past.

Your thoughts? Discuss but keep lets keep it clean and not petty crap like 'AMD pwns you INteL egg fryers.'

 

[DeathReborn]

Intel got lazy, it's that simple. They tend to try and lock themselves into cycles of new CPU needs new Motherboard, RAM or GPU. AMD was more sensible, letting others design the Chipsets threfore keeping to a known platfrom for a long time (Socket A, Socket 939). Intel's methodology is to stick the custome with as many upgrades as possible to make more and more money, AMD try and go for stability with current platforms & technology.

Now Intel are trying to keep shrinking their CPU Die Size to keep up with AMD who have made some great moves like working with IBM on Processor Technologies.

 

[openwheelformula1]

I do believe that Intel is trying hard to innovate. They've just yet to find anything significant to counter AMD's on-die memory controller architecture.

I agree that lately it seems we need to upgrade chipsets just to upgrade Intel CPUs. It's not like the newer chipsets are sigfinicant upgrades either.

 

[Maximilian]

Meh... was probably the case before they announced an architecture change and a whole new direction. Efficiency over mhz basically. Theyre not lazy anymore, theyre getting their act together

 

[AntiStatic]

AMD pwns you INteL egg fryers.

 

[TuxDave]

There are several points in the article which do not make sense. His argument that Intel is reacting to AMD's ideas is based off of release date and not based on proposal of a technology.

"AMD hit 1GHz first and so Intel was just reacting to AMD's 1GHz". I highly doubt that Intel never though about releasing a 1GHz part. It just so happened that AMD executed faster in hitting that target.

I'm not too familiar with exactly when 64-bit support or dual core support was in the works at Intel but just because a company is able to execute an idea faster does not mean that everyone else who comes in 2nd is 'copynig the idea'.

 

[dmens]

I used to read firingsquad, but only for game reviews. People who don't anything about engineering aspects of CPU design shouldn't talk about engineering decisions. I love articles that contain this: "CPU design is a complicated process, BUT..."

The funniest assertion in the entire article is the whole "intel reacts to AMD" thing. I'd agree that dual-core was a reactionary move, but that's about it. EM64T has been on the table for ages, plus it is a cakewalk to implement. SSE was about as reactionary towards 3DNow as 3DNow was reactionary towards MMX, plus both MMX and 3DNow sucked. On-die memory controller was on Timna before the ill-fated Rambus decision (and before K8 for that matter). CSI is almost entirely homegrown, an inevitable development that's about as much of a copy as AMD taking up out-of-order execution.

I'm surprised the author didn't mention NX-bit, but sorry, that was developed in parallel as well. Needless to say, the term "copying" would be a fanboyish stretch.

I won't accuse AMD of "copying" the following intel techs, even if the intel implementation came or will come to market first:

- Speedstep
- Virtualization
- LaGrande & iAMT2 (no AMD response yet afaik)
- Rockton (no AMD response yet afaik)
- SMT (no response and there probably never will be, but that is a perfectly legitimate engineering decision)

In regards to the "core" trademark, it seems like the guy is agonizing over the mere possibility that some "doofus" will buy Intel over AMD because he "doesn't know any better", as if yonah is crappy, or somesuch. First off, "doofa" have been doing that for ages because of AMD's non-brand. Secondly, trademarks are first-come-first-serve, why didn't AMD claim it earlier? Thirdly, that kind of elitist attitude is insanely arrogant. Last of all, what's new? The core trademark is about as legitimate as AMD's early claim of dominion over the 64-bit world. All marketing gimmicks.

Guess my lunch break is up. I advise the firingsquad dude to stick to game reviews, because he really has no idea what he's talking about. The article was written by a fanboi, for fanbois.

 

[sandorski]

Originally posted by: dmens
I used to read firingsquad, but only for game reviews. People who don't anything about engineering aspects of CPU design shouldn't talk about engineering decisions. I love articles that contain this: "CPU design is a complicated process, BUT..."

The funniest assertion in the entire article is the whole "intel reacts to AMD" thing. I'd agree that dual-core was a reactionary move, but that's about it. EM64T has been on the table for ages, plus it is a cakewalk to implement. SSE was about as reactionary towards 3DNow as 3DNow was reactionary towards MMX, plus both MMX and 3DNow sucked. On-die memory controller was on Timna before the ill-fated Rambus decision (and before K8 for that matter). CSI is almost entirely homegrown, an inevitable development that's about as much of a copy as AMD taking up out-of-order execution.

I'm surprised the author didn't mention NX-bit, but sorry, that was developed in parallel as well. Needless to say, the term "copying" would be a fanboyish stretch.

I won't accuse AMD of "copying" the following intel techs, even if the intel implementation came or will come to market first:

- Speedstep
- Virtualization
- LaGrande & iAMT2 (no AMD response yet afaik)
- Rockton (no AMD response yet afaik)
- SMT (no response and there probably never will be, but that is a perfectly legitimate engineering decision)

In regards to the "core" trademark, it seems like the guy is agonizing over the mere possibility that some "doofus" will buy Intel over AMD because he "doesn't know any better", as if yonah is crappy, or somesuch. First off, "doofa" have been doing that for ages because of AMD's non-brand. Secondly, trademarks are first-come-first-serve, why didn't AMD claim it earlier? Thirdly, that kind of elitist attitude is insanely arrogant. Last of all, what's new? The core trademark is about as legitimate as AMD's early claim of dominion over the 64-bit world. All marketing gimmicks.

Guess my lunch break is up. I advise the firingsquad dude to stick to game reviews, because he really has no idea what he's talking about. The article was written by a fanboi, for fanbois.

3DNow didn't suck, in fact it rocked, but was hardly used.

 

[dmens]

As an extension over MMX (copy LOL!!!) it addressed many concerns with those instructions, had some interesting additions, but throughput was still low. Perhaps that is the reason it was seldom used. I've never heard any really good things said about it, but please point me to any positive commentary on 3dnow, thanks.

 

[sandorski]

Originally posted by: dmens
As an extension over MMX (copy LOL!!!) it addressed many concerns with those instructions, had some interesting additions, but throughput was still low. Perhaps that is the reason it was seldom used. I've never heard any really good things said about it, but please point me to any positive commentary on 3dnow, thanks.

Quake 2's 3DNow patch made a k6-2 300 equivalent in performance to a PII 400. A few other games gained 25%ish performance using it, such as Power Slide.

 

[dmens]

Thanks. Do you have a link? So 3DNow wasn't bad, my bad. But the point regarding the MMX, 3DNow and SSE relationship stands.

 

[dexvx]

Originally posted by: sandorski

Originally posted by: dmens
As an extension over MMX (copy LOL!!!) it addressed many concerns with those instructions, had some interesting additions, but throughput was still low. Perhaps that is the reason it was seldom used. I've never heard any really good things said about it, but please point me to any positive commentary on 3dnow, thanks.

Quake 2's 3DNow patch made a k6-2 300 equivalent in performance to a PII 400. A few other games gained 25%ish performance using it, such as Power Slide.

No, the 3dNow patch made a K6-2 300 equivalent to a Pentium-2 of the same clock speed.

 

[sandorski]

My bad, Ace's says PII 375

AMD tested using 3D Winbench that showed a k6-2 300 just shy of PII 400 performance, but I won't include a link as it's a synthetic bench.

 

[Zim]

I find the article puerile. These are just companies like any other, making products and trying to sell them to produce a profit and a dividend for their shareholders. The Intel brand name is not "core", they are "Core Solo" and "Core Duo". If he wants to bitch about stuff, how about bitching about AMD using Pentium performance ratings to name their chips. And speaking of Pentium, I remember people rounding on Intel for that too - they said it was a stooopid name because it was a mix of Ancient Greek and Latin, which upset the scholars.

I have an AMD Opteron that I think is the best chip I have ever owned. But I am NOT an AMD fanboy. If in 12 months Intel brings out a super-duper chip, I'll be all over that, I can promise you. And it could happen. I don't think a 45nm quad-core is going to do it... but you never know. And ask yourself, what do we know of where AMD is going, once they have maxxed out the A64 lines? Do they have another winner up their sleeves, or will Intel pull something new out of the hat. Sit back and enjoy... but don't bitch.

 

[sandorski]

Originally posted by: Zim
I find the article puerile. These are just companies like any other, making products and trying to sell them to produce a profit and a dividend for their shareholders. The Intel brand name is not "core", they are "Core Solo" and "Core Duo". If he wants to bitch about stuff, how about bitching about AMD using Pentium performance ratings to name their chips. And speaking of Pentium, I remember people rounding on Intel for that too - they said it was a stooopid name because it was a mix of Ancient Greek and Latin, which upset the scholars.

I have an AMD Opteron that I think is the best chip I have ever owned. But I am NOT an AMD fanboy. If in 12 months Intel brings out a super-duper chip, I'll be all over that, I can promise you. And it could happen. I don't think a 45nm quad-core is going to do it... but you never know. And ask yourself, what do we know of where AMD is going, once they have maxxed out the A64 lines? Do they have another winner up their sleeves, or will Intel pull something new out of the hat. Sit back and enjoy... but don't bitch.

I'm a diehard AMD fan, but I almost wish Intel would bring out some ownage(at least for a few months) and a bunch of Anandtechers would jump on it. The reason is quite simple and is a result of the Net hitting close to my Real Life when after recommending Anandtech to a friend of mine his response was, "Anandtech is AMD biased!" Inside I was all "WTF?", in reality I whimped out an attempted explanation of why Anandtech appears that way, but the look on his face convinced me to just drop it.

 

[Cooler]

I better Trademark Core Quad quick!.

 

[Marmion]

Originally posted by: Cooler
I better Trademark Core Quad quick!.

And I'll trademark Core Octo

 

[dexvx]

Originally posted by: sandorski
My bad, Ace's says PII 375

AMD tested using 3D Winbench that showed a k6-2 300 just shy of PII 400 performance, but I won't include a link as it's a synthetic bench.

According to the article, it says a fully optimized 3d-Now application would result in a 10-30% performance advantage over a Pentium-II.

A lot of tests show that a K6-2 300 runs a bit faster than a PII-300 and sometimes it even tops a PII-333.

When a FPU-intensive game or application is well optimized for 3Dnow! you can expect that the K6-2 will outperform a equivalent PII by about 10-30%, but most games are not enough optimized for 3Dnow! to show its potential. Should K6-2 owners be dissapointed in 3Dnow!?

But there are many other variables. VIA/ALI AGP implementation was lackluster compared to the 440BX. And then there was the FSB of the P2 (either 66 or 100Mhz). Look at tomshardware's earlier massive CPU roundup's (which even include a pentium 100). The K6-2/3 are generally a notch down from the P2 series.

 

[sandorski]

Originally posted by: dexvx

Originally posted by: sandorski
My bad, Ace's says PII 375

AMD tested using 3D Winbench that showed a k6-2 300 just shy of PII 400 performance, but I won't include a link as it's a synthetic bench.

According to the article, it says a fully optimized 3d-Now application would result in a 10-30% performance advantage over a Pentium-II.

A lot of tests show that a K6-2 300 runs a bit faster than a PII-300 and sometimes it even tops a PII-333.

When a FPU-intensive game or application is well optimized for 3Dnow! you can expect that the K6-2 will outperform a equivalent PII by about 10-30%, but most games are not enough optimized for 3Dnow! to show its potential. Should K6-2 owners be dissapointed in 3Dnow!?

But there are many other variables. VIA/ALI AGP implementation was lackluster compared to the 440BX. And then there was the FSB of the P2 (either 66 or 100Mhz). Look at tomshardware's earlier massive CPU roundup's (which even include a pentium 100). The K6-2/3 are generally a notch down from the P2 series.

I don't deny that k6-2 was weaker than its' PII equivalent, just stating that 3DNow wasn't a crap feature, just under utilized. Also, 30% better than a PII 300 would be approximately PII 400 performance.

 

[dexvx]

Originally posted by: sandorski
I don't deny that k6-2 was weaker than its' PII equivalent, just stating that 3DNow wasn't a crap feature, just under utilized. Also, 30% better than a PII 300 would be approximately PII 400 performance.

SIMD is the future, at least imo.

However, what I generally find amusing is that the AMD fanbois are claiming lack of optimizations are hurting them back then, and take no muse to the fact that the lack of SSE2/3 optimizations are hurting to Pentium-4, as evident in many commercial builds of games using ancient 4 year old compilers.

Thus, thats why I'm going to assume AMD took the K7 route with the don't need no optimizations approach.

 

[sandorski]

Originally posted by: dexvx

Originally posted by: sandorski
I don't deny that k6-2 was weaker than its' PII equivalent, just stating that 3DNow wasn't a crap feature, just under utilized. Also, 30% better than a PII 300 would be approximately PII 400 performance.

SIMD is the future, at least imo.

However, what I generally find amusing is that the AMD fanbois are claiming lack of optimizations are hurting them back then, and take no muse to the fact that the lack of SSE2/3 optimizations are hurting to Pentium-4, as evident in many commercial builds of games using ancient 4 year old compilers.

Thus, thats why I'm going to assume AMD took the K7 route with the don't need no optimizations approach.

Agreed. 3DNow was the great White Hope that never really materialized. I think most of us AMDers from them just gave up waiting and are cynical about anything other than raw power. Potential is no good until it's realized.

 

[DeathReborn]

Intel does it again with the "Platform Refresh" to support Conroe.

http://www.dailytech.com/article.aspx?newsid=789

It's a mystery why people haven't kicked up a huge stink at Intel for doing this for almost every new chip.

 

[dexvx]

Originally posted by: DeathReborn
Intel does it again with the "Platform Refresh" to support Conroe.

http://www.dailytech.com/article.aspx?newsid=789

It's a mystery why people haven't kicked up a huge stink at Intel for doing this for almost every new chip.

No, you can actually blame the motherboard manufacturers because they're the ones that are profiting the most from this.

I bought an Asus P4P800, 865 chipset, over 2 years ago and it supports Prescotts and even Dothans. Whereas most other 865 boards cant support either.

 

[openwheelformula1]

The mobo manufacturers still need to purchase chipsets to make motherboards you know. Intel is requiring new chipset updates to support newer CPUs WAY TOO OFTEN.

 

[dexvx]

Originally posted by: openwheelformula1
The mobo manufacturers still need to purchase chipsets to make motherboards you know. Intel is requiring new chipset updates to support newer CPUs WAY TOO OFTEN.

All 915 and above chipsets can support Conroe.

In fact, even the ancient 865 can support Pressler.