It seems like AMD is getting crushed...

[dexvx]

Originally posted by: Viditor
Well, as dexvx points out, it is just speculation at the moment and we should keep that in mind. But the core improvements of note for K8L (grabbed hastily from Wiki) are:

More aggressive prefetching (16 bytes to 32 bytes)
Out of order loads
128 bit wide Floating point units
Larger Out of Order (OoO) buffers
Greater number of entries in Branch Target Buffer
Probable new additions to micro-ops ROM

Add to that the reduced latency of HT and ODMC, and it seems to me that K8L will be crowned the new champ next summer...

The 128bit wide FPU is interesting. If it can do single pass 128bit SSE2 instructions in 1 pass, it'll have SSE2 performance on par with Core2. Everything else just seems more or less tweaks that should gain any feasible performance beyond the spread of error. But those are just too vague.

However, just a latency decrease doesn't seem to be worthy of any performance differential. Memory latency decrease is still no match for cache latency. This is where the L3 will get interesting (again depending on how slow). Again, this is all speculation and you are assuming nothing actually gets slower. Remember, Conroe's L2 is actually a little slower than Yonah's (on a per clock basis).

Originally posted by: Viditor
For Conroe, the major advantages appear to be:

Can decode and execute 4 commands per clock cycle (though I have yet to see an example where it actually does this, it has the ability and hence the headroom)

Process 128-bit SSE3 instructions without slowing down

While we certainly won't know how these 2 play out until samples are released, my own opinion is that K8L comes pretty darn close to Conroe in performance.

If you compare Yonah to Conroe (closest derivative), the main performance increase resulted in the 4 issue wide buffer compared to 3. "Simple" math would yield an optimal 33% increase in per clock improvement, and real world was more like 20% per clock on average.

Another performance issue is the larger, but slower (per clock), L2 cache as well as more L1 and L2 bandwidth.

There were also a series of tweaks here and there, but seriously, I doubt those wouldnt be noticeable beyond the spread of error.

 

[zsdersw]

Originally posted by: harpoon84
I think what Viditor is trying to say is that K8L's core features are just as impressive as Conroe, and with the added benefits of an IMC will enable it to exceed Conroe levels of performance.

I know what Viditor is trying to say... I'm saying he's wrong. I don't think the IMC will be what, if anything, puts the K8L ahead of Conroe/Woodcrest. If anything, it'll be the improvements to the core.

 

[Toadster]

Originally posted by: SparkyJJO

Originally posted by: OhioState
everyone likes to discuss the merits of X2-VS-C2D but can we agree that semperon totally pwns cerleron >?

Agreed. Celeron is pretty bad. Only once CPU series I know of is worse - VIA processors

do you remember the Celeron 300A?

 

[Stumps]

Originally posted by: Toadster

Originally posted by: SparkyJJO

Originally posted by: OhioState
everyone likes to discuss the merits of X2-VS-C2D but can we agree that semperon totally pwns cerleron >?

Agreed. Celeron is pretty bad. Only once CPU series I know of is worse - VIA processors

do you remember the Celeron 300A?

I still have one of those in my stockpile...it easily made it up to 450mhz no worries at all.

I remember back in 98 when I was using it and my friend had a "new" P2 450 that cost him nearly 3 times as much and yet it couldn't keep up with my "lowly" celeron.

ahhh those were the days

 

[dexvx]

Originally posted by: Stumps

Originally posted by: Toadster

Originally posted by: SparkyJJO

Originally posted by: OhioState
everyone likes to discuss the merits of X2-VS-C2D but can we agree that semperon totally pwns cerleron >?

Agreed. Celeron is pretty bad. Only once CPU series I know of is worse - VIA processors

do you remember the Celeron 300A?

I still have one of those in my stockpile...it easily made it up to 450mhz no worries at all.

I remember back in 98 when I was using it and my friend had a "new" P2 450 that cost him nearly 3 times as much and yet it couldn't keep up with my "lowly" celeron.

ahhh those were the days

No different than a E6300 @ $183 being overclocked and beating stock X6800's and FX-62's easily.

AFAIK, the price on the 300a @ release was MSRP'ed around $150ish and the P2-450 was MSRP'ed at around the $500 mark.

 

[Stumps]

Originally posted by: dexvx

Originally posted by: Stumps

Originally posted by: Toadster

Originally posted by: SparkyJJO

Originally posted by: OhioState
everyone likes to discuss the merits of X2-VS-C2D but can we agree that semperon totally pwns cerleron >?

Agreed. Celeron is pretty bad. Only once CPU series I know of is worse - VIA processors

do you remember the Celeron 300A?

I still have one of those in my stockpile...it easily made it up to 450mhz no worries at all.

I remember back in 98 when I was using it and my friend had a "new" P2 450 that cost him nearly 3 times as much and yet it couldn't keep up with my "lowly" celeron.

ahhh those were the days

No different than a E6300 @ $183 being overclocked and beating stock X6800's and FX-62's easily.

AFAIK, the price on the 300a @ release was MSRP'ed around $150ish and the P2-450 was MSRP'ed at around the $500 mark.

at the time, in Australia the Celery went for around AU$250 the P2 cost upwards of AU$800

 

[dexvx]

Originally posted by: Stumps

Originally posted by: tylerdustin2008
No different than a E6300 @ $183 being overclocked and beating stock X6800's and FX-62's easily.

AFAIK, the price on the 300a @ release was MSRP'ed around $150ish and the P2-450 was MSRP'ed at around the $500 mark.

at the time, in Australia the Celery went for around AU$250 the P2 cost upwards of AU$800

Same ratio: 500/150 = 3.3, 800/250 = 3.2

I suppose half the glory of the 300A was that eventually, it came down to about the $60 USD mark for a brand new one (when P3-550's were top end). For some reason, I doubt the E6300 will come down so sharpely anytime soon.

However, its getting a little off-topic. In general, I think its too soon to speculate on performance gains. However, even assuming its going to have the same IPC as a Core2, it would need to debut at the same clock frequency as the Core2 to remain competitive (which I would imagine you need the mid-3Ghz range). And then theres a HUGE question of 4 core scaling from K8L. How will it scale? MAny people assumed that Kentsfield will scale poorly due to a "FSB bottleneck" which the folks at XS completely debunked, even at the stock 1066FSB it proved to scale just as well as 2x DC Opterons on a single channel FSB.

 

[Viditor]

Originally posted by: harpoon84

Originally posted by: Viditor
While we certainly won't know how these 2 play out until samples are released, my own opinion is that K8L comes pretty darn close to Conroe in performance.
Add to that the reduced latency of HT and ODMC, and it seems to me that K8L will be crowned the new champ next summer...

And you really think Intel will do nothing from now until Q3 07? Does 45nm or Yorkfield ring a bell?

You keep going on about how K8L will beat Conroe, but you fail to mention that it won't even be competing with Conroe (65nm) when it is launched but rather Yorkfield @ 45nm.

According to the latest reports, it'll be Yorkfield @ 3.4 - 3.7GHz vs K8L @ 2.7 - 2.9GHz, K8L better well be significantly faster than Conroe (and Yorkfield) clock for clock if it wants to compete.

Of course, I'll be expecting you to say you are sceptical of a Q3 Yorkfield release, that the slated clockspeeds are unrealistic, etc etc.

Of course I don't think Intel will sit still, but they are certainly limited in what they can do in such a short period of time...
1. Yorkfield is a quad core design, K8L (or Rev B if you prefer) is both quad and dual core which is really where mainstream (and even most performance) desktop will be till 2008 (IMHO). Quad core will not really help in gaming at least...
2. 45nm doesn't add a thing to performance, though it will add headroom for clockspeed.
3. K8 is already coming out at 3GHz on 90nm...do you really doubt that at least the DC version of K8L won't be at 3.4-3.7 on 65nm? In other words, do you think that AMD's power profile will be sitting still?

As to the timeframe and clockspeeds, I have yet to see any statements from Intel that contradicts their end Q4 early Q1 release...the only contradiction I've seen to that schedule is from a piece released by HKEPC which is dubious at best.
As to clockspeed, I don't dispute that it's certainly possible. However, it's a fairly wild conjecture at this point.

 

[liebremx]

Originally posted by: Viditor
2. 45nm doesn't add a thing to performance, though it will add headroom for clockspeed.

Not directly. Indirectly it will allow larger caches wich do affect performance.

Originally posted by: Viditor
3. K8 is already coming out at 3GHz on 90nm...do you really doubt that at least the DC version of K8L won't be at 3.4-3.7 on 65nm? In other words, do you think that AMD's power profile will be sitting still?

K8L will probably be taken to 3.4-3.7 by enthusiasts but I doubt you will see those as stock clock speeds. Also I think AMD will strongly follow the path of reducing power consumption which will probably hinder overclocking potential as well.

We'll see. It's going to be interesting.

 

[Nil Einne]

Time for a quick economics lesson. Even if AMD went out of business, prices would not "skyrocket" for Intel processors.

Enter our friend Supply and Demand. If Intel cranked up the prices, then there would be a large surplus of Intel processors, and many consumers would be rationed out of the market by the high price. It doesn't make economic sense for Intel to drive prices up. Suppose that Intel does crank up prices. How will that affect the consumers? Probably a longer lifespan for their computers, so they don't have to buy the expensive processors as often. This lowers demand, which forces prices down, as well as quantity. Ultimately Intel loses if they drive prices up.

I've never studied economics but I think you're missing something here. I've never actually quite understood why (perhaps it has to do with the perception they have large room for growth) but many companies do in fact keep prices high even though it means they're selling less and could probably make more money by selling more at a cheaper price. Case in point is software, music and movies. For that matter, try going to a country without an unbundled local loop where the telco is able to control the broadband market and look at the prices there (like here in NZ). Point being, monopolies do tend to keep the prices up even if they could make more money by selling more at a lower price.

But I digress, the stronger/more important evidence is when you actually look at the history. There is strong evidence to suggest that many of Intel's price cuts were driven by AMD and the average price has come down thanks to an increasingly competitive AMD. Converstly if we look at AMD there is strong evidence to suggest many of their price cuts were strongly influenced by Intel. Case in point, AMD prices especially dual core have been very high. Does anyone doubt that if Intel had still been fooling around with netburst AMD's prices would be anything like they are today? No of course not. They would be a lot higher. Whatever your economics tells you, the evidence is that monopolies do tend to keep prices high. Intel may not necessarily raise prices per se, but they will not cut prices by much and when new products come along they will be more expensive then they would be otherwise.

Of course, I don't get why everyone is so convinced AMD is screwed. Intel was in that boat for years but they didn't die. True they had a much larger market share, a better image and they were a much bigger company but they survived for years with their problems. And AMD's problems don't seem anywhere nearly as bad. AMD have an excellent architecture as far as we know, a team which has been shown to be able to deliver and a roadmap which suggests they will deliver. And despite the fact Intel is better clock for clock, it doesn't really matter. What matters are things like price/performance, what the OEMs want, etc. AMD might not be able to compete at the top at the moment but they don't really need to.

 

[Nil Einne]

Originally posted by: zsdersw

Originally posted by: harpoon84
I think what Viditor is trying to say is that K8L's core features are just as impressive as Conroe, and with the added benefits of an IMC will enable it to exceed Conroe levels of performance.

I know what Viditor is trying to say... I'm saying he's wrong. I don't think the IMC will be what, if anything, puts the K8L ahead of Conroe/Woodcrest. If anything, it'll be the improvements to the core.

So you're saying you think if the K8L didn't have an IMC it would be just as fast it is going to be with the IMC? I.E. The IMC doesn't benefit AMD's dualcore, single chip line at all?

 

[Nil Einne]

Originally posted by: Regs
Cache latency is very important. Conroe's memory sub-system can beat AMD's IMC with "brute force" cache as AMD calls it. Whatever AMD wants to call it, Intel's system seems to be working more efficiently than AMD's. AMD has grown too large now to ignore the desk top market and they can no longer depend on Opteron revenue for future growth. They will have to deliver a new product the will satisfy their customers across the board. They know this and they plan to deliver. The only question is when.

There is no problem with brand loyalty. It's perfectly natural and it shows that they are doing their jobs trying to keep their customer base. Though to wait over a year for a upgrade seems like a severe penalty to pay for loyalty.

Well I guess with AMD at least you know you'd probably be able to upgrade like they promised unlike with Intel where you can be resonably sure the next CPU won't work in current chipsets for whatever reason

Originally posted by: dmens
well imo neither bandwidth nor latency is really that important when viewed relative to other factors. it'd be a waste for amd to spend resources to tune that metric.

as for the list of improvements, that's all nice, but how high can it clock, and how high will be power be at given frequencies? i noticed the stallion codename cores have really high power targets compared to intel planned products in the same time frame. merom is a prime example of frequency scalability of a part with a low power design target.

whatever, data should be coming along in 6 months.

Are you talking about the TDP? because if you are let's not forget they aren't the same thing

Originally posted by: dexvx
No, Celeron-M's squarely and soundly beat Sempron's.

Are you sure? I thought Celeron-Ms didn't even have most power saving modes?

 

[Nil Einne]

 

[Nil Einne]

 

[myocardia]

Originally posted by: Nil Einne
Well I guess with AMD at least you know you'd probably be able to upgrade like they promised unlike with Intel where you can be resonably sure the next CPU won't work in current chipsets for whatever reason

Actually, Intel is the one where you've been able to keep the same chipset for a very long time (recently, that is). If you bought a 975x chipset when they first came out, you're able to run any C2D processor now, and if it's the right motherboard, you can also run the Kentsfield ES's in them, if Intel was nice enough to ship you one.

 

[dexvx]

Originally posted by: liebremx

Originally posted by: Viditor
3. K8 is already coming out at 3GHz on 90nm...do you really doubt that at least the DC version of K8L won't be at 3.4-3.7 on 65nm? In other words, do you think that AMD's power profile will be sitting still?

K8L will probably be taken to 3.4-3.7 by enthusiasts but I doubt you will see those as stock clock speeds. Also I think AMD will strongly follow the path of reducing power consumption which will probably hinder overclocking potential as well.

Remember that even at 130nm, K8 was at the 2.6Ghz mark already. There have been conflicting roadmaps. HKEPC reported that the K8L will top at 3Ghz on initial introduction. But just because its coming out at 3Ghz in a lab is meaningless. Intel has already publically demo'ed the Kentsfield at 3.73Ghz on pre-release commercial software, but its introduction will only be in the mid-2Ghz range.

Originally posted by: Nil Einne

Originally posted by: dexvx
No, Celeron-M's squarely and soundly beat Sempron's.

Are you sure? I thought Celeron-Ms didn't even have most power saving modes?

Performance wise, the Celeron-M can soundly beat Sempron 64's clock for clock in most situations. Despite having no power saving features, it can also beat Sempron 64's in battery life in most situations (same general model but different platform).

See:

http://www.anandtech.com/mobile/showdoc.aspx?i=2625

for more information. Pay special attention to the Compaq V2000/V2000Z because they are essentially the same notebook with a different platform: Celeron-M and Sempron64. Note that a Celeron-M 360 = 1.4Ghz/1MB and the Semperon64 2800+ operates at 1.6Ghz/256KB.

 

[zsdersw]

Originally posted by: Nil Einne
So you're saying you think if the K8L didn't have an IMC it would be just as fast it is going to be with the IMC? I.E. The IMC doesn't benefit AMD's dualcore, single chip line at all?

You're not making a valid comparison. The K8L's performance vis-a-vis the Intel chips of that time will be compared to what came before it, the K8. Both have IMC's. The question will be whether K8L's IMC is different enough from the K8's to make it a significant factor in any performance advantage K8L may have over what's available from Intel at the time. Viditor thinks it will be.. I don't.

 

[javiervi]

It's very simple, Intel has way more money for RD and that will let them generally have better products. I dont think they will every fall behind as much as they did with the P4 vs A64.

 

[OcHungry]

Originally posted by: Frackal
I'm glad AMD got their ass handed to them. I am still pissed about the prices they charged for X2s when they first came out and they couldn't overclock for sh*t either. Not that I want to see AMD gone, but my Conroe experience has been 10000x better than my X2 experience was

oh really?

 

[harpoon84]

Originally posted by: OcHungry

Originally posted by: Frackal
I'm glad AMD got their ass handed to them. I am still pissed about the prices they charged for X2s when they first came out and they couldn't overclock for sh*t either. Not that I want to see AMD gone, but my Conroe experience has been 10000x better than my X2 experience was

oh really?

Oh really what? Which part of his post do you disagree with? This should be interesting...

 

[sandorski]

Originally posted by: RichUK
Why do you care whats happening to AMD? If you can?t get the best product from them you?ll go somewhere else, simple. It?s AMD?s problem not yours.

I don't care about either company, i'm just going to look out for my best interests.

Intel ran over my dog, AMD nursed it back to life. Damn right I care!!

 

[harpoon84]

Originally posted by: sandorski

Originally posted by: RichUK
Why do you care whats happening to AMD? If you can?t get the best product from them you?ll go somewhere else, simple. It?s AMD?s problem not yours.

I don't care about either company, i'm just going to look out for my best interests.

Intel ran over my dog, AMD nursed it back to life. Damn right I care!!

LOL!