Official Phenom 2 Review Thread

[Peter Trend]

Remember? You say that like I might have actually changed cpus before, lmao. Will that solve it? I thought clearing the cmos was for if you wrecked it by overclocking.
Thanks I will give it a shot

 

[Idontcare]

Originally posted by: Peter Trend
Remember? You say that like I might have actually changed cpus before, lmao. Will that solve it? I thought clearing the cmos was for if you wrecked it by overclocking.
Thanks I will give it a shot

Clearing the CMOS is tech support item #1 on the list, kinda like how anytime your windows install is acting funny the first thing we do is reboot and see if that fixes it.

Yeah even if it doesn't turn out to be the solution to your mobo issue it is generally recommended that anytime you do major change like new CPU or bios update your "clear the cmos" to ensure all those assumed zero values are actually initialized to be zero, etc.

 

[Peter Trend]

It worked!
http://i147.photobucket.com/al...an/RestartRequired.jpg
thankyou Idontcare
Im off to bed now. Btw, I have some "interesting"data to show you for the phenom 9500. Will see how well PhII runs MetaTrader tomorrow.
I better load that folding@home SMP client now hadn't I? Lol

 

[Idontcare]

Super! Glad that was all that was holding you back.

 

[Cookie Monster]

Originally posted by: Idontcare
It's true, each wolfdale dual-core chip has a shared 6MB L2$ seen by just the two cores on that die and the data are only accessible via FSB should a thread on another core and another die need to access data on the other die.

core0,1 can access the L2$ on the other chip in that they access data stored there but they cannot address the L2$ in a manner that would make it usable as if it were more L2$ outside its own 6MB L2$. (this is not a mandatory limitation by the way, Intel could have made it work this way albeit with the extra mammoth latency involved but they chose not to)

But counting all 12MB of L2$ is relevant in the sense that it still communicates meaningful data to the consumer. From a technical standpoint they could mislabel it all they want, call it a 12 byte L2$ for all I care, the performance in my apps is all that matters and my apps happen to fit well into the 2x4MB L2$ on Kentsfield so they'll probably fit OK on the shared L3$ of either a PhII or an i7.

Thanks for the explanation IDC

 

[Peter Trend]

I have a load of real benchmarks for the PI 9500 vs PII 940 in an excel spreadsheet, could anybody advise how best to show the people of Anandtech? I mean if people PM me I can email it as an attachment, but there must be a better way surely?

 

[Idontcare]

Originally posted by: Peter Trend
I have a load of real benchmarks for the PI 9500 vs PII 940 in an excel spreadsheet, could anybody advise how best to show the people of Anandtech? I mean if people PM me I can email it as an attachment, but there must be a better way surely?

Open a photobucket account, its free but only hosts images.

Then take a screenshot of your spreadsheet table, crop in your fav editing application and upload to photobucket.

Here's a spreadsheet example: http://i272.photobucket.com/al.../BandwidthAnalysis.jpg

You can do the same thing with your graphs: http://i272.photobucket.com/al...rk_Clock_Scaling-1.gif

 

[Peter Trend]

Thanks I'm just going to need to have a break from the screen for a while before I finish up the Phenom II OC on air. So far seems stable at 3.4GHz, and it was easy, all I did was push the multiplier up from 15 to 17 and left all the voltages on Auto!
But I may hit a temperature wall soon as Im at 39-40C idle 55-57C load. I don't think I dare risk it beyond 60 really.
But the rest of my h2o kit is due to arrive tomorrow, I'll see if I can find time over the weekend to assemble it (That's doubtful though as I have exams to revise for).

The most noticable jump was on GTA IV from ~7fps to 40+fps! 9500 was definitely a massive bottleneck for that game. Sorry I can't give you all the stats quite yet

 

[Martimus]

Originally posted by: Peter Trend
Thanks I'm just going to need to have a break from the screen for a while before I finish up the Phenom II OC on air. So far seems stable at 3.4GHz, and it was easy, all I did was push the multiplier up from 15 to 17 and left all the voltages on Auto!
But I may hit a temperature wall soon as Im at 39-40C idle 55-57C load. I don't think I dare risk it beyond 60 really.
But the rest of my h2o kit is due to arrive tomorrow, I'll see if I can find time over the weekend to assemble it (That's doubtful though as I have exams to revise for).

The most noticable jump was on GTA IV from ~7fps to 40+fps! 9500 was definitely a massive bottleneck for that game. Sorry I can't give you all the stats quite yet

I don't think it is a good idea to leave the voltages on AUTO. When you are overclocking, you should manually adjust them or they can really skyrocket on you.

 

[Peter Trend]

Yeah I couldn't get any more than 3418MHz before hitting 60C. I'm not pushing it too hard now because this stock cooler isn't doing the job well enough.

 

[cmdrdredd]

Originally posted by: alexruiz

Originally posted by: cmdrdredd
I'm glad everyone is so blind that they really believe Intel sent free mobos and memory :roll:

Let me tell you what happened. They tested using the current High end components for either configuration.

Oh, I forgot the reviewers bought the parts, as they were available that very same day at midnight and they picked "next 2 hours" shipping.... talk about retardness.... :roll:

No, most of the big sites receive SAMPLES for testing and reviews. Those samples are usually picked by the manufacturer. It is up to the reviewers to change those configurations. The samples are, officially, not free... but if you keep them for a while for continued reviews, then they are in practice free samples.

As it has been pointed out, if you are already spending money in an X48/X58 mobo and DDR3, why settle for a Penryn chip?

To make a proper assesment of value for the money, a DDR2 configuration should have been used as baseline. In the old days, Anand tested P4 "Northwoods" with i850 chipsets running RDRAM-800 (Rambus) and also tested the P4 in more mundane i845 boards running plain old DDR266.

That doesn't matter. As I said, they test using the current high end components on either platform. They aren't going to go back and pop in a P35 board.

 

[dingleberrydorkbutt]

Originally posted by: xusaphiss
...
AMD is lapped!

It's time for them to die!
...

Then things really would get bad. Intel would charge whatever the hell they want to, and would spend less R & D money, because they wouldn't have to with no competition.

 

[JackyP]

Your comment is as simple-minded and wrong as if I said "AMD going chapter 11 won't affect Intel at all". Please consider both sides of the argument, Intel dominates already, interestingly their biggest competitor is Intel, i.e. their own install and customer base, inventories, etc. Then there's still the law preventing them from charging whatever they want. IIRC.

 

[Idontcare]

Originally posted by: dingleberrydorkbutt

Originally posted by: xusaphiss
...
AMD is lapped!

It's time for them to die!
...

Then things really would get bad. Intel would charge whatever the hell they want to, and would spend less R & D money, because they wouldn't have to with no competition.

Or more to the point of cause-and-effect, in the absence of robust competition Intel's decision makers have less justification to keep pouring shareholder equity into R&D expenses as well as having less justification to not raise prices and maximize shareholder returns.

See Exxonmobile.

edit:

Originally posted by: JackyP
Your comment is as simple-minded and wrong as if I said "AMD going chapter 11 won't affect Intel at all". Please consider both sides of the argument, Intel dominates already, interestingly their biggest competitor is Intel, i.e. their own install and customer base, inventories, etc. Then there's still the law preventing them from charging whatever they want. IIRC.

There is no law preventing Intel or any company from charging whatever they want when we are discussing the "upside" potential.

There is the ever-present threat of congressional action and DOJ investigation.

Where "charging whatever they want" gets problematic is in situations of price-gouging on items the government deems are a necessity (gas, water, food) during short-term periods of crisis (hurricanes, etc).

It is also a good way to be investigated for price-fixing if you and your competitors seemingly raise prices in unison. If both Intel and AMD raised prices on all processors by $100 for example...this is not illegal in its own right but if it can be proven they arranged to do it in unison (i.e. collusion) then it is illegal. (see Samsung)

 

[cmdrdredd]

Originally posted by: Idontcare

Originally posted by: dingleberrydorkbutt

Originally posted by: xusaphiss
...
AMD is lapped!

It's time for them to die!
...

Then things really would get bad. Intel would charge whatever the hell they want to, and would spend less R & D money, because they wouldn't have to with no competition.

Or more to the point of cause-and-effect, in the absence of robust competition Intel's decision makers have less justification to keep pouring shareholder equity into R&D expenses as well as having less justification to not raise prices and maximize shareholder returns.

See Exxonmobile.

edit:

Originally posted by: JackyP
Your comment is as simple-minded and wrong as if I said "AMD going chapter 11 won't affect Intel at all". Please consider both sides of the argument, Intel dominates already, interestingly their biggest competitor is Intel, i.e. their own install and customer base, inventories, etc. Then there's still the law preventing them from charging whatever they want. IIRC.

There is no law preventing Intel or any company from charging whatever they want when we are discussing the "upside" potential.

There is the ever-present threat of congressional action and DOJ investigation.

Where "charging whatever they want" gets problematic is in situations of price-gouging on items the government deems are a necessity (gas, water, food) during short-term periods of crisis (hurricanes, etc).

It is also a good way to be investigated for price-fixing if you and your competitors seemingly raise prices in unison. If both Intel and AMD raised prices on all processors by $100 for example...this is not illegal in its own right but if it can be proven they arranged to do it in unison (i.e. collusion) then it is illegal. (see Samsung)

True.

Lets go over a hypothetical situation here. PII with AM3 is still a ton slower than i7. New i7 parts overclock to 4Ghz on air very easily on 32nm, Intel charges $500 for the cheapest one. People still buy it because the cheapest i7 is guaranteed 4Ghz on air with stock voltages. There is nothing against that at all. Nvidia did similar things on the GPU side for quite some time.

Now lets say AMD files chapter 11, Intel is the only CPU manufacturer left and they raise prices on their CPUs as a result of increased demand for their products as AMD no longer can provide anything to OEMs. Again, nothing wrong with this. Increased demand, lower supply = higher price.

 

[William Gaatjes]

Originally posted by: cmdrdredd

Originally posted by: Idontcare

Originally posted by: dingleberrydorkbutt

Originally posted by: xusaphiss
...
AMD is lapped!

It's time for them to die!
...

Then things really would get bad. Intel would charge whatever the hell they want to, and would spend less R & D money, because they wouldn't have to with no competition.

Or more to the point of cause-and-effect, in the absence of robust competition Intel's decision makers have less justification to keep pouring shareholder equity into R&D expenses as well as having less justification to not raise prices and maximize shareholder returns.

See Exxonmobile.

edit:

Originally posted by: JackyP
Your comment is as simple-minded and wrong as if I said "AMD going chapter 11 won't affect Intel at all". Please consider both sides of the argument, Intel dominates already, interestingly their biggest competitor is Intel, i.e. their own install and customer base, inventories, etc. Then there's still the law preventing them from charging whatever they want. IIRC.

There is no law preventing Intel or any company from charging whatever they want when we are discussing the "upside" potential.

There is the ever-present threat of congressional action and DOJ investigation.

Where "charging whatever they want" gets problematic is in situations of price-gouging on items the government deems are a necessity (gas, water, food) during short-term periods of crisis (hurricanes, etc).

It is also a good way to be investigated for price-fixing if you and your competitors seemingly raise prices in unison. If both Intel and AMD raised prices on all processors by $100 for example...this is not illegal in its own right but if it can be proven they arranged to do it in unison (i.e. collusion) then it is illegal. (see Samsung)

True.

Lets go over a hypothetical situation here. PII with AM3 is still a ton slower than i7. New i7 parts overclock to 4Ghz on air very easily on 32nm, Intel charges $500 for the cheapest one. People still buy it because the cheapest i7 is guaranteed 4Ghz on air with stock voltages. There is nothing against that at all. Nvidia did similar things on the GPU side for quite some time.

Now lets say AMD files chapter 11, Intel is the only CPU manufacturer left and they raise prices on their CPUs as a result of increased demand for their products as AMD no longer can provide anything to OEMs. Again, nothing wrong with this. Increased demand, lower supply = higher price.

And you are paying that higher price

You see, Intel is not doing anything wrong. It is just the way the market works. But on a technological side things will start to stagnate. You are happy because you have a cpu that overclocks well. Let's think of this hypothetical situation :

1 Intel raises prices.
2 people buy the cheapest version because it overclocks so well.

we go to split scenario's here :


Scenario A :
a3 Intel notices that people do not buy the better cpu's anymore because of stagnating sales.

a4 Intel builds a clock check circuit in the normal cpu's that checks if the cpu is running on proper clocks and adjust it to it's proper programmed value when the cpu is being overclocked. This is very easy, nowadays rc clock circuits can be made purely in silicon ( this means on chip) accurate enough to be used as reference clock for comparison. These rc clock generators are calibrated during testing of the cpu at the factory and is programmed with use of blowing internal fuses . Just another part of binning as is determining the clockspeed.

a5 The extreme versions do not have this limitation but a very high price.



Scenario B :
b3 Intel already has foreseen this would happen and has raised the prices enough that there is nothing to worry about for the shareholders of Intel.

b4 The extreme version and the higher clocked version are made in such small quantities that the prices of these cpu's skyrocket.

b5 New technologies are delayed since the market does not want to pay for these new technologies. The only way we see this technologies is in high end hardware for the military or companies and not for us humble customers.


Scenario C :
c3 Intel already has foreseen this would happen and has raised the prices enough that there is nothing to worry about for the shareholders of Intel.

c4 The prices still stagnate and Intel introduces a new cpu.

c5 Only a few people can afford it.

c6 A deal is made with software companies like for example microsoft to introduce something revolutionary new software but it needs the new cpu to work for optimal performance.

c7 you still pay the higher price.

 

[William Gaatjes]

Originally posted by: cmdrdredd

Originally posted by: alexruiz

Originally posted by: cmdrdredd
I'm glad everyone is so blind that they really believe Intel sent free mobos and memory :roll:

Let me tell you what happened. They tested using the current High end components for either configuration.

Oh, I forgot the reviewers bought the parts, as they were available that very same day at midnight and they picked "next 2 hours" shipping.... talk about retardness.... :roll:

No, most of the big sites receive SAMPLES for testing and reviews. Those samples are usually picked by the manufacturer. It is up to the reviewers to change those configurations. The samples are, officially, not free... but if you keep them for a while for continued reviews, then they are in practice free samples.

As it has been pointed out, if you are already spending money in an X48/X58 mobo and DDR3, why settle for a Penryn chip?

To make a proper assesment of value for the money, a DDR2 configuration should have been used as baseline. In the old days, Anand tested P4 "Northwoods" with i850 chipsets running RDRAM-800 (Rambus) and also tested the P4 in more mundane i845 boards running plain old DDR266.

That doesn't matter. As I said, they test using the current high end components on either platform. They aren't going to go back and pop in a P35 board.

You are trying to preach people into buying Intel. That's ok.
But why don't you have the best hardware then ?
As i see it you think Intel is the best but you do not buy the best hardware from Intel.
You buy lower priced cpu's and you overclock them.
I do not find that supportive.

A statement from you :

Originally posted by: cmdrdredd

Originally posted by: William Gaatjes

Originally posted by: cmdrdredd

Originally posted by: William Gaatjes

Originally posted by: cmdrdredd

Originally posted by: William Gaatjes

Originally posted by: cmdrdredd

Originally posted by: Ocguy31

Originally posted by: SickBeast

Originally posted by: cmdrdredd

Originally posted by: qurious63ss

Originally posted by: SickBeast

Originally posted by: cmdrdredd I didn't say it was a scam only misleading and purposefully done. Here's a fact check for everyone here in case it doesn't click. 1) the PII isn't a magic chip that brings AMD to Intel's level. 2) The PII is and always will be slower than i7, and hopefully you understand that the i7 isn't their competition with this chip. 3) the PII is actually slower overall than a C2Q which has been available for months, sometimes not by a large margin but slower is still slower any way you slice it. 4) The names chosen for the PII were deliberately done so to mislead people into buying them thinking they were getting i7 performance for hundreds less. Nothing more and nothing less in my view. That's how it is. You can argue that competition is good and blah blah but the simple fact still remains, AMD is behind Intel. Months behind, perhaps years...i'm not an engineer so I cannot say.

I agree with you, but you should point out that the Phenom 2 is a slightly better value than a C2Q when you take everything into account. The Phenom 2 920 in particular is a very good value and would be hard to beat with a C2Q at the same pricepoint.

Wrong. The Phenom 2 is not a better value. The only people that should consider upgrading to P2 is someone who already owns an AM2+ system. If you own an Intel 775 system then switching to P2 is stupid. If you are in the market for a new systems then you have a whole bunch of variables that can sway you back and forth but there is no definitive answer that makes P2 a better value then C2Q.

Right, and time will tell how far Intel is willing to drop prices on their C2Qs which may further push the PII out of the running for many people.

There is no denying that the P2 is cheaper than and performs better than the C2Q chips within their respective price markets. They also generally overclock better. Plus there is the fact that AMD platform is cheaper, and you don't need to worry about DDR3 or a $300 motherboard.

I can deny it all I want, because it is not true. I ran a C2Q just fine without DDR3, and with a $110 mobo.

Yes, I just stability tested my Q9550 C1 with this old P35 board in my sig at 3.8Ghz. 448x8.5 @ 1.38v in bios (1.35v idle and 1.29v load). The PII is cheaper only for those who are upgrading AM2+ boards. Everyone else will get better performance from Intel parts, and like I said...we will see how low intel cuts prices. I am willing to bet that a DDR2 based P45 or x38/x48 board running a C2Q will still outpace a PII. Memory speed doesn't play that great a role in the performance of the LGA775 chips.

The core generation cpu's of intel always have been memory bandwidth starved. That is why Intel introduced the cpu's with so much cache. It is also the reason why core i7 has an integrated memory controller. Because those core's are so fast they need every little bit of bandwidth they can get. When originally the core architecture was designed Intel already knew it loved memory bandwidth. See for example this page about the architecture of the core. Penryn is the second generation core. Memory disambiguation: the data stream version of speculative execution

DDR3 =/= uber gains. I don't care what people say, the numbers speak for themselves. Lets not beat around the bush...AMD CPUs are still slower than Intel CPUs.

Depends which cpu you mean. What i am seriously interested in is where the limitation lie in the amd cpu Ph2. I wonder what happens with the Ph2 when the bandwidth of this cpu also increases by going to DDR3 with the AM3 socket. I am sure the Ph2 has some bottlenecks hidden in it's architecture and i feel the most obvious one is the low clock of the IMC and the L3 cache. When compared the i7 has much higher uncore clocks. If AMD can scale the "uncore" clocks up of phenom2 i am sure we can see some increase in performance. The core i7 maybe the fastest but the price range is a lot different. For the enthusiast only the core i7 is good enough but for most people and that includes me too the Ph2 is plain good.

Plain good but slower than a C2Q clocked comparatively.

Look i don't really mind what you say. I prefer the Ph2 because overal it provides a platform what i want and those Qx chips don't. And before i forget the Ph2 has Memory disambiguation too it is only called different and it is more conservative. This is what could be improved too with the Ph2, to be more speculative. non-speculative memory access re-ordering

So you want something slower...I get it. Thanks for sharing

[/quote]


Yet you do not buy the best cpu out there which is the i7 965 together with the best mainboard and the best ddr3.
You buy an Intel cpu as long as it is a little bit faster and higher priced then the offerings of AMD . I find that strange and laughable at best.

:laugh: :laugh: :laugh:

 

[dennilfloss]

I what I can afford and fits my taste.

For example, the PH2 940 was supposed to be $300CAN with free shipping at DirectCanada yesterday. Very tempting at that price.

Ha! Now it's $290. with $10 shipping.

http://www.directcanada.com/pr...CGIBOX&manufacture=AMD

Edit: And my mom & pop store said they will match this, with no shipping charge, so there's a 940 BE for $289+taxes order on my to-do list in a couple of weeks.

 

[William Gaatjes]

Originally posted by: Idontcare

Originally posted by: Cookie Monster
This maybe abit offtopic, but shouldn't penryn quads count as having only 6MB of L2$ since the total amount of L2$ isnt "shared" by the 2 penryn dies. For intel MCM quads, they communicate via FSB (i.e introduces some serious bottlenecks in some situations) but can say core0,1 access and use core2,3's L2$ or am i understanding this incorrectly?

It's true, each wolfdale dual-core chip has a shared 6MB L2$ seen by just the two cores on that die and the data are only accessible via FSB should a thread on another core and another die need to access data on the other die.

core0,1 can access the L2$ on the other chip in that they access data stored there but they cannot address the L2$ in a manner that would make it usable as if it were more L2$ outside its own 6MB L2$. (this is not a mandatory limitation by the way, Intel could have made it work this way albeit with the extra mammoth latency involved but they chose not to)

But counting all 12MB of L2$ is relevant in the sense that it still communicates meaningful data to the consumer. From a technical standpoint they could mislabel it all they want, call it a 12 byte L2$ for all I care, the performance in my apps is all that matters and my apps happen to fit well into the 2x4MB L2$ on Kentsfield so they'll probably fit OK on the shared L3$ of either a PhII or an i7.

I once read that the 2 penryn dies replicate eachothers data to keep the traffic over the fsb bus as low as possible. Is this true ? It would sure explain why the core i7 is much more efficiënt with smaller caches because the data traffics over the internal cpu busses instead of the external fsb bus. It would also explain why Intel went for huge L2 caches in general before the core i7.

The i7 has L1 cache of 32kB I cache and 32kB of D-cache. L2 caches of 256kB a core.
And a shared L3 of 8MB. I am sure the cores can snoop around in eachothers L2 cache
But i wonder if it is the L3 cache that is meant as the real one and only data exchanger between the cores to keep thing easy for the cache control logic.

 

[Idontcare]

Originally posted by: William Gaatjes
I once read that the 2 penryn dies replicate eachothers data to keep the traffic over the fsb bus as low as possible. Is this true ? It would sure explain why the core i7 is much more efficiënt with smaller caches because the data traffics over the internal cpu busses instead of the external fsb bus. It would also explain why Intel went for huge L2 caches in general before the core i7.

I have not heard of an explicit duplication of data in the two L2$ sets in a yorkfield but there is nothing preventing this from happening just by happenstance as the prefetchers in both penryns are going to be busy pre-fetching data they think will be relevant to future processing and because threads migrate from core to core constantly the pre-fetchers are going to think the same data is relevant to future processing for all cores.

My understanding was that Intel went with the huge L2$ to take advantage of their rather well designed and aggressive pre-fetchers. The pre-fetchers were implemented not so much to get around the FSB but more to keep the IPC of the core2 architecture as high as possible (which it does).

Originally posted by: William Gaatjes
The i7 has L1 cache of 32kB I cache and 32kB of D-cache. L2 caches of 256kB a core.
And a shared L3 of 8MB. I am sure the cores can snoop around in eachothers L2 cache
But i wonder if it is the L3 cache that is meant as the real one and only data exchanger between the cores to keep thing easy for the cache control logic.

On nehalem the prefetcher is still there, it is just fetching into 6.75MB of 36-cycle L3$ instead of 6MB of 11-cycle L2$. (the remaining 1.25MB of L3$ is consumed by mirror copy of the L1$ and L2$ contents)

But yes the L2$ snoops are eliminated as the L3$ is inclusive. A miss in the L3$ means the system is assured the data lies in the ram or beyond.

Checkout this Nehalem video, it has a cool segment on cache snooping for nehalem. Worth the time to watch it. If you want to zip ahead the "Smart Cache" in the menu is the portion which is relevant to our discussion here.

http://www.intel.com/technolog...next-gen/demo/demo.htm

 

[William Gaatjes]

Originally posted by: Idontcare

Originally posted by: William Gaatjes
I once read that the 2 penryn dies replicate eachothers data to keep the traffic over the fsb bus as low as possible. Is this true ? It would sure explain why the core i7 is much more efficiënt with smaller caches because the data traffics over the internal cpu busses instead of the external fsb bus. It would also explain why Intel went for huge L2 caches in general before the core i7.

I have not heard of an explicit duplication of data in the two L2$ sets in a yorkfield but there is nothing preventing this from happening just by happenstance as the prefetchers in both penryns are going to be busy pre-fetching data they think will be relevant to future processing and because threads migrate from core to core constantly the pre-fetchers are going to think the same data is relevant to future processing for all cores.

My understanding was that Intel went with the huge L2$ to take advantage of their rather well designed and aggressive pre-fetchers. The pre-fetchers were implemented not so much to get around the FSB but more to keep the IPC of the core2 architecture as high as possible (which it does).

Originally posted by: William Gaatjes
The i7 has L1 cache of 32kB I cache and 32kB of D-cache. L2 caches of 256kB a core.
And a shared L3 of 8MB. I am sure the cores can snoop around in eachothers L2 cache
But i wonder if it is the L3 cache that is meant as the real one and only data exchanger between the cores to keep thing easy for the cache control logic.

On nehalem the prefetcher is still there, it is just fetching into 6.75MB of 36-cycle L3$ instead of 6MB of 11-cycle L2$. (the remaining 1.25MB of L3$ is consumed by mirror copy of the L1$ and L2$ contents)

But yes the L2$ snoops are eliminated as the L3$ is inclusive. A miss in the L3$ means the system is assured the data lies in the ram or beyond.

Checkout this Nehalem video, it has a cool segment on cache snooping for nehalem. Worth the time to watch it. If you want to zip ahead the "Smart Cache" in the menu is the portion which is relevant to our discussion here.

http://www.intel.com/technolog...next-gen/demo/demo.htm

The pre-fetchers were implemented not so much to get around the FSB but more to keep the IPC of the core2 architecture as high as possible (which it does).

You are right but you have to agree that the prefetchers are used to eliminate any bottleneck in getting the data as fast a possible to the execution units. The fsb is just one of the many bottleneck as is "slow" main memory. Intel does a great job with the prefetchers tho, i agree totally. Fun part is Intel always seems to take the difficult way to solve problems. But because they do, Intel has much more knowledge. Intel really seems to spend a lot of time simulating and analyzing data streams from and to memory. And what kind of instructions are used the most.

 

[Zstream]

Intel has much more knowledge. Intel really seems to spend a lot of time simulating and analyzing data streams from and to memory. And what kind of instructions are used the most.

Intel has more knowledge then whom? I believe you are referring to AMD, should we all remind you AMD has done more to spur the technical revolution then Intel has by a large sum. In terms of ratio/ratio is concerned.

Rather it is out of necessity or whatever, the fact remains the same.

 

[Idontcare]

Originally posted by: Zstream

Intel has much more knowledge. Intel really seems to spend a lot of time simulating and analyzing data streams from and to memory. And what kind of instructions are used the most.

Intel has more knowledge then whom? I believe you are referring to AMD, should we all remind you AMD has done more to spur the technical revolution then Intel has by a large sum. In terms of ratio/ratio is concerned.

Rather it is out of necessity or whatever, the fact remains the same.

What are you ratio'ing? Patents? Just US or ROW? Or perhaps a likely incomplete (but popular nonetheless) itemized list of architecture achievements?

Bold claims require robust data, otherwise they are nothing more than bold opinion.

 

[cmdrdredd]

Originally posted by: William Gaatjes

You are trying to preach people into buying Intel. That's ok.
But why don't you have the best hardware then ?
As i see it you think Intel is the best but you do not buy the best hardware from Intel.
You buy lower priced cpu's and you overclock them.
I do not find that supportive.


Yet you do not buy the best cpu out there which is the i7 965 together with the best mainboard and the best ddr3.
You buy an Intel cpu as long as it is a little bit faster and higher priced then the offerings of AMD . I find that strange and laughable at best.

:laugh: :laugh: :laugh:

Here's a reality check for you dodos out there just because someone says Intel is better and they should spend the $40 extra for the better CPU, doesn't mean they're telling everyone to buy an i7. That's precisely what you implied and it's really stupid of you to even put those words in the text box.

Now, when I went with intel is was MUCH MUCH faster than anything AMD had available anywhere. Now it's still faster and all i did was swap the CPU.

Who says I have to buy the best stuff? I'm saying that saving $100 on a new build is not what I'd do. I don't go out and buy the newest stuff just because it's new. You're reaching here and it's quite pathetic to witness. How long as the P35 chipset been available? I've had this motherboard THAT LONG! and I have just last month replaced my E6400 which was at 3.2Ghz with this current Q9550. Why should I swap my mobo and reinstall windows? I don't have to and I still have a faster CPU than AMD has.

The reason I got this board was because the P965 board I had went dead and I RMAd it to Asus, so I bought a new mobo to use and sold the P965 board when it returned to me.

If I was building a new system right now, I'd go with Intel because it's faster than AMD and if I'm already planing to spend $700+ I may as well spend a few extra dollars for the better performing part.

And you bet your ass if I could afford to build an i7 940 system I would do it, and have a pair of GTX280s in SLI with a 30" monitor and a couple Raptor drives in RAID, and a Blu-Ray burner...

get the point? I buy what's faster that I can afford, but I'm not going to be trying to save a few dollars and lose performance. I would spend the bit extra for the better part, as long as it's within the budget. You can't convince me, no matter how hard you try that someone spending $275 on a CPU and $500 on motherboard and memory can't afford another $40 for an Intel CPU setup which outperforms the $275 AMD CPU.

 

[William Gaatjes]

Originally posted by: cmdrdredd

Originally posted by: William Gaatjes

You are trying to preach people into buying Intel. That's ok.
But why don't you have the best hardware then ?
As i see it you think Intel is the best but you do not buy the best hardware from Intel.
You buy lower priced cpu's and you overclock them.
I do not find that supportive.


Yet you do not buy the best cpu out there which is the i7 965 together with the best mainboard and the best ddr3.
You buy an Intel cpu as long as it is a little bit faster and higher priced then the offerings of AMD . I find that strange and laughable at best.

:laugh: :laugh: :laugh:

Time for some responding.

Here's a reality check for you dodos out there just because someone says Intel is better and they should spend the $40 extra for the better CPU, doesn't mean they're telling everyone to buy an i7. That's precisely what you implied and it's really stupid of you to even put those words in the text box.

That is not what i am implying. I wrote that when one wants the best x86 performance out there the i7 965 is the way to go. And that is not a lie. Learn to read.

Now, when I went with intel is was MUCH MUCH faster than anything AMD had available anywhere. Now it's still faster and all i did was swap the CPU. Who says I have to buy the best stuff? I'm saying that saving $100 on a new build is not what I'd do. I don't go out and buy the newest stuff just because it's new. You're reaching here and it's quite pathetic to witness. How long as the P35 chipset been available? I've had this motherboard THAT LONG! and I have just last month replaced my E6400 which was at 3.2Ghz with this current Q9550. Why should I swap my mobo and reinstall windows? I don't have to and I still have a faster CPU than AMD has.

I never mentioned such a thing. You live in your own reality.
I am planning to buy a complete new system. Not just exchange the cpu.
And you are still not able to understand why i want an AMD system. It is the overal performance of AMD (that means gfx, chipset with onboard gfx and cpu) that i am very satisfied with. If anybody wants the fastest out there i promote the i7 965 or the i7940 with the best performing MB. If that is not the case i will make a price /performance comparision based on what people want to do with the pc when they ask me for advice.

The reason I got this board was because the P965 board I had went dead and I RMAd it to Asus, so I bought a new mobo to use and sold the P965 board when it returned to me.

What has this to do with the phenom 2 ?

If I was building a new system right now, I'd go with Intel because it's faster than AMD and if I'm already planing to spend $700+ I may as well spend a few extra dollars for the better performing part.

That can be a reason.

And you bet your ass if I could afford to build an i7 940 system I would do it, and have a pair of GTX280s in SLI with a 30" monitor and a couple Raptor drives in RAID, and a Blu-Ray burner...

You would pair up 2 GTX280 but not go with the i7965. I find it strange. I have not done much homework on how much the i940 can keep the 3sli GTX280 (that was what i was talking about) busy with respect to the i965. Nor do i know if the GTX280 in sli or 3sli mode can saturate the nehalem architecture. That depends entirely on for example the game engine and the drivers of the gfx cards.

get the point? I buy what's faster that I can afford, but I'm not going to be trying to save a few dollars and lose performance. I would spend the bit extra for the better part, as long as it's within the budget. You can't convince me, no matter how hard you try that someone spending $275 on a CPU and $500 on motherboard and memory can't afford another $40 for an Intel CPU setup which outperforms the $275 AMD CPU.

That's my point. The intel setup is more expensive then the AMD setup. Yet for real world use there is not that much difference in my opinion. Therefore i go for AMD.
And don't start with: " as the prices come down".

Nothing you have sad made me think : Hmmmm, he has a point there.
And your choice of words in your response explains enough.