AMD shows off "barcelona" Quad core CPU
- Thread starter SickNic
- Start date
ah the sweet smell of fresh blood dripping from intel's nostrils...
the more they punch each other the more performance we get and the less money we pay
the more they punch each other the more performance we get and the less money we pay
Very true. Gotta love the price drops! I must admit, until Intel had the core 2 duo, I was an AMD fanboi, now I got my core 2, and I guess I am intel fanboi for the time being. i was so used to taking sides for AMD, I had to read up on those core 2 benchmarks for a few days before I could finally admit AMD had lost that battle. =P
Mike
Mike
I wish they would have shown off some more or gave us a bit more we could use....otherwise it is pure PR babble...
2 dual cores slapped together or not, if the c2d still have a clock for clock advantage I could care less what is native or not....I think where Intel may hit the wall is their lack of a system like AMD's direct connect architecture....
2 dual cores slapped together or not, if the c2d still have a clock for clock advantage I could care less what is native or not....I think where Intel may hit the wall is their lack of a system like AMD's direct connect architecture....
Duvie... for gods sake, its I couldn't care less not I could care less.
If you say I could, it means that you are caring now, and you could care less if you wanted to.
I couldnt, means that you dont have any care left for the subject.
PEOPLE STOP SAYING I COULD CARE LESS. IT MAKES NO SENSE IN ENGLISH.
If you say I could, it means that you are caring now, and you could care less if you wanted to.
I couldnt, means that you dont have any care left for the subject.
PEOPLE STOP SAYING I COULD CARE LESS. IT MAKES NO SENSE IN ENGLISH.
I know, but i just felt like saying because so many freaking people are picking up the this saying and everytime i hear it it pisses me off.
sorry if i took my rage on you
sorry if i took my rage on you
Originally posted by: JAG87
I know, but i just felt like saying because so many freaking people are picking up the this saying and everytime i hear it it pisses me off.
sorry if i took my rage on you
NP...really just PM me next time....I am sure it isn't the first time nor the last time I will slaughter my own language.....
Saying "I could care less" is like saying it is possible for you to care less than you already do. To say "I couldn't care less" would mean you are at the bottom of the caring totem pole. Correct?
Originally posted by: JAG87
I know, but i just felt like saying because so many freaking people are picking up the this saying and everytime i hear it it pisses me off.
sorry if i took my rage on you
i could care less about ppl freaking you out
i could but i don't want to.
goddamn... PR bots drag out a design presentation and takes pot shots?
maybe intel PR will retaliate by dragging out a dataflow sheet of the nehalem design, or even gesher... see see, it will rock!
idiots, all of them.
maybe intel PR will retaliate by dragging out a dataflow sheet of the nehalem design, or even gesher... see see, it will rock!
idiots, all of them.
Seemed like this was a pretty in depth presentation. Reading through Extremetech's article this chip looks alot like the Conroe architecture except for 3 IPC rather than 4.
http://www.extremetech.com/article2/0,1697,2027633,00.asp
Seems to corroborate Viditor's previous assertions on this so called K8l.
http://www.extremetech.com/article2/0,1697,2027633,00.asp
Seems to corroborate Viditor's previous assertions on this so called K8l.
Back to topic.....
I agree with you guys.... AMD is resorting to this native and non-native design word war. Point is they lost the time to market delivery of quad core.... It sthe same when they released the 64 bit capability. To be honest, I think Intel is correct that industry (client side) is not ready for 64-bit. I cant even see a system/desktop more than 4GB of memory.
Its all about performance. Before I was thing that they did a great job in IMC and hyper transport which if Intel will not implement that, they will be doomed. But Israel team proved us wrong. With FSB bottleneck they still managed to outperform AMD. We know FSB still have this limitation bt Intel did a great job on delivering best processor in their architecture.... I hope AMD is not sleeping a well as Intel not to be complacent again. This is good for us consumers.
Now I wish software can really catch up with hardware.
I agree with you guys.... AMD is resorting to this native and non-native design word war. Point is they lost the time to market delivery of quad core.... It sthe same when they released the 64 bit capability. To be honest, I think Intel is correct that industry (client side) is not ready for 64-bit. I cant even see a system/desktop more than 4GB of memory.
Its all about performance. Before I was thing that they did a great job in IMC and hyper transport which if Intel will not implement that, they will be doomed. But Israel team proved us wrong. With FSB bottleneck they still managed to outperform AMD. We know FSB still have this limitation bt Intel did a great job on delivering best processor in their architecture.... I hope AMD is not sleeping a well as Intel not to be complacent again. This is good for us consumers.
Now I wish software can really catch up with hardware.
Regs
Lifer
- Aug 9, 2002
- 16,666
- 21
- 81
The point is we have more info on AMD's plans and not just some vague specs that we got months ago.
I personally do not care if Intel has two cores glued together. It's faster in gaming and that's all I look for in purchasing a CPU. Other than the obvious power and heat problems if they arise.
The only question in my head still remains. Will AMD be able to take advantage of the memory bandwidth it has to play with? If not - they still might have a competitive product if the price is right. If it does - then you have a Conroe killer.
Hightlighted list of improvements:
[*]AMD estimates power consumption will fit within a 95W envelope
[*]allow individual core frequencies to lower while other cores may be running full bore
[*]SSE execution 128 bits wide (doubled from K8)
[*]Instruction fetch bandwidth 32 bytes/cycle (doubled from K8)
[*]Data cache bandwidth 2 x 128 bits loads/cycle (doubled from K8)
[*]L2 cache/memory controller bandwidth 128 bits/cycle (doubled from K8)
[*]Floating-point scheduler depth 36 dedicated x 128-bit ops (doubled from K8)
[*]SSE MOV instructions can be performed in the floating-point "store" pipe
[*]Improved branch prediction. AMD's architects have doubled the return stack size, added more branch history bits, and built in a 512-entry indirect branch predictor
[*]32-byte instruction fetch. Increases efficiency by reducing split-fetch instruction cases
[*]Sideband stack optimizer. Adjustments to the stack don't take up functional unit bandwidth.
[*]Out-of-order load execution
[*]Other enhancements have also been introduced, including optimizations to the TLBs (translation lookaside buffers), additional Fastpath instructions, and extensions to bit manipulations and SSE instructions
[*]The die now has independent memory controllers, which enables more memory pages to remain open
[*]the memory controllers now support full 48-bit hardware addressing (but who needs it?)
[*]AMD feels that this cache architecture, with dedicated L1 and L2 caches and a shared L3 cache, is better suited for the coming age of virtualization.
I personally do not care if Intel has two cores glued together. It's faster in gaming and that's all I look for in purchasing a CPU. Other than the obvious power and heat problems if they arise.
The only question in my head still remains. Will AMD be able to take advantage of the memory bandwidth it has to play with? If not - they still might have a competitive product if the price is right. If it does - then you have a Conroe killer.
Hightlighted list of improvements:
[*]AMD estimates power consumption will fit within a 95W envelope
[*]allow individual core frequencies to lower while other cores may be running full bore
[*]SSE execution 128 bits wide (doubled from K8)
[*]Instruction fetch bandwidth 32 bytes/cycle (doubled from K8)
[*]Data cache bandwidth 2 x 128 bits loads/cycle (doubled from K8)
[*]L2 cache/memory controller bandwidth 128 bits/cycle (doubled from K8)
[*]Floating-point scheduler depth 36 dedicated x 128-bit ops (doubled from K8)
[*]SSE MOV instructions can be performed in the floating-point "store" pipe
[*]Improved branch prediction. AMD's architects have doubled the return stack size, added more branch history bits, and built in a 512-entry indirect branch predictor
[*]32-byte instruction fetch. Increases efficiency by reducing split-fetch instruction cases
[*]Sideband stack optimizer. Adjustments to the stack don't take up functional unit bandwidth.
[*]Out-of-order load execution
[*]Other enhancements have also been introduced, including optimizations to the TLBs (translation lookaside buffers), additional Fastpath instructions, and extensions to bit manipulations and SSE instructions
[*]The die now has independent memory controllers, which enables more memory pages to remain open
[*]the memory controllers now support full 48-bit hardware addressing (but who needs it?)
[*]AMD feels that this cache architecture, with dedicated L1 and L2 caches and a shared L3 cache, is better suited for the coming age of virtualization.
Zenoth
Diamond Member
- Jan 29, 2005
- 5,202
- 216
- 106
But, by mid-2007, aren't Intel going to release something even better than Conroe anyway ? A "refresh" of it or something ? I'm just asking.
If not then those specifications for Barcelona look awesome. But I still think mid-2007 is too long. They should try for Q1 2007 instead. People will look for Windows Vista in early 2007, and a lot of them (well, me for example) will want to completely upgrade their system once it's released.
If not then those specifications for Barcelona look awesome. But I still think mid-2007 is too long. They should try for Q1 2007 instead. People will look for Windows Vista in early 2007, and a lot of them (well, me for example) will want to completely upgrade their system once it's released.
Originally posted by: JAG87
I know, but i just felt like saying because so many freaking people are picking up the this saying and everytime i hear it it pisses me off.
sorry if i took my rage on you
I think you need an English lesson yourself. You put 2 "it" contiguously in an ambiguous sentence.
AMD can't just accelerate something becuase it is taking a long time. Barcelona is targeting servers so it will be going against Clovertown and the future Tigerton processor and maybe Harpertown as well.
Conroe isn't an issue because Barcelona isn't targeting that CPU. By the time Mid 2007 rolls around were very close to the Conroe shrink Wolfdale/Ridgefield with 6MB of cache, 45nm process, and apparently SSE4.
Regs
Lifer
- Aug 9, 2002
- 16,666
- 21
- 81
'Antares', 'Arcturus' and 'Spica' are the desktop CPU's of AMD's next-gen coming in Q3 07.
Antares will be dual core, Altair will be quad, and im guessing Spica will either be single core or the "duron" of next generation. All will sport 1-2MB of shared L3 cache and the core speeds will range from 2.6-3.0 GHz with HT 3.0 running from 3.0 to 4.0 GHz. The quad will have a TDP of 125W or lower.
Antares will be dual core, Altair will be quad, and im guessing Spica will either be single core or the "duron" of next generation. All will sport 1-2MB of shared L3 cache and the core speeds will range from 2.6-3.0 GHz with HT 3.0 running from 3.0 to 4.0 GHz. The quad will have a TDP of 125W or lower.
AMD is really in no position to go throwing jokes at Intel.
It should rather insert its energy at thinking up how to survive more people opting for Intel Core 2 Duo's until they have a product to directly compete with it.
With the amount of money AMD has been throwing/indexing lately, they are in a volatile situation financially speaking. AMD is about 1/10th the size of Intel, too small to jump and critisize.
Unfortunately, a large percentage of consumers will eat such comments about AMD's take on Intel's quad core. In reality however, such opinion is nonsense. As long as it's a real quad core (4x 1 core), that's all the consumer wants/needs to know.
If Intel was as kid like as AMD, they could tell the public how AMD was caught without having an answer for a quad core system or 65nm production. Rebranding an Opeteron dual CPU motherboard is just that, a quick measure to have something on the market that correlates to the competitor.
The fact of the matter is AMD is mad it hasn't got 65nm because if it did, it would certainly do the same as Intel, i.e. take an X2 and put two of them on the same socket. AMD is also mad that within 6 months of it having 65nm Intel will be rolling out 45nm. Should they be paranoid? It would be the sensible thing to act, yes.
It should rather insert its energy at thinking up how to survive more people opting for Intel Core 2 Duo's until they have a product to directly compete with it.
With the amount of money AMD has been throwing/indexing lately, they are in a volatile situation financially speaking. AMD is about 1/10th the size of Intel, too small to jump and critisize.
Unfortunately, a large percentage of consumers will eat such comments about AMD's take on Intel's quad core. In reality however, such opinion is nonsense. As long as it's a real quad core (4x 1 core), that's all the consumer wants/needs to know.
If Intel was as kid like as AMD, they could tell the public how AMD was caught without having an answer for a quad core system or 65nm production. Rebranding an Opeteron dual CPU motherboard is just that, a quick measure to have something on the market that correlates to the competitor.
The fact of the matter is AMD is mad it hasn't got 65nm because if it did, it would certainly do the same as Intel, i.e. take an X2 and put two of them on the same socket. AMD is also mad that within 6 months of it having 65nm Intel will be rolling out 45nm. Should they be paranoid? It would be the sensible thing to act, yes.
No disrespect GFORCE100, but you really have to understand how things work in the industry first...
For example:
1. I assume that it is general consensus that Core2>AMD>Netburst still...given that only 20% of Intel's shipments are Core2 (and even in Q1 next year it will only hit 40%), then you can understand why AMD is still gaining marketshare and will continue to do so.
2. Almost all consumers neither need nor want quad core at all...
3. If the definition is 4 x 1 core for quad core, then AMD's 4x4 is every bit a quad core...in fact more so as the 2 dual core chips are directly connected via HT rather than the common FSB. Of course a populated 4p single core mobo would be a quad core as well...
4. Intel won't be shipping 45nm till the end of next year (which makes it 12 months), but why should any of that upset AMD at all?
1) As long as there is enough supply to meet demand, then there's enough shipments to allow consumers to buy the processors should they choose to. There are no major shortages of Core 2 Duo's in the channel, it's there to buy today. Intel could make more Core 2 Duo's now but it's you who possibly has got the wrong end of the stick here. The reason to produce less of a new product is to allow all your partners to clear their inventory of older chips hence the Pentium D in most cases. If they ramped up production too quick, this would make it harder to sell off the Netburst CPU's, which for the average consumer are fast anyway. AMD on the other hand doesn't have the production capacity, which is recently best seen by dry spots in the channel. Dell is taking too many shipment for this not to translate into the consumer namespace via online outlets etc. Y
2) That is beyond the point. If this is what's available, and for the right price, people will buy them, especially if the media do their bit to prove it's better than the dual core. Consumers also didn't need 64bit CPU's but they bought them simply because this is what was on offer. Multi-core technology is a much more guaranteed technology, guaranteed to bring immediate benefits, at least in how smooth one's system performs, even when not running specfically optimised software.
3) FSB or specifically GTL+ is really only a bottleneck on 4 or more cores all fighting for the same bandwidth. You must note that a) Intel's quad chip uses 2 dual core chips so at maximum, only 2 cores fight for the bandwidth. The other two cores simply talk to each other via the L2 cache, which is much faster than AMD's caches anyway. At worst this performane delta in comparison to Intel Vs AMD on the FSB Vs HT front is neglible. This would only be a hot debate in the server market, and only then in MP systems. You must also note that a lot of the talk about the limitation of GTL+ was at the time CPU's only had 256K L2 cache. Today CPU's have a lot more, plus it's faster too and so accessing the FSB is not as frequent as before.
4) 2007 Q3 to be precise which means any time from July to September. AMD cannot reliably ship 65nm until Q1 2007. AMD has bought ATI and invested in NY, a lot of money for a small company, or otherwise, a lot of money in proportion to its capital. To allow 65nm AMD needs also a lot of money, and more fabs since its bottleneck is production capacity. Because it has to rely on IBM and others, this instantiates more delay to rollout production to a level satisfied by the Q&A department. Whatever one wishes to say about AMD production, the bottom line is AMD is about 6-12 (conservative estimate) to 9-12 (more realistic) months benhind Intel when it comes to 65nm. AMD knows that if Intel gets to 45nm, it can start add to the architecture by increasing the transistor count without worrying about excessive production cost or heat dissipation. It can also go for faster clock speeds. All of these three aspects mean a faster/better product to the market, a market AMD wants to gain and hold onto, but not lose the market share it's been fighting for these past years.
The situation now is kind of like back in 1997-1998 when AMD just had a complete flop with the K5 series (brought too late to market and wouldn't scale in MHz to the levels of the the then P54/P54C by Intel, (Pentium and Pentium MMX) and couldn't get good yields with the K6 either. To make things worse, Intel then already had the Pentium II what was faster. While the K6 was a good deal compared to a Pentium MMX, the Pentium II was a more advanced product. The same is today, Core 2 Duo is better, and at 90nm AMD cannot pull any more magic for the Athlon X2/FX line-up. 14 stage pipelines and 90nm at 3GHz just do not go.
zephyrprime
Diamond Member
- Feb 18, 2001
- 7,512
- 2
- 81
64bit is early to the desktop but it's not early to the x86 server market which has been a growing and important piece of the market in recent years. More than 4GB of memory in a x86 server is not uncommon and there are many situations where such a machine is a much better value than a Sun or IBM. Since the x86 server chips are essentially the same as their desktop breathren plus a few goodies, I don't think the push to 64bit was made too early. Besides, software support always trails hardware support for these sorts of things by a few years.
You're missing the point...I agree that Intel is doing the right thing. But it's your assumption that it's costing AMD that is incorrect! if there's enough supply to meet demand, and that supply is only 20% of Intel's output, then I don't think demand is anywhere NEAR as high as you are assuming it is...
Ummm...64bit chips were the same or less expensive (at least AMD's were), quad core chips are 100% more expensive!
As to your use of the term "multicore", it's far too general...it won't help anyone who isn't using it professionally, and in that case a 2P dualcore is easily as effective and reduces the effect of a defective chip for mission critical workstations. (being one of those professionals, this is an area I am quite familiar with...). Quad core won't speed up games, rip videos faster, multitask quicker, etc... It will render much quicker (but then so does dual-dual).
Ummm...how is connection via the L2 in C2D "much faster" than the connection via crossbar of AMD?
I think you missed my point here...I wasn't trying to say that in a single quad core scenario one would end up being faster than another, I was saying that your definition of quad core covers more scenarios (like 4P, and 4x4)
I have no idea why you believe this, as shipments for November this year have been confirmed in articles all over the net...they've been in production for months now!
As to the 45nm in Q3 rumour, only HKEPC has said so...if you look at the foils from IDC, you'll note that 45nm is still scheduled for a Q1 08 launch.
I'll take these one at a time, as they are common misconceptions...
In this case, while it's true that AMD paid $5B for ATYT, almost everyone forgets that they also recieve ATI's income as well...that was $2.39B over the last 12 months!
Again, most people forget that AMD has tripled their production capacity over the last year. They currently have sufficient capacity to easily supply 50% of the world's CPUs...and that's before any of the future Fabs are built.
I frankly have no idea what you mean here...if what way does AMD have to rely on IBM?
If you are referring to their co-development of 65nm, 45nm, and 32nm at East Fishkill, that portion was done a very long time ago (AMD was testing their production level 65nm chips in October 05...).
Say what you want about Mike Dell, but there's is no way in Hell he would have made the move unless he were convinced of the viability of AMD's future...he's already lost a fortune to HP in marketshare by waiting for the proof, and I can't see him flip-flopping at this late stage without that assurance.
I would certainly accept 9-12 months for release on 65nm (closer to 11 actually). The question you have to ask is "why?"...this is important because it gives clues for the next node shrink.
If you have 2 guys who see food down the street, and 1 is starving while the other one is full (because he just ate the other guys lunch), then I would bet that the starving guy will get the food first.
Intel had absolutely no choice but to expend every resource they had on both 65nm and Core technology (remember how many projects they had to cancel to get there)...without both they would continue in a death spiral. The 90nm Netburst had hit a hard wall, and even at 65nm they weren't quite competitive. That said, it cost them dearly...for more on this, look very closely at their financials for the year (and follow the downgrades and performance of their stock) and remember that CSI was originally to be launched THIS year.
They absolutely made the right decision, and Otellini is to be congratulated...but it doesn't mean that they're out of the woods.
AMD didn't and doesn't need 65nm even now...don't get me wrong, it would be nice. But AMD is still gaining marketshare, they have new architecture coming in time to counter Intel when they are able to get larger volume on C2D, they have the ability to change-over faster to 65nm than Intel, and they have some very big irons in the fire like Torrenza as backup. They also continue to expand their markets (from 5% to 20% in China alone).
Firstly, AMD is scheduled to ship 45nm ~6 months after Intel (end of Q2 08). But let's look at the things a node shrink can get you (I believe you listed 2 of the 3).
1. Savings (the one you didn't list) - AMD already has a leg up on this as Intel must spend half of it's die on a large cache in order to keep up with AMD's latency advantage from the ODMC and HT.
2. Space - in balance with the savings, the question is what does Intel intend to use this space for? The more that they put on the die, the more expensive it is...so it still needs to be cost effective. Larger caches mask the need for reduced latency, but there is a point of diminishing returns here...
3. Power/heat - I will certainly cede this as an advantage, but there are so many other variable here that predicting it would be very difficult at this point...
If you will recall, AMD had a complete changeover in engineering and manufacturing staff at the end of this period...followed by a changeover in management. In addition, Intel controlled the market with an iron fist (reducing AMD's ability to get capitol for development). In fact, this is why Wells Fargo values a judgement on the anti-trust to be in the $4B range for AMD...
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products- Started by poke01
- Replies: 358
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 22K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
Facebook
Twitter