Things are looking better for AMD

[mosox]

AMD is very competitive and will be so even if Intel releases 3 more generations like that before the Bulldozer. There are 1 billion computers in the world most of them in underdeveloped/developing countries. The number increases fast. No way those people (I live in E Europe and I know what I'm saying) would EVER cough up $400 for a CPU/mobo. They're selling Raidmax and Delux PSUs in here like hot cakes. And all the bunch wants to play Crysis and whatnot - in India, China, Africa, etc etc. That's all.

IF the AMD guys are smart they would leave the performance crown to Intel who knows best in higher end CPUs and do their thing. That's cheap CPUs. Cheap quads. Triple cores. Socket stability (very good for budget builders). The works.

Do you think that someone who is running an Athlon II X3 is crying now because of the Sandy Bridge? The dude got his CPU/mobo/RAM for less $$ than a quad from Intel. What $75 Intel is better? What $100 Intel CPU is better than the Athlon II X4?

Let Intel make the cake, AMD should make bread.

 

[Nemesis 1]

It's better than your endless pro-intel drivel.

Where's your larrabee now, tough guy?



:thumbsup::thumbsup:

Well knights ferry be in developers hands hence the ferry word . Knights corner is for the HPC crowed. on 22nm. Hence the corner word . Be nice feed the birds .

 

[Soleron]

What we actually know about Bulldozer on the desktop (confirmed facts):

- Intel will only go up to 6 SB cores on the desktop; AMD will have 8 BD cores in silicon smaller than Istanbul
- AMD gets to go to 32nm HKMG from 45nm
- Bulldozer will have higher IPC per core than Phenom II (according to JF)
- Bulldozer is targeted at very high clockspeeds, 3.5GHz+ and many changes from Phenom II to allow it to clock higher such as pipeline length
- Bulldozer's Turbo is improved relative to Phenom II
- Bulldozer implements instructions that Intel previously had but AMD didn't

All of these factors together make it seem likely to me that an 8-core BD will be competitive with a 6-core SB, even if BD isn't a revolution. For that to not happen BD would have to be a Barcelona and have a publicised bug and much lower clocks than expected.

That fair?

 

[JFAMD]

Yes, those statements are all true.

You forgot some new instructions that Intel won't support for a few years - if at all.

But everyone is just speculating at this point because people have neither in their hands. It will be interesting when both are able to compete head to head, but until that time, all of this is just speculation.

 

[Ajay]

Yes, those statements are all true.

You forgot some new instructions that Intel won't support for a few years - if at all.

But everyone is just speculating at this point because people have neither in their hands. It will be interesting when both are able to compete head to head, but until that time, all of this is just speculation.

But are any compiler companies (i.e. M$) going to support the new instructions? Will these instructions be used by .NET or DirectX? Obviously, Intel's compiler won't.

 

[JFAMD]

Well, there is already support for SSSE3, SSE4.1, SSE4.2, AES acceleration and PCLMULQDQ in all of the current stuff.

Software will need to be updated to support AVX but that will be the same for intel and AMD.

XOP and FMA4 are the 2 that intel is not, at this point, implementing. We are obviously working with the tools vendors, but I cannot comment on their products because I do not work for them and I am pretty sure they would not want me making statements on behalf of their products.

But if you think about it, FMA4 is going to be pretty powerful and there will be tools vendors that will really want to take advantage of that instruction set because it will give them an advantage in the market (compiled code will run a lot faster).

 

[Dadofamunky]

Sandy Bridge brings high-end performance to the mainstream segment. To me this is a huge advance. .0001% (figuratively) of users have a Gulftown. Now you can have the Gulftown equivalent for the price of a Micro Center CPU special. That's not the $50 market like you get in the cheap laptops, but it locks AMD out of competing in that area and ensures that no one will think of AMD as anything except the low-cost low-performance vendor.

For my standpoint, this pretty much buries AMD as a possibility for me anytime soon. Intel is destroying them on the manufacturing side these days. I just hope AMD finds a way to stay alive, or we'll be playing "monopoly" again in the CPU market. Bulldozer? Please. We've heard it all before.

 

[Nemesis 1]

Man I be the first to ask I guess . What exactly is it that SB 2011 will have that SB1155 doesn't have? We know 1155 has IGP . But what does 2011 SB have other than more cores than SB 1155, I mean people are making SB 2011 out to be a new Arch . I am sure there will be something extra for server market , But were is all this performance going to come from other than more cores. I see SB 2011 as a bit of a let down if its built up to be faster than present SB1155. If its faster because of more cores big deal. 2600K does 8 threads now.

 

[Nemesis 1]

Sandy Bridge brings high-end performance to the mainstream segment. To me this is a <b>huge</b> advance. .0001% (figuratively) of users have a Gulftown. Now you can have the Gulftown equivalent for the price of a Micro Center CPU special. That's not the $50 market like you get in the cheap laptops, but it locks AMD out of competing in that area and ensures that no one will think of AMD as anything except the low-cost low-performance vendor.

For my standpoint, this pretty much buries AMD as a possibility for me anytime soon. Intel is destroying them on the manufacturing side these days. I just hope AMD finds a way to stay alive, or we'll be playing "monopoly" again in the CPU market. Bulldozer? Please. We've heard it all before.

Be carefull! AMD is 1 step ahead of Intel on the gate first thing . If AMD gets good yields There in a good position. If gate first causes low yields than AMD is in a world of hurt. But until those facts are known don't X AMD out of the game just yet.

AMDs new 975 seems to be a nice processor That should be had for $100 shortly . A very good buy at $100.

As for BD cores lets hope they outperform intel . Because as fast as Intel is moving AMD needs a break . It won't be a long lasting victory all things considered. But I am pulling for AMD . Even tho I willl stay with Intel. Choice its a wonderful gift .

 

[SickBeast]

Well knights ferry be in developers hands hence the ferry word . Knights corner is for the HPC crowed. on 22nm. Hence the corner word . Be nice feed the birds .

Please explain your post to me as I do not understand any of it.

I never feed birds BTW. D:

 

[Maximilian]

Please explain your post to me as I do not understand any of it.

I never feed birds BTW. D:

Ye they just keep coming back for more if you feed em :\

 

[grimpr]

Please explain your post to me as I do not understand any of it.

I never feed birds BTW. D:

Nemesis_1 is our dear familiar Intel Supremacist and an AMD stock shorter, you should have figured that a long time ago!

 

[Edrick]

But everyone is just speculating at this point because people have neither in their hands.

True, because AMD is 6 months late to the dance.

 

[Edrick]

- Intel will only go up to 6 SB cores on the desktop

Really? Intel said this directly? Link?

 

[Nemesis 1]

Ye they just keep coming back for more if you feed em :\

Yes they do. I love humming birds there cool alittle sugar water they be happy , Cardinals and blue jays are cool Woodpickers are great. Bird feeders are wonderful .

I just got up and looked . I have 8 deer in the backyard right now . A little corn every night and salt blocks makes for wonderful beings of nature right outside the door. My grandson loves it . Nature something that needs mending.

Even tho I use intel I don't want to see AMD go away. So now we can only hope that AMD staff of engineers comes up with something really good or Apple will buy them . Which is likely to happen anyway now that AMD is fabless.

 

[Nemesis 1]

Please explain your post to me as I do not understand any of it.

I never feed birds BTW. D:

You talk like Larrabee is a dead project . Knights ferry is real its in developers hands . Knights corner is the first retail Chip on 22nm. According to Intel anyway. If Its just HPC thats fine . Seems to me intel could make a GPU based on the IGP they have on SB . Say 96 Eu and more of whatever it takes it could go head to head with NV /AMD

 

[Jovec]

IF the AMD guys are smart they would leave the performance crown to Intel who knows best in higher end CPUs and do their thing. That's cheap CPUs. Cheap quads. Triple cores. Socket stability (very good for budget builders). The works.

AMD sells cheap because they have to, not because they want to. The problem only gets worse as the IPC gap widens - with enough IPC, Intel can sell a (much smaller) 2c/4t CPU that beats a 4c/4t AMD CPU. Intel will probably always have a process advantage, but even at the same process Intel CPUs are smaller (and thus cheaper to manufacture) while being as fast or faster.

While losing the "speed race" might be an issue for AMD's marketing department, from a practical standpoint it really only hurts AMD in the area of gaming performance. For your typical desktop, anything AMD sells you in your price range is fast enough. For your typical server environment, cores scale, and more AMD cores per dollar roughly equates to less, more expense Intel cores. It's not quite that simple, but it's true enough.

I would wager that IPC isn't at the top of AMD's priority list. I think they are more concerned with power consumption for the mobile space, total package cost (eg. SoC, Llano), having a viable product(s) for the ever-expanding mobile markets (netbooks, laptops, tablets, maybe cell phones, etc), and finally keeping die size down.

The one area where AMD has really misjudged (IMHO) is the IGP. They should have continued to put pressure on Intel by making the 800 series 80SP, with a 40SP IGP as OEM only if needed. The vast majority of users were better off with a slower AMD CPU and faster IGP compared to a faster Intel CPU and (much) slower IGP. Now with the Core I3/5 IGP they are both roughly equal. They lost their ace in the hole.

 

[ydnas7]

If 1 x SandyBridge core approximates the 1 x BullDozer module
Then per mm2 Bulldozer will offer superior multi-threaded performance while Sandybridge offers superior single thread performance. (given similar litho...)
A Sandybridge core maximises single thread performance and uses SMT to improve its inferior multithreaded performance
A Bulldozer module maximises multi threaded performance and uses its inverse of 'SMT' to improve single threaded performance

so for the lower 50&#37; of the market (dual cores and lite quad cores) then Bulldozer is probably slightly superior, for the cloud computing/web server Bulldozer will be very competitive.
but for the higher performance individual users, Sandybridge is stronger as it is for businesses and users who really are subject to diminshing returns of increasing core count.

 

[Nemesis 1]

Well I would have to debate the server market with you as AMD loses more and more every Qt. Thats with magna cores. AMD needs a home run . AMD going after the lowend market may pay dividends but only if they survive the storm .

 

[nyker96]

I don't understand how can people here like declare victory for AMD when BD isn't even a real product yet!? And now all that's available from AMD are paper drawing of architectures etc. I mean at best we just DON'T KNOW YET.

 

[JFAMD]

I don't understand how can people here like declare victory for AMD when BD isn't even a real product yet!? And now all that's available from AMD are paper drawing of architectures etc. I mean at best we just DON'T KNOW YET.

It is no different than those that say that intel will always have the advantage, no matter what.

Competition is good.

And in case you are wondering, bulldozer is more than a paper drawing at this point, it has been demonstrated in public already, OEMs are finalizing platforms and it will be launching next quarter.

 

[RussianSensation]

And in case you are wondering, bulldozer is more than a paper drawing at this point, it has been demonstrated in public already, OEMs are finalizing platforms and it will be launching next quarter.

That's fine, but we have literally just seen Intel struggle to get more than 15&#37; IPC improvement from Nehalem in 2008 to SB in 2011 (that's more than 2 years). Nehalem is already 20% faster in IPC compared to Phenom II (Stars). So Bulldozer will need a 35% improvement in IPC just to match SB in IPC. Since Intel also has a manufacturing process advantage, this basically guarantees that AMD is going to compete by selling more cores at similar or even lower prices as they are already doing with X6s. It's a decent strategy to maintain market share but it sure isn't helping them solidify their profit margins. At the same time, they become less and less competitive on the mobile side (which is more important than desktops imo).

IMO, AMD needs to do something very radical. Personally, if they delivered a notebook that guaranteed 8-10 hours of battery life with 30% lower performance than Intel, I bet you tons of university and college students and people in business/on the go would be willing to get that. AMD is neither the leader in performance, nor in power consumption. They need to be a leader in at least 1 of these - and then they will start eating away at Intel's market share by at least capturing attention of consumers that always choose Intel for mobile systems. What Apple has taught us is that you don't necessarily need to get the fastest product on the market to have a great (if not the best) product. If AMD became the company that manufactures notebooks with the longest battery life, this competitive advantage would spread like wildfire!! <<< AMD is already dominating in the discrete GPU mobile space with > 60% market share, despite the fact that NV has more powerful GPUs. >> now they just need to make a very power efficient processor.

 

[Makuab]

That's fine, but we have literally just seen Intel struggle to get more than 15% IPC improvement from Nehalem in 2008 to SB in 2011 (that's more than 2 years). Nehalem is already 20% faster in IPC compared to Phenom II (Stars). So Bulldozer will need a 35% improvement in IPC just to match SB in IPC. Since Intel also has a manufacturing process advantage, this basically guarantees that AMD is going to compete by selling more cores at similar or even lower prices as they are already doing with X6s. It's a decent strategy to maintain market share but it sure isn't helping them solidify their profit margins. At the same time, they become less and less competitive on the mobile side (which is more important than desktops imo).

IMO, AMD needs to do something very radical. Personally, if they delivered a notebook that guaranteed 8-10 hours of battery life with 30% lower performance than Intel, I bet you tons of university and college students and people in business/on the go would be willing to get that. AMD is neither the leader in performance, nor in power consumption. They need to be a leader in at least 1 of these - and then they will start eating away at Intel's market share by at least compaturing attention of consumers the always choose Intel for mobile systems. What Apple has taught us is that you don't necessarily need to get the fastest product on the market to have a great (if not the best) product. If AMD became the company that manufacturers notebooks with the longest battery life, this competitive advantage would spread like wildfire!! <<< AMD is already dominating in the discrete GPU mobile space with > 60% market share, despite the fact that NV has more powerful GPUs. >> now they just need to make a very power efficient processor.

The average consumer doesn't known the difference between AMD and Intel.

 

[Castiel]

That's fine, but we have literally just seen Intel struggle to get more than 15% IPC improvement from Nehalem in 2008 to SB in 2011 (that's more than 2 years)

Struggle? Struggle wouldn't be the term i would use for Sandy Bridge. SB is mainstream. Intel didn't have to pull out all the stops and try to obliterate Nehalem and the performance it offers i think is perfect for the market segment that it's targeted too.

 

[Ajay]

Man I be the first to ask I guess . What exactly is it that SB 2011 will have that SB1155 doesn't have? We know 1155 has IGP . But what does 2011 SB have other than more cores than SB 1155, I mean people are making SB 2011 out to be a new Arch . I am sure there will be something extra for server market , But were is all this performance going to come from other than more cores. I see SB 2011 as a bit of a let down if its built up to be faster than present SB1155. If its faster because of more cores big deal. 2600K does 8 threads now.

Right now it seems that there will be support for triple and quad channel memory; probably 30+ PCIe lanes and support for at least a 6 core CPU (maybe 8 cores, if Bulldozer is looking really good). AFAICT, it's really for the high end enthusiast market (benchmarkers, geeks with too much cash to burn, etc.) and Workstations.