Bulldozer vs SandyBridge

[Cs342]

I'm planning on building a new system in January, but I can't decide whether I should build an AMD or Intel system. Which of these processors do you think will offer the best value for my money and why? Also what are the release dates for these 2 new processor lines, and should I just build a 1366 or AM3 system instead of waiting??

 

[Malloc1024]

So when a new processor is released, how soon afterwords until we can actually build a system using it?

 

[cjl]

So when a new processor is released, how soon afterwords until we can actually build a system using it?

Really soon. I had my current system up and running a week after Bloomfield (Nehalem) came out two years ago.

 

[jf-amd]

We typically are shipping Opteron into large deals before launch. I think we had tens of thousands of processors sold prior to launch.

 

[eyefinity]

He's probably thinking more about the client BD john.

Just out of curiosity, anything at all of the nomenclature of client BD? Phenom again?

 

[jf-amd]

He's probably thinking more about the client BD john.

Just out of curiosity, anything at all of the nomenclature of client BD? Phenom again?

Because the client product is very transactional, they tend to have product available at launch, but I really don't know much about their plans. No idea on the naming.

 

[someguy7]

I'd say the problem with Barcie was that it way underperformed everybody's expectations due to all the pre-launch hype (remember the "dancing in the aisles" stories, plus Allen's interview as documented in the Youtube clip). So when it did debut late, with low clocks and then the TLB bug, and the benchies showed it much lower performance than what we had been lead to believe, it was a huge disappointment.

However I wouldn't call it garbage, plus AMD did get the clocks up and the bug fixed after a few months. In retrospect, AMD should probably have delayed the launch even more and used the extra time to more thoroughly test it. But at the time AMD was used to making a profit, unlikely as that sounds , so obviously they wanted it out the door and earning some return on their investment. And IMO they miscalculated the lasting impression such an underpowered and underperforming debut would have, at least for enthusiasts who are a really hard bunch of nerds to please ..

I disagree, it was garbage. You covered most of the reasons why I think it was garbage so I will just leave it at that. Phenom II on the other hand I love. The last three systems I put together all PII chips.

 

[metallifux]

Because the client product is very transactional, they tend to have product available at launch, but I really don't know much about their plans. No idea on the naming.

Since we have your attention, could i please beg you to push others in your company to release the AM3+ socket earlier than the new BD CPU's. I'm really keen on a good upgrade path and like many others will not buy a new motherboard with a socket that will soon be updated, but i would buy an outdated CPU with a new socket that i can upgrade. I think if the socket specs are ready why not put it out now? It means that people like myself can build a kick ass rig with a 6 core 1100T on an AM3+ board and then sometime next year chuck in a Bulldozer. AMD would make more money out of me and i would have an extremely quick machine for the next few years!

 

[warmon6]

Since we have your attention, could i please beg you to push others in your company to release the AM3+ socket earlier than the new BD CPU's. I'm really keen on a good upgrade path and like many others will not buy a new motherboard with a socket that will soon be updated, but i would buy an outdated CPU with a new socket that i can upgrade. I think if the socket specs are ready why not put it out now? It means that people like myself can build a kick ass rig with a 6 core 1100T on an AM3+ board and then sometime next year chuck in a Bulldozer. AMD would make more money out of me and i would have an extremely quick machine for the next few years!

As much as some of us would like that to happen , You know it wont.

They got plans to release the AM3+ at a particular data and not a moment sooner.

+ anyways, he works in the sever side of things. Meaning if he looking at any sockets, he seeing the G34 sockets a lot more than the AM3+

 

[jf-amd]

Guys, AMD does not build motherboards and we don't get to tell them what to do.

 

[Cs342]

So only Sandy Bridge will have been released by the end of March?? I thought Bulldozer would have been out by then...

 

[Cs342]

When will Bulldozer and Ivy Bridge come out? And if I buy a Core i5/i7 laptop for Christmas will it be a good choice? I don't want to wait for Sandy Bridge laptops.

 

[cheesesubs]

given the smaller cache size/longer stage pipeline bulldozer is a completely wrong design in my book. bulldozer had been confirm to have 22 stage pipeline(at least) with only 16kb data per module(8kb per core) and share 64kb instruction cache(or 16uops trace cache) for entire processor while having gigantic 2mb l2 per module and 12mb share l3 for all cores is opposite to intel's larger dedicate core with 64kb l1 + 256kb l2 per core with 8mb l3 share cache and 14 stage pipeline. intel had learn from their mistake and stick with 14 stage pipeline ever since conroe released. until now nehalem and upcoming sandy bridge are all 14 stage pipeline design while bulldozer is completely opposite in architecture.

there's no telling why amd decide to move to longer instruction pipeline and smaller cache after seeing intel fail with their netburst architecture. one thing for sure they are decide to fight another gigahertz war and multi threading ability after amd fail compete in single thread performance since conroe released in 2006 due to inferior in architecture compare to intel's conroe and nehalem. amd had to make a break through and also make them look different from intel. that is where it give birth of bulldozer. it is mainly design for server market that's purely to run multitasking. however limited to current die size and fabrication the number of core are increase very slow during each generation. first commercial dual core was introduce in 2004 and now we just just start taking advantage from it while quad core and six core still in the air with expansive price and ran very hot by the physical limit. to having more core number without having huge die the only way is to shrink the core itself. to do so the core has to be as small as possible, means the cache must be first to throw a way then the computing unit like mpu and aux will be cut off to maintain the core size and all cache will be "uncore" and run out side the core speed. then to rise the efficiency for their smaller core inside bulldozer the only way to achieve is insanely clock speed(rumor to have 6ghz+). so the stage pipeline will be much longer in order to clock up even higher.

result, a amd version of pentium 4 is stand right front of you. but only different is bulldozer support sunmicrosystem's "net" architect with can issue 3 thread per coreand since orochi is 4 module 8 core then it can issue 24 thread and the single thread performance can be solve through very high clock. i don't know how intel will deal such challenge (back to clock speed + multi thread war)...but intel will definitely keep their p6 for a while...

 

[jf-amd]

I would not say that your assessment is that accurate, but you are entitled to your own opinion.

 

[fazers_on_stun]

When will Bulldozer and Ivy Bridge come out? And if I buy a Core i5/i7 laptop for Christmas will it be a good choice? I don't want to wait for Sandy Bridge laptops.

I would guess we'll see some Sandy laptops out in January. I'm in the market for a new lappy myself, and I plan to wait a month or two. I'm hoping some OEM will make one that can use the onboard GPU for basic stuff with the discrete GPU turned off, and only turn on the discrete GPU when needed for gaming, etc.

 

[fazers_on_stun]

@ cheesesubs: I wouldn't say a longer pipeline is necessarily bad. The reason Netburst got such a bad rep was that its branch prediction wasn't up to the job back in 2000, so a mispredicted branch would cause the pipeline to stall while the bad data was flushed and then reloaded with correct data. In the ensuing decade both Intel and AMD have improved branch prediction considerably.

My guess is that a modern P4 with improved branch prediction, 4-issue decoder, etc. and implemented on 32nm would not be a bad CPU at all. Probably nowhere near SB or BD, but then it might hit that magical 10GHz clockspeed that Intel predicted way back when .

 

[cheesesubs]

@ cheesesubs: I wouldn't say a longer pipeline is necessarily bad. The reason Netburst got such a bad rep was that its branch prediction wasn't up to the job back in 2000, so a mispredicted branch would cause the pipeline to stall while the bad data was flushed and then reloaded with correct data. In the ensuing decade both Intel and AMD have improved branch prediction considerably.

My guess is that a modern P4 with improved branch prediction, 4-issue decoder, etc. and implemented on 32nm would not be a bad CPU at all. Probably nowhere near SB or BD, but then it might hit that magical 10GHz clockspeed that Intel predicted way back when .

well it is depends on how long the pipeline is. there's several rumor of pipeline size between 18, 22, 30, 36. mine theory is 22 stage because longer 30+ pipeline will cost more die space than on 22 pipelines which is not what amd would expect from their die-reduction strategy.

on bulldozer, even with modernized branch prediction improvement it still have clock up higher due to longer pipeline to avoid the instruction latency. branch prediction is just to prevent the bobble create in the pipeline when bad instruction is loaded. but latency is still exist. 5 stage pipeline will have far less latency than common 14 stage pipeline. to reduce that disadvantage is to clock up high enough to minimize the latency impact. in other side the processor that has 5 stage pipeline won't be able to clock up higher. even without data bobble in netburst processor it would have to take higher clock speed to beat shorter pipeline processor in slower clock speed. this is speed vs latency trade off.

disadvantage for sandy bridge from going higher clock is not because of short stage pipeline but the on-die pcie controller. sandy bridge's internal bus speed is extremely narrow 100mhz which similar to pcie's lane speed and worse part is pcie controller become bottleneck that keep processor from having wider bus back in conroe's day and overclock its internal bus. sdb is being handicap by it's bus speed and oc ability. bulldozer has bus wide advantage over it and can clock very high. but it may take beyond 10ghz to beat a 2.4ghz sandy bridge due to poor core feature and smaller cache and cycle latency.

 

[metallifux]

Guys, AMD does not build motherboards and we don't get to tell them what to do.

No, but don't you choose when to release the socket and chipset?

 

[jf-amd]

If an AM3+ can handle an AM3, don't you think the socket will be the same? Can't speak to the chipset.

The goal is never to build products that sit on shelves, the goal is to start shipping products as soon as possible.

 

[Cs342]

It think I'll just buy an i7 laptop, I can't be bothered to wait for Sandy Bridge

 

[metallifux]

If an AM3+ can handle an AM3, don't you think the socket will be the same? Can't speak to the chipset.

The goal is never to build products that sit on shelves, the goal is to start shipping products as soon as possible.

not necessarily it could have a few extra pins.

 

[jf-amd]

It does not.

 

[mdd1963]

so if i build a 1366 rig for christmas/late january will it last me 2 years? i don't think so but just asking...

In my opinion, any decent quad core buit today with an upper mid range video card (GTX570 or ATI 6970) will easily last two years.

However, if you can wait a few weeks for Sandy Bridge, I would consider waiting....

 

[cheesesubs]

In my opinion, any decent quad core buit today with an upper mid range video card (GTX570 or ATI 6970) will easily last two years.

However, if you can wait a few weeks for Sandy Bridge, I would consider waiting....

6970 is no where near gtx 570..only equal to gtx 480.

anyway i don't want to go off topic which i 'm just try to correct you that's all

 

[cjl]

6970 is no where near gtx 570..only equal to gtx 480.

anyway i don't want to go off topic which i 'm just try to correct you that's all

If you're going to bother to correct people, you might try being correct first.

The 6970 is quite competitive with the 570 - they test about equal in most of the reviews, with the 570 seeming to have a slight edge at lower resolutions, and the 6970 having an edge at higher resolutions.

 

[cheesesubs]

with the 570 seeming to have a slight edge at lower resolutions, and the 6970 having an edge at higher resolutions.

yeah...try to compare a 1gb version 6970 to gtx 570 1280mb, it will be an onslaught...like good old day 2900xt did only beat 8800gts 320mb at highest resolution.....when it equip 1GB of ram....320mb vs 1gb!! wow

i don't need to correct myself because amd sells junky graphic card/processor most of time(except 3870 and 6850/6870...they are very good mid range card at its time) or refresh the same product/architecture to milk more money from consumer when they were upper hand in k8's day until core 2 came out. intel/nvidia's tick&tock roadmap will really kill amd let alone these garbage mid-low range market because this forum is full enthusiast that only care raw performance over price/ power consumption.