CPU/Mobo AMD Bulldozer Discussion Thread

[mrcool63]

so in effect you are saying they should shorten the pipelines and increase the l2/l3 cache.

that would imply that with BD architecture they can shift to the 28nm dye.

 

[cranky]

Nope, I think the architecture is fine.

What it needs is higher clocks, more optimised software (remember the early days of SSE2?) and a less leaky CPU. Process improvement should take care of most of these issues, and a helping hand to software developers will help.

28nm is not something BD needs, it is something it needs desperately, if the process works well.

If you look at how long Stars lasted (in one form or the other) it is a fair bet AMD is looking to squeeze 4-5 years out of this microarchitecture. Thread location awareness and leakage/power consumption seem to be where AMD needs most work as of now.

So I'm skipping BD as well, maybe AM3+ and a big quad is what I'll be doing, or SB 2500k with a decent motherboard. The second is about 30% more expensive, so I'll weigh that into the decision as well.

2560x gaming and desktop work (surf, music and a bit of movies, photoshop and pdf) doesn't seem to indicate a beefy CPU anyway.

 

[Aces170]

CS4 vs CS5.1, and different filters.

I would be very surprised if the results were identical.

Hmm so is CS 5 better at distributing workload? I had serious slowdown issues when editing RAW files on my X2 4000+, hence I am inclined for a faster CPU this time around.

 

[thebanik]

There is one more thing.

The desktop CPU market is less than 1% of the total CPUs sold by the industry. Even with 40% removed for embedded and OEM, that's 60% in institutional and server markets. Frankly, neither Intel nor AMD would give a damn if all of us perished tomorrow. CPUs aren't designed for us, so we choose what's best for each other.

You have to understand where BD came from, it was a clocked-up server chip which doesn't work well in single-unit installations like a consumer desktop. Add this to the above paragraph and the picture begins to emerge.

I dont know from where you got the data but latest IDC report the PC CPU market revenue for 2011 at 43billion. I doubt that accounts for 1% or not, but I am sure neither Intel nor AMD would want to lose that share. Maybe you are talking about enthusiast segment and confused it with Desktop,

 

[cranky]

Hmm so is CS 5 better at distributing workload?

Ummm, no idea. I was coming from the simple fact that you can't compare benchmarks done on two different versions of any software.

I dont know from where you got the data

Nor did I.

But the channel + OEM makes up about 90% of the offtake, the number released in secondary retail is usually much, much lower.

What the figure pointed out was the number of people walking into a shop and choosing a CPU was.

There is also some amount retained for restocking and warranty service needs. This is usually very small, to the order of 4-5%

I think the majority of shipments are for notebook.

I'm inclined to agree without even looking at the data.

 

[thebanik]

Nor did I.

But the channel + OEM makes up about 90% of the offtake, the number released in secondary retail is usually much, much lower.

What the figure pointed out was the number of people walking into a shop and choosing a CPU was.

There is also some amount retained for restocking and warranty service needs. This is usually very small, to the order of 4-5%

The desktop CPU market is less than 1% of the total CPUs sold by the industry. Even with 40% removed for embedded and OEM, that's 60% in institutional and server markets.

Comeon now, you are just shooting figures out of your a**, . In a serious thread where you were blaming people for not able to identify their needs or whatever such casual attitude towards numbers doesnt make for a good choice,

 

[mrcool63]

what i am saying is if AMD shifts to a smaller fabrication die won't these leaks decrease as there is a closer proximity of working components.. I might be shooting in the dark here but doesn't that help?
Also shorter pipelines will help decrease power consumption won't they... the present problems arise as you have said due to longer pipelines and the inability of the architecture to perform shorter threads more efficiently..

Even if that proccy (FX 8150) was available for 7-8k, I would yet not touch it, as I leave my comp on most of the time.

Thats a very shortsighted statement.. power footprint is large i agree but not massive... let me point out a few flaws in your argument... the core 2 duo e6400,i5 760, x4 975, the i7 970 and many more have an even bigger power consumption at idle.. what you see as load there is a cinebench rendering which i can say with 99 percent guarantee than nobody will do it 24 hours a day... for 7-8k that proccy is a steal!!

 

[cranky]

Comeon now, you are just shooting figures out of your a**, . In a serious thread where you were blaming people for not able to identify their needs or whatever such casual attitude towards numbers doesn't make for a good choice,

It was a little more involved than that. Do note that IDC, for example, reports unit and revenue only to paid subscribers.

But you can make a pretty good estimate by clubbing together a bunch of data sets.

Let's start with the semiconductor industry as a whole. This generates about $25bn in sales per month (documented on the SIA site). Note this is all sales of all semis. This is roughly $300bn. I find it a bit tough to imagine sales of desktop PC CPUs being 15% of this total at $45bn. Imagine if that were true. You're looking at AMD getting about $9-10bn, if this were true, and only from sales of desktop CPUs given their ~20% share.

Obviously there's some problem with these numbers as AMD is currently at about $6.5bn total revenue, and Intel is at $43 billion. This would mean that they make almost every bit of money only off CPUs, which we know is not true.

The $43bn number (even as a forecast) doesn't sound right.

So let's dig deeper.

AMD's market share tiptoes higher, Intel still ruler of the roost -- Engadget

So we have another figure, this time the pie size is $9.5 billion, a number I can frankly believe easily.

Now let's use this as a base, which makes AMD's desktop CPU business about $2bn and Intel's about $7.5bn. Given that AMD's 'Desktop solutions' space is about $4.8bn including chipsets, embedded and server (hey, don't ask me!!), a fifty-fifty split sounds good.

Let's take a quick look at some proof of channel splits:

Channel Partners Expected to Dominate PC Sales Through 2012-www.itchannelplanet.com

direct marketers will comprise less than five percent of the worldwide channel market by 2012

So we're down to 5%. Obviously this includes entities such as 'white box' providers or your friendly neighborhood assembler.

5% of (let's say) AMD total desktop CPU biz ($2bn, remember) is $100 million.

This is roughly 0.01% of their total revenue ($6.5bn). And it includes assemblers.

So if anything, I overestimated it.

 

[thebanik]

Any statement about the architecture is a fools errand, AMD has got the best of engineers in the world working for them. They have been working on this CPU for atleast 5 Years, and our keyboard warriors after reading few reviews have already figured the problem and suggested solutions. All that is left is to implement it in the next stepping or maybe give it a codename and launch a new more efficient CPU within the next 6 months....Hoorah!!!!! :hap2:

Thats a very shortsighted statement.. power footprint is large i agree but not massive... let me point out a few flaws in your argument... the core 2 duo e6400,i5 760, x4 975, the i7 970 and many more have an even bigger power consumption at idle.. what you see as load there is a cinebench rendering which i can say with 99 percent guarantee than nobody will do it 24 hours a day... for 7-8k that proccy is a steal!!

The flaw with that statement is that you are comparing older generations of Intel CPU's with the latest generation from AMD. If you compare it with latest generation of Intel CPU's then you will find that Power Usage on Idle has increased by 131% and under load by ~160%. All the while not able to match(or just about match) the performance of a CPU which is cheaper (ie 2500K) while being costlier. You have taken a hypothetical statement by Aces to show his disappointment and based your whole argument around that. :no:

--- Updated Post - Automerged ---

The article you linked to from engadget doesnt even mention what kind of Pie that is. Revenue, Profit or something else, below I am linking just two which clearly states the revenue of PC Processors from IDC, there are lot more which will substantiate that claim....

PC processor market up 17.6% in 2011, says IDC

Intel Ends 2010 with 81.0 Percent Share of CPU Market - Broadband & Consumer Electronics Market Research at iSuppli

2010 around 40 Billion $, 2011 around 43 Billion $.

Anyways no point in arguing, the point I was trying to make was, you can easily discount enthusiast segment as not being important, but saying the whole PC segment is not worth it for Companies like Intel, AMD, Nvidia is a bit far fetched is all I would say....

It was a little more involved than that. Do note that IDC, for example, reports unit and revenue only to paid subscribers.

But you can make a pretty good estimate by clubbing together a bunch of data sets.

Let's start with the semiconductor industry as a whole. This generates about $25bn in sales per month (documented on the SIA site). Note this is all sales of all semis. This is roughly $300bn. I find it a bit tough to imagine sales of desktop PC CPUs being 15% of this total at $45bn. Imagine if that were true. You're looking at AMD getting about $9-10bn, if this were true, and only from sales of desktop CPUs given their ~20% share.

Obviously there's some problem with these numbers as AMD is currently at about $6.5bn total revenue, and Intel is at $43 billion. This would mean that they make almost every bit of money only off CPUs, which we know is not true.

The $43bn number (even as a forecast) doesn't sound right.

So let's dig deeper.

AMD's market share tiptoes higher, Intel still ruler of the roost -- Engadget

So we have another figure, this time the pie size is $9.5 billion, a number I can frankly believe easily.

Now let's use this as a base, which makes AMD's desktop CPU business about $2bn and Intel's about $7.5bn. Given that AMD's 'Desktop solutions' space is about $4.8bn including chipsets, embedded and server (hey, don't ask me!!), a fifty-fifty split sounds good.

Let's take a quick look at some proof of channel splits:

Channel Partners Expected to Dominate PC Sales Through 2012-www.itchannelplanet.com

So we're down to 5%. Obviously this includes entities such as 'white box' providers or your friendly neighborhood assembler.

5% of (let's say) AMD total desktop CPU biz ($2bn, remember) is $100 million.

This is roughly 0.01% of their total revenue ($6.5bn). And it includes assemblers.

So if anything, I overestimated it.

 

[mrcool63]

banik dude i was just pointing out the futility of the statement made by aces. power consumption cannot be compared to sandy but i wanted to show that BD is still not as bad as projected. 975BE was released towards the end of phenom II era. so it is not an archaic chip

 

[cranky]

The article you linked to from engadget doesnt even mention what kind of Pie that is. Revenue, Profit or something else

Can't be profit.

Intel's total revenue is $43bn. (annual report, 2010)

AMDs is $6.5bn. (annual report, 2010), their profit was $43mn only.

The $43 bn CPU market (which I can actually believe) may be the total of every kind of processor made by every manufacturer, including the bits that go into washing machines et al.

Definitely not PC processors. It would mean that Intel makes more than 80% of their money on PC CPUs alone. We know this is not true, their embedded and memory business is almost as big.

 

[raj_pol]

Bulldozer Has Arrived: AMD FX-8150 Processor Review - X-bit labs

This is another review of the same. My apologies if it has already been mentioned. I did not go through the entire thread. Basically inspite of those who wish to believe in the contrary I think AMD is a loosing proposition. They have frankly nothing to compete against Intel and it is frankly a little depressing. Just my thoughts. Peace

 

[PoBoy]

now that the dust has settled :

AMD Bulldozer Blues: Performance, Power, Price... which way to fight into the market? by VR-Zone.com

OK, enough complaints on this - what can AMD do with the Bulldozer core as it is now? First, the core seems to be doing very well, including power consumption wise, on lower clocks. So, the mobile versions of Bulldozer, like the one combined with GPU in the 'Trinity' successor to Llano, could do just fine. Second, the same benefits apply to enterprise servers, especially those in virtualised or cloud environments, or even 'throughput' HPC use, where many small threads run at the same time, a lot of them without any FP use.

That's why AMD launches 'Interlagos', a dual-die, 16 core total, chip, to be the first in its Bulldozer server line up. Yes, the core speeds will be lesser, below 3 GHz, but the power usage will be lower, the core density per chip the highest in X86 world, and with enough shared memory bandwidth, all of 4 DDR3-1600 channels per socket, to feed the cores well.

Bulldozer design compromises offer mixed bag for desktop use

... The processor die contains plenty of high-bandwidth HyperTransport hardware for multiprocessor setups (which is actually disabled on desktop parts)...

AMD to Turn to TSMC for ''Bulldozer'' Manufacturing | techPowerUp

AMD is rumored to be seeking ties with TSMC, Taiwan's premier semiconductor manufacturing foundry, for future manufacturing of its "Bulldozer" architecture processors, according to a report by DonanimHaber.

 

[Gannu]

AMD's only hope now could be price revisions the way I see it.

 

[Parix]

AMD's only hope now could be price revisions the way I see it.

true.. AMD has always been the budget choice for the past 3 years.. i guess they need to change the pricing to still remain budget king..

But wonder what will happen to their profitability though.. they must produce BD in large quantities to meet the gap.. but they were having issues with their yields...

 

[mrcool63]

if this chip is priced below a 2500k even by 200-300rs then it is a definite thumbs up in my books

 

[baccilus]

if this chip is priced below a 2500k even by 200-300rs then it is a definite thumbs up in my books

And what about the power efficiency part? What's the difference in the Power consumption between the two chips?

 

[Aces170]

And what about the power efficiency part? What's the difference in the Power consumption between the two chips?

Huge....

Frankly I don't think the chip with its current performance, and power requirements is worth it. The sides have changed from A64 days when Pentium was pushing its net burst architecture, and just claiming higher speeds in certain applications. While totally discounting performance for watt.

The only reason to even consider BD is as Cranky pointed out, a more robust platform as AMD has more options at a lower price-point for the motherboards. However I would skip BD completely and wait for Pile-driver where they have more time to iron out the short comings. However Intel is not lying idle, and Ivy Bridge looks very promising.

Just hope they have cheaper BD "quad-core/hexa core" variants sub 5k. That seems the only price-point where AMD can survive atm.

 

[mrcool63]

it has been shown that the IIPC increases by around 20 percent when you disable 4 cores in BD that is core 2,4,6 and 8. the sharing is supposedly decreasing the IIPC

http://www.xtremesystems.org/forums...ew-(4)-!exclusive!-Excuse-for-1-Threaded-Perf

Check this out.. now are you convinced Aces??? BD is not totally useless its just not fabricated with you iin mind:bleh:

to quote rage3d

FX is not a Sandy Bridge killer, nor is it killed by Sandy Bridge - they're both horned, winged beasts.

 

[mrcool63]

the reviews for 8120 on newegg are quite promising.. check this out

Clocks like mad...assuming your BIOS will let you. From my very limited testing in Windows using AMD Overdrive I hit 4.9 GHz just by cranking the multiplier and the voltage to 1.4125

When it did work, I did see an improvement at stock when compared to my 1090t at stock. Gaming just felt smoother. Also, thanks to Turbo-Core this usually sits at around 3.4Ghz as opposed to 3.1Ghz it's advertised as, which is VERY impressive, unlike the 1090t that would only turbo up from the 3.2Ghz to 3.6Ghz when an app used 2 or 3 cores

but support on asus mobos and their bios is very shady ...

But I just called ASUS tech support today and they will not verbally confirm if the FX CPU's will work in 890FX Motherboards

Errors all over Windows. It kept going between using Aero and then the gross basic version of windows 7. Stressed tested it using AMD OverDrive at stock speeds and the program would crash out and close.

all in all a mixed package