AMD starts shipping "Bulldozer" Opterons, FX Series delayed

Jos

Posts: 3,073   +97
Staff

AMD yesterday announced that it had begun shipping its first chips based on the new Bulldozer core. Unfortunately for PC desktop enthusiasts, the company was referring to Opteron server parts and not the Zambezi (FX Series) desktop chips we've been waiting for months. Codenamed Interlagos, the former will arrive as part of AMD's Opteron 6200 Series platform for enterprise-class two-socket and four-socket servers.

The new server chip is based on a 32nm fabrication process and is touted as the world's first 16-core processor. Specifically, the Opteron 6200 has eight Bulldozer modules and each module contains two independent integer processors. However, each module only has a single floating point processing unit (FPU) and fetch/decode/execute unit to share between the two processing cores.

The new chips should offer up to 50% more throughput as its current 12-core Opterons while sitting in the same power envelope, according to AMD. In addition, a redesigned memory controller will offer 30% more memory performance when compared to its predecessor and a flexible 256-bit floating-point unit.

AMD said in a statement that initial production of Interlagos processors began in August, and with shipments now underway to OEM customers, Bulldozer-based systems should arrive by the fourth quarter of 2011.

Their desktop counterparts were supposed to launch in Q3 2011 with the FX Series but the company is now saying these will launch in Q4. While AMD didn't offer any explanations for the delay there has been some speculation suggesting the company is struggling to clock Bulldozer cores to speeds that would be competitive with Intel's Core i7 chips.

Permalink to story.

 
I hope BD is a huge breakthrough in power consumption for mobile segment and gains market share in servers as a result of giving more cores for less $. However, on the desktop its future is bleak. By the time it launches in Q4 2011, Ivy Bridge is just 1 quarter away. Given that AMD has been unable to launch lower clocked BD chips in Q1-Q3, it is almost certain that BD will be like Phenom II in that AMD will sell us more cores against Intel because they won't be able to compete in performance per core. I foresee FX-4100 to go against i3, FX-6100 to go against 2500k and FX-8100 to go against 2600k. As such, BD's only shining area is likely to be rendering, video encoding and some 8 threaded workstation apps.

The problem is for majority of users, Quick Sync is good enough to encode video to your tablet or smartphone. So really BD sounds like a very niche product in much the same way that Phenom II X6 was.

For office, general performance, games, SB will clean up due to huge overclocking headroom and insanely superior instructions per clock. Ironically, ALL of that was available since January 11, 2011 or so while AMD "fanbois" continued to spew propaganda of how BD is worth waiting for.....and here we are 9 months later with 0 to show for BD.
 
This is disappointing news but numerous sources have been hinting at this delay for the past few weeks. Now we wait and see if Bulldozer will in fact be able to perform anywhere within the range of Sandy Bridge. Suspect that many of those that have been waiting on this processor will now switch over to Intel or wait for Sandy Bridge E.
 
Why is there a excavator on the chip? Why not stick a John Deere or a big Massy on the side of the yoke.
 
This is disappointing news but numerous sources have been hinting at this delay for the past few weeks. Now we wait and see if Bulldozer will in fact be able to perform anywhere within the range of Sandy Bridge. Suspect that many of those that have been waiting on this processor will now switch over to Intel or wait for Sandy Bridge E.

On that note...
Sandy Bridge-E has it's coming-out ball on the 23rd October....and...
Scuttlebutt release date 15th November.

Looks like BD's time in the spotlight is going to be very short indeed.

Add in that
1. There seems to be a bug/coding issue with BD that can't be fixed until after the CPU's launch -which might take the gloss of some review benches, and...
2. The initial lineup are sceduled to go EOL between three and six months after launch, so obviously (it would seem) there are some real issues with the launch stepping.

Shaping up to be a very interesting holiday season.
 
SB-E is not going to compete with BD though for 2 reasons:

1) SB-E's cheapest CPU will be > $300, motherboard > $200, quad-channel ram, etc. Also, you would be stupid to purchase a SB-E quad core. So that means realistically the cheapest worthwhile SB-E is $500+!!!

2) BD was never meant to compete with SB-E since it targets Sub $300 prices. So you are either trolling or have no idea what BD's purpose is.
 
There is a 3rd point why SB-E is irrelevant:

According to Intel's internal estimates, Sandy Bridge E-series microprocessors will account for about 1% - 2% of Intel's desktop processor shipments by volume in 2H 2011. By contrast, Sandy Bridge chips for mainstream PCs will represent a half of Intel's desktop shipments in the second half of 2011.
 
SB-E is not going to compete with BD though for 2 reasons:
1) SB-E's cheapest CPU will be > $300, motherboard > $200, quad-channel ram, etc. Also, you would be stupid to purchase a SB-E quad core. So that means realistically the cheapest worthwhile SB-E is $500+!!!*
2) BD was never meant to compete with SB-E since it targets Sub $300 prices. So you are either trolling or have no idea what BD's purpose is.
Sandy Bridge-E i7-3820.....$294

16GB (4x4) of DDR3-1600 CL7....$105

Guest trolling....priceless.

Oh, forgot about the enthusiast grade (>$200) motherboard- can't forget about that:

$240 and $250

* Newsflash junior, BD will likely be ~$300 (top bin)
BD+8GB RAM+ 990FX board ($150-310 = $500 and up.

Oh, and "Guest" - Thanks for the textbook AMD fanboy rant - equal parts smugness, whining and comprehension fail. Re-read my post and point out where I placed BD and SB-E in the same market segment. I think you'll find that what I wrote was that BD's PR longevity is going to be foreshortened due to the imminent release of a competitiors CPU and chipset.
 
Boy oh Boy I hope Ivy Bridge comes out fast and has great overhead room for overclocking. I think Intel is really stretching the market here with its SB-E's and their high prices. I dont think I can fork out top dollar for something thats going to be demolished 4 or 6 months later. I suppose to buy SB was when it came out or when the K series came out.
 
"Sandy Bridge-E i7-3820.....$294"

No unlocked multiplier. Have fun overclocking using BLK on your $300 LGA board. Can you provide any reasons why anyone would be dumb enough to pay $294 for 3820 without an unlocked multi when you can get 2600k + $150 top notch Asrock Extreme 4 board?

"16GB (4x4) of DDR3-1600 CL7....$105"

Ok, why would we spend $50+ extra over 8GB of RAM? Good one. Performance benefit = 0. Here we go again, $50 extra wasted on RAM you won't use (ala Core i7 920 LGA1366 vs. i7 860).

"Oh, forgot about the enthusiast grade (>$200) motherboard- can't forget about that: $240 and $250"

Who are you kidding? You want 14 SATA ports? 10 USB ports for $150-200 more? Not only does a $150-160 LGA1155 have everything you need and the kitchen sink, but you provided top of the line series for AM3+ line. Guess what the LGA1155 equivalents of the boards you linked are:

Asus P67 Maximum IV = $330 http://www.newegg.com/Product/Product.aspx?Item=N82E16813131700

Asus Z68 Maximus IV = $360 http://www.newegg.com/Product/Product.aspx?Item=N82E16813131760

Gigabyte Z68 UD7 = $350 http://www.newegg.com/Product/Product.aspx?Item=N82E16813128499

So LGA2011 UD7 and Maximum series will be >$400.

So let's add this up:

- $50 extra for wasted RAM
- $200-250 extra over LGA1155 Extreme 4 Gen3 board that has it all
- A processor that isn't better than 2600k
- Top of the line air cooling heatsinks clearance limitations for memory since you will have 2 banks to the left of LGA socket and 2 banks to the right (that pretty much discounts Silver Arrow, NH-D14, etc. leaving you either buying noise H50/60/70/80 kits of $100 H100 kit.
================================
+ $250-300 extra over pretty much identical 2600k setup

No thanks, that's like getting a "free" HD7950.

Nice try though. LGA2011 ONLY makes sense for 6 cores ($583+ CPUs).
 
Yup, comprehension fail part deux.

Re-read the last paragraph of post #13, then re-read post #10.

And you could well be right....Buying Sandy Bridge ( and why bother with a 2600K when hyperthreading is of such limited value) 2500K + mainstream board beats all comers. Let's cancel the BD and SB-E party.

Don't forget the emails to AMD and Intel telling them not to bother.

BTW...X79 baseclock overclocking options : 100MHz (standard), 125MHz, 166MHz, 200MHz and 250MHz.. multiply those by the nominal 36 multiplier of a i7-3820. I'll leave you to the math if you can. Oh, and here's a pretty picture if you need to have the info in pictoral form :
http://limages.vr-zone.net/body/13175/sandy_bridge_e_oc.jpg.jpeg

We wont mention the three and four-way graphics card support of X79 or the DDR3-2666 support since enthusiasts don't require multi-card or high bandwidth.
 
All these Bulldozer launch delays are really bad PR for AMD considering how far behind Intel they are in terms of CPU performance. They have better hit this one out of the ballpark if they are to be least competitive with the current Intel line-up. Intel, on the other hand, has not even bothered to drop the prices of its flagship Sandy Bridge processors which is pretty indicative of their confidence that their performance lead will not be threatened by AMD's coming launch.
 
On that note...
Sandy Bridge-E has it's coming-out ball on the 23rd October....and...
Scuttlebutt release date 15th November.

Looks like BD's time in the spotlight is going to be very short indeed.

Add in that
1. There seems to be a bug/coding issue with BD that can't be fixed until after the CPU's launch -which might take the gloss of some review benches, and...
2. The initial lineup are sceduled to go EOL between three and six months after launch, so obviously (it would seem) there are some real issues with the launch stepping.

Shaping up to be a very interesting holiday season.

AMD has clearly missed their window of opportunity to make a splash with Bulldozer. If in fact the FX series can't perform on par with Sandy Bridge this launch will start with a resounding thud. At this point I'm pretty much resigned to waiting for the 8170P in the hopes that the outstanding issues will be resolved. Looks like AMD will once again have to play the price / performance game.
 
Four way videocard support? Who cares about that? 0.0001% of consumers?

Looks to me like you missed the key point I started: LGA2011 ONLY makes sense for 6 cores or higher. Hence, it will not compete with BD since a 6 core SB-E is $580+.

Yet, you continue to spew nonsense regarding 4 GPU support. No one is going to buy 4x GTX680s or HD7970s ($2k in GPUs) and pair it with a $294 lower end SB-E CPU.

Not to mention, good luck getting X79 revision 1 with PCIe 3.0:

http://techreport.com/discussions.x/21613

What do I care? Go ahead and spend $$ on LGA2011 and get that $300 SB-E 4 core. It's not my $. I'll be laughing when your $200-250 EXTRA "futureproofing" motherboard and more ram investment makes 0 difference when IB debutes or when I can reinvest that $200-250 towards a Haswell upgrade. For now the "low end" LGA1155 makes SB-E irrelevant for anything but a 6-core setup. Good thing you waited almost 12 months until crippled PCIe 2.0 X79 $350 boards arrive when you could have been enjoying 2600k all this time.

Sounds to me like you buy for E-peen because a true enthusiast of your 'caliber' would have purchased a 2500-/2600k already and upgraded to a 6-core the day LGA2011 debutes.

Are you also one of those guys who thinks quad-channel memory bandwidth will make any difference for a quad core SB-E in real world ? LOL
 
AMD has clearly missed their window of opportunity to make a splash with Bulldozer. If in fact the FX series can't perform on par with Sandy Bridge this launch will start with a resounding thud. At this point I'm pretty much resigned to waiting for the 8170P in the hopes that the outstanding issues will be resolved. Looks like AMD will once again have to play the price / performance game.
That probably comes down to corporate strategy to a degree. At some point, it was decided that Llano should take precedence in time-to-market (probably the reason that Dirk and a raft of other execs disappeared). Since GloFo has limited 32nm wafer capacity it probably boiled down to Llano or BD production - Llano's higher demand probably looks better to the bean counters.
Factor in that GloFo apparently has some yield issues (note Llano's moderately slow ramp) which would have a knock-on effect with BD if Llano contracts to OEM's (esp. laptops where a missing APU means processor-less laptops stacking up in inventory) needed to be fulfilled first.

All in all, I wouldn't think the BD is going to be very far removed -performance or cost- from Sandy Bridge, so should still be a very capable performer when it arrives. My point, that you've also noted ( overlooked by some anonymous special needs contributor) is that AMD aren't going to realise the full marketing/PR value of the CPU's due to (probably) unfortunate timing.
 
stan4 said:
Poor AMD fanboys, hang tight, hold to your 1090t as long as you can, LOL.

Poor Intel fanboys who really has no clue that it's just computer processors not the Super Bowl.
 
dividedbyzero,

Are you going to be a real man and accept that you were wrong on all accounts regarding SB-E?

1) Lacking USB 3.0
2) Only 2x SATA 6 ports natively
3) Lacking PCIe 3.0
4) Quad-channel memory bandwidth has no impact on desktop performance
5) The only CPU that makes sense in the entire lineup is the 3930K (a hexa core); the 3820 makes NO sense

http://www.tomshardware.com/reviews/core-i7-3960x-x79-performance,3026-3.html

Based on all of these factors, the LGA2011 platform Revision 1 is a failure against 2600k / LGA1155 setup unless you really feel like dropping $600 on the 3930K.

Also, it's 100% obvious that LGA2011 is positioned for server/workstation market and won't compete with AMD's next generation CPU in any shape or form, and esp. not when it costs nearly 3x more!
 
Back