Bulldozer, Massively unoptimized

[beginner99]

lol. nice marketing. Like 24 GB RAM for Xeon compared to 64 GB of RAM for BD does not have any influence on the results...

 

[SickBeast]

This website has gone commercial.

There used to be good Linux and Server articles. Now there's tons of Apple and Microsoft stuff instead.

If Bulldozer actually *is* a good server CPU, I would like to see a good article on it. The same goes for workstation performance.

Those 2 billion transistors have to be doing *something*, and I'm sure that there are apps out there that can push them, perhaps even to the point of the power consumption being a non-issue.

 

[exar333]

More to read on the Opteron:
http://hexus.net/tech/news/cpu/32595-amd-bulldozers-raison-detre/

AMD announced today the immediate launch and availability of its Opteron 6200 and 4200 series of Bulldozer-based server processors, previously codenamed Interlagos and Valencia, designed for Socket G34 and C32 motherboards respectively.

The new silicon offers exceptional value for money for those that need the horse-power, with AMD claiming up to 84 per cent greater performance and 73 per cent greater memory bandwidth for its Operton 6276 when benched against the Intel Xeon X5670, made possible thanks to the Inerlagos' monstrous, hexadeca-16-core design. Quite rightfully, by cramming such performance in to a single chip, cost effectiveness for large server farms goes through-the-roof as you require fewer racks to achieve comparable performance, with AMD also claiming that the technology requires two third's less floor space and money, in comparison to the Intel Xeon 5600 series.

AMD is also claiming the greatest performance-per-core with its 4200 EE series. The math works out at 4.375W per core, with the lowest comparable Intel server offering at 10W per core, the Xeon L5630. The firm has also announced that the new line of Opterons are the only server processors currently supporting ultra-low 1.25v memory.
Initial pricing appears more than competitive, with, for example, 1k unit costs for the AMD Opteron 6276 at $788, compared to $774 for an Xeon E5640. AMD's frontrunner will be the Opteron 6282 SE 16-core, 4-way Interlagos Bulldozer, clocking in at 2.6GHz per core, with a turbo frequency of 3.3GHz, 16MB level three cache and a 140 Watt TDP, with a 1k unit cost of $1019.

It's about time to see AMD's Bulldozer architecture finally do what it was made for, which as we've always suspected, is highly-threaded server work; with AMD's initial figures seeming to show that it does this job exceptionally well. This new product line does, however, give us an interesting insight into the large scalability of the Bulldozer design in general; proof that it can utilise low-power memory, range from 4 to 16 cores, scale from TDPs of 35W to 140W and from previous over-clocking evidence, scale incredibly well from low high-core-count server frequencies up to record-breaking single module frequencies. Along with our prior post on scheduler performance optimisations, perhaps this all shows that there's plenty more left to come out from the Bulldozer architecture.

Doesn't really seem that impressive IMHO. You are talking about 16C vs 6C (166% more cores) for only AT BEST 84% more performance. Probably more like 30-40% average performance gains, as the 84% is likely cherry-picked. The clocks are similar (2.6BD vs. 2.93Xeon).

Also, power usage per core is pretty meaningless when each of your cores is only 1/3 the power of the competition. Let's say a Gulftown Xeon = 2 BD cores (this is pretty accurate). The power usage per core is then very close...

Once BS-based Xeons replace the existing Gulftowns, expect performance gains to be very solid on the Xeon side, plus an increase of 10% or so in clocks. That will translate to likely somewhere between 20-25% total performance improvement with better thermals. I just don't see the BD Interlagos being a great option unless price is the only thing that matters to you.

 

[AtenRa]

OK guys, my review (AMD Opteron 6200 and 4200 Series) is up

njoy

Link in my sig

 

[Phynaz]

Posting AMD slides in your blog is a review?

 

[formulav8]

Edit: I kinda see what you mean :\

 

[exar333]

OK guys, my review (AMD Opteron 6200 and 4200 Series) is up

njoy

Link in my sig

I think the AT review says enough.

16C @ 2.3ghz(BD) vs 12C @ 2.2ghz(Magney)

BD is 4% better. Pathetic.

 

[AtenRa]

I think the AT review says enough.

16C @ 2.3ghz(BD) vs 12C @ 2.2ghz(Magney)

BD is 4% better. Pathetic.

Sorry to say but the workloads at AT review are not what the new Opterons were created for.

We already know from the desktop Zambezi that Rendering is not what those CPUs were designed for.

 

[exar333]

Sorry to say but the workloads at AT review are not what the new Opterons were created for.

We already know from the desktop Zambezi that Rendering is not what those CPUs were designed for.

So more clock speed and 33% more cores is acceptable? Only an idiot would say that is an improvement.

 

[Idontcare]

So more clock speed and 33% more cores is acceptable? Only an idiot would say that is an improvement.

Surely there is a more cordial way for you to make your point without resorting to equating individuals who disagree with it as being "idiots", yes?

 

[SickBeast]

Sorry to say but the workloads at AT review are not what the new Opterons were created for.

We already know from the desktop Zambezi that Rendering is not what those CPUs were designed for.

AtenRa, what then was this CPU designed for? Servers? What type of servers?

It seems pretty apparent that AMD messed up with this design. It will serve a niche market for the time being.

 

[taltamir]

OK guys, my review (AMD Opteron 6200 and 4200 Series) is up

that wasn't a review, I just finished reading and you just rehashed AMD marketing data. Nowhere did I see you perform a single test of either CPU. I strongly suspect you don't have said CPUs and simply use AMDs published marketing claims as fact to draw conclusions from.

 

[SickBeast]

This thread is dying.

 

[Smartazz]

While transistor counts aren't everything, it's interesting to see that both SB-E and Bulldozer both have 8 cores on die and 2 billion transistors. I'm certain that an 8-core Sandy Bridge would thrash Bulldozer.

 

[Idontcare]

This thread is dying.

Its not dying, its just massively unoptimized. :sneaky:

While transistor counts aren't everything, it's interesting to see that both SB-E and Bulldozer both have 8 cores on die and 2 billion transistors. I'm certain that an 8-core Sandy Bridge would thrash Bulldozer.

1-part superior process tech, 1-part superior microarchitecture, all bought and paid for by a generously larger R&D budget combined with an employee moral that only job security and the comfort of knowing there are billions of profit per quarter can bring to your everyday Intel engineer.

 

[blckgrffn]

Its not dying, its just massively unoptimized. :sneaky:



1-part superior process tech, 1-part superior microarchitecture, all bought and paid for by a generously larger R&D budget combined with an employee moral that only job security and the comfort of knowing there are billions of profit per quarter can bring to your everyday Intel engineer.

What's interesting is that AMD remains remotely relevant.

There could be some uses for BD in certain cases - facebook and memcached come to mind.

It's in our best interest for AMD to stick around, hating them for not being Intel is likely a shortsighted way of looking at things.

This isn't meant to be an apology for AMD... I'm just saying

Also, as I dig into distributed computing here @ Anandtech, I think it will be very interesting how AMD can make inroads there. Surely even a single socket Interlagos is going to crush some WU's... Let's hope piledriver brings reasonable power consumption (manufacturing maturity) and single core umph (architectural refinement) to the table. It is possible that we will get both.

 

[dac7nco]

Its not dying, its just massively unoptimized. :sneaky:

Nice. A colder fact is this one: why is Interlagos (16c) cheaper than Magny-Cours (12c)? Interlagos doesn't sound french, so admittedly I want to buy it, but it, unfortunately, may swirl AMD into the shitter. Johan isn't an idiot. Anandtech may not be a place for serious IT people to hang, but it didn't pretend to be. If you're in the market for serious HPC gear, the application/AIW solution vendors are where you shop - not a place where 95% of us think $999 for a CPU is overkill. HPC guys think about the power bill, infrastructure, power bill.....

Cray has changed hands more often than I have fingers, which amazes me, because when they decided to build a giant AMD cluster, the rest of us said: SO? Cray who? (It's the same as DEC who, and Thinking Machines who, and MIPS who.......... and Sun and SGI and KSR who.)

I'm not terribly worried that AMD may go down. To me, it's like a girlfriend who miscarries due to drugs, who I was leaving anyway. Wow, that sounds harsh.

 

[ocre]

I'm not terribly worried that AMD may go down. To me, it's like a girlfriend who miscarries due to drugs, who I was leaving anyway. Wow, that sounds harsh.

Daddy?

 

[AtenRa]

Posting AMD slides in your blog is a review?

that wasn't a review...

Sorry guys, just realized that.

My mistake, I have corrected it (Preview).

 

[AtenRa]

AtenRa, what then was this CPU designed for? Servers? What type of servers?

Data centers, Virtualization, Web/Clound and HPC.

Problem is, most of the people here at AT forums are thinking in Desktop terms and conditions. That is Higher performance no matter what, something that is not applied to the entire Server market or if you prefer, the Server market AMD's new Server CPUs were made for.

 

[sm625]

Data centers, Virtualization, Web/Clound and HPC.

I think that about sums up what BD is designed for. Clowns

 

[busydude]

I think that about sums up what BD is designed for. Clowns

You guys need to take a break. Judging a person based on the CPU they purchase.. seriously?

 

[taltamir]

Sorry guys, just realized that.

My mistake, I have corrected it (Preview).

Fair enough. It was presented well enough that I will read the actual review when you manage to get your hands on the actual CPUs.

 

[Idontcare]

You guys need to take a break. Judging a person based on the CPU they purchase.. seriously?

That's something only an owner of a Via Nano would say :\ how lame :sneaky:







j/k

 

[Dadofamunky]

Its not dying, its just massively unoptimized. :sneaky:

Nice. A colder fact is this one: why is Interlagos (16c) cheaper than Magny-Cours (12c)? Interlagos doesn't sound french, so admittedly I want to buy it, but it, unfortunately, may swirl AMD into the shitter. Johan isn't an idiot. Anandtech may not be a place for serious IT people to hang, but it didn't pretend to be. If you're in the market for serious HPC gear, the application/AIW solution vendors are where you shop - not a place where 95% of us think $999 for a CPU is overkill. HPC guys think about the power bill, infrastructure, power bill.....

Cray has changed hands more often than I have fingers, which amazes me, because when they decided to build a giant AMD cluster, the rest of us said: SO? Cray who? (It's the same as DEC who, and Thinking Machines who, and MIPS who.......... and Sun and SGI and KSR who.)

I'm not terribly worried that AMD may go down. To me, it's like a girlfriend who miscarries due to drugs, who I was leaving anyway. Wow, that sounds harsh.

Speaking of which, even SGI still exists!

Surely there is a more cordial way for you to make your point without resorting to equating individuals who disagree with it as being "idiots", yes?

:thumbsup: :thumbsup: