Opteron A1100 is shipping

[Vesku]

This isn't competing with generic x86 servers. Big data companies are looking to take some of the wind out of Intel's pricing sails by running a mixed ISA environment. They are targeting relatively narrow datacenter uses atm: memcached and similar RAM hosted applications, storage, and web services.

Same basic reason "OpenPower", diversification of IBM Power platform, came about.

 

[Sweepr]

You forgot performance/watt that favours the Xeons.

So if we look 6 months ahead, how will these A1100 sales been?

AMD didn't even dare to include such a compqrison in their marketing slides.
And 6 months ahead it will be decimated by similar/lower priced 16-core Denverton.

The Xeon will be better at CineBench of course.

Unless you have anything useful to prove otherwise, faster everywhere else too.

 

[ShintaiDK]

Same basic reason "OpenPower", diversification of IBM Power platform, came about.

OpenPower have been anything but a success. So if its like that, then we see the same story as usual. Going nowhere.

 

[krumme]

One can say there is a consolidation on the hosting market. With the big guys (amazon, fb, google) there is more organizational headroom and ressources to differentiate solutions on the server hardware level. Then it makes more sense to operate with far more different hardware platforms, and therefore selecting eg. 1100 as a solution for that type of workload.
One can also say the mobile devices forces the cloud solutions and the above market tendency as a consequence.
We are still in the first step. Like the first x86 server. BM will be interesting, but i think there is much more to it, deciding what is going to happen.

 

[Vesku]

OpenPower have been anything but a success. So if its like that, then we see the same story as usual. Going nowhere.

Different approach looking to address the same concerns of the dominant data companies. For AMD the issue will be continued development, i.e. updating to 14FF with A72 or similar within 12-18 months.

 

[Thala]

For AMD the issue will be continued development, i.e. updating to 14FF with A72 or similar within 12-18 months.

The update might include K12 ARM cores, which supposed to have higher IPC than Zen. At least that was the plan.

 

[AtenRa]

You forgot performance/watt that favours the Xeons and most likely avoton.

So if we look 6 months ahead, how will these A1100 sales been?

From TCO to MT workloads to Perf/watt to sales. What is next ??

 

[ShintaiDK]

From TCO to MT workloads to Perf/watt to sales. What is next ??

You tell me, its you doing the goalpost tapdance as usual.

So if the A1100 cant compete in any CPU load, TCO and performance/watt. Not to mention its ARM based with all its limitations. Then what will sell it?

What will make this the next AMD golden cow?

Even your beloved company says it cant compete with Xeon-D.

 

[ShintaiDK]

The update might include K12 ARM cores, which supposed to have higher IPC than Zen. At least that was the plan.

Another AMD dream product made on a shoehorn budget in the same timeframe?

 

[Arachnotronic]

Can't believe anybody is defending Seattle at this point. It is DOA. I guess when a product is uncompetitive and essentially garbage, it gets called a "development platform". Spin at its finest.

This is the same crap that Intel did with Knights Ferry, calling it a "development platform" too when in reality it got thrashed by modern GPGPUs making it utterly unsaleable.

 

[Vesku]

Can't believe anybody is defending Seattle at this point. It is DOA. I guess when a product is uncompetitive and essentially garbage, it gets called a "development platform". Spin at its finest.

This is the same crap that Intel did with Knights Ferry, calling it a "development platform" too when in reality it got thrashed by modern GPGPUs making it utterly unsaleable.

It's always been a development platform first with transition to normal product. The funny bit has been the multiple "now available for testing" launches. Challenge of making sure it could work without requiring a single custom Linux distribution was vastly underestimated apparently.

 

[mrmt]

Xeon D is in a different price range. Avoton and presumably its successor lines are actually in the same price tier.

What do you mean by different price range? Orders of magnitude bigger or a few percent bigger? If not it's better to check your premises because the acquisition costs won't matter as much as TCO costs.

This isn't competing with generic x86 servers. Big data companies are looking to take some of the wind out of Intel's pricing sails by running a mixed ISA environment. They are targeting relatively narrow datacenter uses atm: memcached and similar RAM hosted applications, storage, and web services.

Same basic reason "OpenPower", diversification of IBM Power platform, came about.

The "basic reason" for OpenPower is IBM not to take wind off Intel sails, but IBM itself wanting to put money on POWER development. POWER is being crushed, with revenue shrinking each generation and development costs raising each generation, so it's either open up or die.

It was the same with their foundry, when they hit the same dilemma they opened up and built the common platform. Not only this didn't grant the survival of the IBM foundry business it also didn't muster enough critical mass for continuous development. It postponed the inevitable fate of IBM foundry business for one or two generations.

 

[mrmt]

You tell me, its you doing the goalpost tapdance as usual.

Don't go down this rabbit hole. Our resident AMD reseller/agitprop does't have a clue on what TCO is.

 

[krumme]

It's always been a development platform first with transition to normal product. The funny bit has been the multiple "now available for testing" launches. Challenge of making sure it could work without requiring a single custom Linux distribution was vastly underestimated apparently.

Its kind of a soft launch for K12 one can say Looking at the difficulties it sure already had its rights. One can hope bringing it to market this late is an indication K12 is still very much strong and alive.

 

[krumme]

Intel killed nvidia entire mobile strategy paying customers to use atom. And the killing was just a sideeffect. Collateral damage.
It will be an easy task to do the same for select segments and customers for a small part of the server market against an economic weak oponent as amd.

Amd needs something that is simply fastest. Sounds impossible but if they want to make a dent in this conservative and risk adverse monopoly market thats whats is demanded.

 

[Vesku]

What do you mean by different price range? Orders of magnitude bigger or a few percent bigger? If not it's better to check your premises because the acquisition costs won't matter as much as TCO costs.

Xeon D board ~= $800, Avoton and presumably A1100 (given $150 chip price) ~= $400.

Not sure where the idea $100s per unit is only a minor blip on TCO comes from, the performance/watt differences aren't that drastic for these products. It's not as if they are keeping these units for decades.

 

[Arachnotronic]

Intel killed nvidia entire mobile strategy paying customers to use atom. And the killing was just a sideeffect. Collateral damage.
It will be an easy task to do the same for select segments and customers for a small part of the server market against an economic weak oponent as amd.

Amd needs something that is simply fastest. Sounds impossible but if they want to make a dent in this conservative and risk adverse monopoly market thats whats is demanded.

NVIDIA's mobile exit had nothing to do with Atom.

 

[AtenRa]

does't have a clue on what TCO is.

Where is that TCO analysis on A1100 you are talking about ???

 

[krumme]

NVIDIA's mobile exit had nothing to do with Atom.

Well nv had to exit anyway - as Intel will have to. Atom just made it go faster. Denver didnt get a fair chance. More time and 16nm would have suited it. But the money canon prevented that

 

[krumme]

Where is that TCO analysis on A1100 you are talking about ???

Tco is not as simple as raw hardware cost and running cost vs perf. A solid tco calculation is a cost in itself. How much valuable top management time is available for IT. Hardware plays zero role here.
You need bulldozer p4 vs k7 on soi to get to a situation where it moves the market radically. And even then we have Michael from Dell calling Otellini for money to flatten the effect - so to speak.

A1100 makes no difference on the market.

 

[Shivansps]

It cost more, it consumes more and it also has higher TDP than the A1100.

All the above make TCO higher than A1100, so now TCO is not a metric ???

Thats also true when comparing A1100 to Avotons if the cost is about 300-400

 

[Vesku]

Tco is not as simple as raw hardware cost and running cost vs perf. A solid tco calculation is a cost in itself. How much valuable top management time is available for IT. Hardware plays zero role here.
You need bulldozer p4 vs k7 on soi to get to a situation where it moves the market radically. And even then we have Michael from Dell calling Otellini for money to flatten the effect - so to speak.

A1100 makes no difference on the market.

Big data have specific tasks they are looking to have multiple ISA support in. They want this multiple ISA support largely as part of a strategy to reduce unit costs. They are running many thousands of units they need to update at a roughly 3-5 year cycle.

As an example, Facebook was quiet pleased ditching a few dollars in ICs that were on generic server boards as part of their Open Compute project.

AFAIK, no company is yet trying to take the "ARM on Desktop" strategy that would be necessary to compete with x86 in "regular" IT roles.

 

[Phynaz]

Intel killed nvidia entire mobile strategy paying customers to use atom. And the killing was just a sideeffect. Collateral damage.
It will be an easy task to do the same for select segments and customers for a small part of the server market against an economic weak oponent as amd.

Amd needs something that is simply fastest. Sounds impossible but if they want to make a dent in this conservative and risk adverse monopoly market thats whats is demanded.

Crappy CPUs killed Nvidia's mobile strategy. Did you try the first few go arounds of Tegra?

 

[Vortex6700]

Intel killed nvidia entire mobile strategy paying customers to use atom. And the killing was just a sideeffect. Collateral damage.
It will be an easy task to do the same for select segments and customers for a small part of the server market against an economic weak oponent as amd.

Amd needs something that is simply fastest. Sounds impossible but if they want to make a dent in this conservative and risk adverse monopoly market thats whats is demanded.

Intel has done much worse to products that were much better.

It is not profitable to make a better product than Intel.

 

[AtenRa]

Tco is not as simple as raw hardware cost and running cost vs perf.

I know very well what TCO is, but he was the one that said A1100 brings nothing to TCO vs the competition. And it seams that he considers the XEON-D as the competition.

So im asking, where is his TCO analysis on the A1100 ??