AMD demos 32-core Bulldozer server

[deputc26]

Sadly it is a race to the bottom between the US gov versus the PIIGS to see who can destroy their national currency the fastest.

Been saying this for years.

Anyway I haven't been on the forum much lately but I didn't know you did quant trading. I just wrote a couple of hf trading algorithms for a friend of mine and was surprised by how easy it was but wished his platform (tradestation) had lower latency. How's Metatrader4 in that regard?

 

[Idontcare]

Been saying this for years.

Anyway I haven't been on the forum much lately but I didn't know you did quant trading. I just wrote a couple of hf trading algorithms for a friend of mine and was surprised by how easy it was but wished his platform (tradestation) had lower latency. How's Metatrader4 in that regard?

MT4 sucks if you want to do anything with low latency. Tradestation is better.

Just to set your expectations, while the platform does allow you trade at the tick frequency, the minimum timeframe for charting purposes is 1min.

Even worse (and this is worse) is the platform is intentionally setup to be ok with "missing ticks". It's designed to be robust, doesn't fail when ticks go missing from bad internet connections or high latency situations, but the net of it is that your broker knows they don't need to send you all the ticks which means you have to program your algo's to be successful in an environment where it can safely be assumed the algo is missing market data that anyone else's algo's have access to.

Honestly I would not be on MT4 if it weren't for the fact that it is a relatively widely available platform (lots of broker's have it, and your code is directly transferable across brokers) and it works well for trading on longer timeframes (daily charts and the like) where your algo's are sensitive to missing ticks.

If you are already on tradestation then MT4 will just be a step back.

 

[Joseph F]

Quote frankly, I'm bored with the whole Bulldozer thing. The lack of information coming from AMD leads me to suspect that it's going to be a repeat of the Phenom launch.

Methinks this as well.

 

[Topweasel]

Sooooo... Upgrade my Win 7 Home Prem to Pro, get a 2S 256C, plunk in a few hundred GB of RAM... my wife would murder me, but it would be a great way to go! 🙂

As for the Euro I can't repeat what my friend said in confidence but let's just say that one of the Is in the PIIGS is in pretty rough shape and it might make Greece look like a Gyros & Ouzo picnic. I have little or no faith in USD either, so I think the best thing would be to start accumulating food, water, and ammo for the cabin in the woods. Can I trade chickens for a 2S 256C? 🙂

Pretty sure even PRO/Ultimate 64-bit there is an artificial limit on max memory.

hmmm. More then I thought. I thought it was around 92GB, but its 192GB. Where as different server versions can go up to 2TB.

 

[PlasmaBomb]

They reside in my underground lair...or "the basement" as my wife insists on calling it...so "noise" tolerance is a fair amount higher than what would be acceptable for an office environment.

How many cores can we license under a single Win7 x64 license?

Will these 4S 64-core beasties start to run afoul the existing software licensing models?

(my app, Metatrader4, is strictly windows-based for the purposes that I use it)

You are limited to one socket with Win 7 Home Premium and two sockets with Win 7 Pro or greater...

 

[Minerva]

I know this vid is fake but I want to see it run this fast lol!

http://www.youtube.com/watch?v=dZBgt9dRArk

 

[ElFenix]

Quote frankly, I'm bored with the whole Bulldozer thing. The lack of information coming from AMD leads me to suspect that it's going to be a repeat of the Phenom launch.

there wasn't any info from them about the 5000 series and the 6000 series graphics cards, maybe they've just learned to control leaks better.


damned if they do, damned if they don't. if they give information they're accused of paper launching or vaporware, and if they don't give information they're obviously hiding something (bad).

 

[Idontcare]

damned if they do, damned if they don't. if they give information they're accused of paper launching or vaporware, and if they don't give information they're obviously hiding something (bad).

It matters though, financially, as to what/who AMD is being damned by.

Paper launch and vaporware is bad for business on the accounting side of it. Not good to be damned like this.

Every person who gets an MBA these days is taught to be mindful of the so-called Osborne effect.

The Osborne effect occurs when this pre-announcement is made either unaware of the risks involved or when the timing is misjudged. Customers react immediately by canceling or deferring orders for the current product, knowing that it will soon be obsolete. Inventories increase and the company must react by either discounting or lowering production of the current product. Either of these choices depresses cash flow. In the actual case of Osborne Computer Corporation, the company took more than a year to make its next product available. It ran out of cash and went bankrupt in 1985.

Its a tough line to draw between letting out some information so as to bolster market confidence in your business model (mobo makers aren't going to invest in your future platform if you can't convince them it is a viable platform, for example) and the risk you take of encouraging customers to avoid today's products in lieu of purchasing the super-ultra-plus versions of them tomorrow.

Apple and Intel have mastered this balancing act. AMD has mostly elected to be conservative about it, probably because unlike Apple or Intel they simply cannot afford to bungle it and find themselves sitting on a pile of inventory they must mark down and sell for losses. (AMD does not have the cash reserves of Intel and Apple)

Withholding info and being damned as hiding something bad limits the damage from the damning to that of harmless words printed in forums by the fanboy crowds and possibly impacting the stock price in the near term as presumed failure on the future product is priced into today's stock price. (but this would be expected to be limited to the relatively minor percentage of traders who are prone to taking rumors as facts, and if the rumors are not backed by facts when BD launches then the stockprice in the long run will rise to be exactly where it should be)

Stock price does not affect AMD's ability to service its debt and pay the power bills, moving product does. So it stands to reason that given the choice of the two, letting the stock price take a hit from the speculation that BD will be dead on arrival is to be preferred to the possibility of having 3 months of 45nm product go unsold as everyone waits for BD.

 

[Terzo]

IDC, based on what you posted I have some speculation of my own.

This is assuming that Bulldozer will be shown at the July 15th HardOCP event. I doubt H would be advertising that unless they were 99.9% sure they'd have something to show, as their reputation would take a hit if Bulldozer didn't make an appearance.

Ok, continuing on, here's what I gathered from your post.
(A) enthusiasts are a non-factor in decisions for launch dates and releasing information
(B) for AMD, the consequences of releasing information on Bulldozer then missing launch dates are much worse (basically lost sales) than saying nothing at all (temporary fluctuations in stock prices)
(C) they do need to keep motherboard makers involved, but wouldn't a simple NDA solve that?

Anyways, if AMD is willing to show Bulldozer at the H event, they're on track for the updated August release...if they weren't, then showing Bulldozer would just cannibalize sales. Is this a decent conclusion? I can think of two alternative explanations.

First, maybe the showing will yield little useful information (i.e. play some games running on our Bulldozer processor).
Or second, maybe they're targeting the event towards enthusiasts, thinking businesses wont pay attention to the results. (I don't believe this, but I suppose it's possible).

Just having some fun with this, I'm sure there are a bunch of holes in my argument.

 

[veri745]

IDC, based on what you posted I have some speculation of my own.

This is assuming that Bulldozer will be shown at the July 15th HardOCP event. I doubt H would be advertising that unless they were 99.9% sure they'd have something to show, as their reputation would take a hit if Bulldozer didn't make an appearance.

Ok, continuing on, here's what I gathered from your post.
(A) enthusiasts are a non-factor in decisions for launch dates and releasing information
(B) for AMD, the consequences of releasing information on Bulldozer then missing launch dates are much worse (basically lost sales) than saying nothing at all (temporary fluctuations in stock prices)
(C) they do need to keep motherboard makers involved, but wouldn't a simple NDA solve that?

Anyways, if AMD is willing to show Bulldozer at the H event, they're on track for the updated August release...if they weren't, then showing Bulldozer would just cannibalize sales. Is this a decent conclusion? I can think of two alternative explanations.

First, maybe the showing will yield little useful information (i.e. play some games running on our Bulldozer processor).
Or second, maybe they're targeting the event towards enthusiasts, thinking businesses wont pay attention to the results. (I don't believe this, but I suppose it's possible).

Just having some fun with this, I'm sure there are a bunch of holes in my argument.

All the major OEMs and direct customers of bulldozer already have NDAs and parts in hand. They need to, otherwise there wouldn't be motherboards ready by the time the product is launched.

 

[Terzo]

All the major OEMs and direct customers of bulldozer already have NDAs and parts in hand. They need to, otherwise there wouldn't be motherboards ready by the time the product is launched.

So wouldn't the example given below be a non-issue? Or is there more to it that I don't know about?

Its a tough line to draw between letting out some information so as to bolster market confidence in your business model (mobo makers aren't going to invest in your future platform if you can't convince them it is a viable platform, for example) and the risk you take of encouraging customers to avoid today's products in lieu of purchasing the super-ultra-plus versions of them tomorrow.

 

[Idontcare]

IDC, based on what you posted I have some speculation of my own.

This is assuming that Bulldozer will be shown at the July 15th HardOCP event. I doubt H would be advertising that unless they were 99.9% sure they'd have something to show, as their reputation would take a hit if Bulldozer didn't make an appearance.

Ok, continuing on, here's what I gathered from your post.
(A) enthusiasts are a non-factor in decisions for launch dates and releasing information
(B) for AMD, the consequences of releasing information on Bulldozer then missing launch dates are much worse (basically lost sales) than saying nothing at all (temporary fluctuations in stock prices)
(C) they do need to keep motherboard makers involved, but wouldn't a simple NDA solve that?

Anyways, if AMD is willing to show Bulldozer at the H event, they're on track for the updated August release...if they weren't, then showing Bulldozer would just cannibalize sales. Is this a decent conclusion? I can think of two alternative explanations.

First, maybe the showing will yield little useful information (i.e. play some games running on our Bulldozer processor).
Or second, maybe they're targeting the event towards enthusiasts, thinking businesses wont pay attention to the results. (I don't believe this, but I suppose it's possible).

Just having some fun with this, I'm sure there are a bunch of holes in my argument.

To (A) I would say we have to accept that there are no absolutes, its a matter of balancing the pro's and con's of all the various cause and effects summed up. Enthusiasts are a factor, just not to the extent that we'd like to apportion ourselves as being IMO.

To (B) I would say yes.

To (C) I would say the distinction needs to be made between leaked info (coming from necessary ES seeding for platform development) versus officially sanctioned information released to the public domain.

Assuming BD makes a showing, and I see no reason why it would not, we can be reasonably assured it will be showcased in the absolute best light possible in a very controlled situation. As such I'm not sure what value will come of it for the everyday enthusiast.

It will boost confidence in BD, if this wasn't assured beforehand then AMD simply wouldn't organize the showcasing. And for the small percentage of the market that is already intentionally sitting on their hands waiting for BD to arrive before upgrading to either SB or BD this showcasing will go some distance towards justifying that hand sitting.

[note: I am in this situation, a member of the "sitting on my hands" club waiting for BD to either justify me buying SB or compelling me to buy BD]

To be sure the [H] event does involve some risk as it will result in some consumers choosing to postpone July/Aug CPU purchases in lieu of the BD launch that is in the offing.

On the other hand the obvious intended upside is the prospect of garnering more visibility for Bulldozer amongst the consumers that currently associate AMD with second-best in CPU's. When the competition sells 5 cpus for your every 1 cpu, that's a lot of sales going to the competition that could be converted into the "sitting on their hands" club in a way that undermines the competition and not yourself.

Its a double-edged sword, but the edges are not equally sharp. One is slightly more dull compared to the other. (in a 5:1 ratio) So the key is to do your best to only cut yourself with the duller blade while wielding the thing.

 

[bridito]

Pretty sure even PRO/Ultimate 64-bit there is an artificial limit on max memory.

hmmm. More then I thought. I thought it was around 92GB, but its 192GB. Where as different server versions can go up to 2TB.

I'm maxing out my 12GB on my i7 940 so I'm sure that 192GB will keep me happy for a long while!

You are limited to one socket with Win 7 Home Premium and two sockets with Win 7 Pro or greater...

Upgrade to Win7 Pro, buy 2S 192GB system. Check! In my dreams at least. If I spent that kind of money my wife would make sure that my corpse could never be found! 🙂

[note: I am in this situation, a member of the "sitting on my hands" club waiting for BD to either justify me buying SB or compelling me to buy BD]

I knowest thereof thou speakest. I was only looking at SB-E and really wasn't paying any attention to BD until I caught wind of wild expectations issuing from some quarters of the setting of a new computing paradigm with the release of BD. The more information or alleged information that comes out the more I'm leaning towards the conclusion that BD is going to be no competition at all to SB-E, but I'm waiting as well. Should the indy benchies prove that BD really is a great outperformer, I'm going to buy it shortly after launch. Otherwise, I'll just hold off for a couple of months and get an SB-E Quad system as soon as it's out.

 

[mnewsham]

Quote frankly, I'm bored with the whole Bulldozer thing. The lack of information coming from AMD leads me to suspect that it's going to be a repeat of the Phenom launch.

I haven't seen "frankly" post in this thread yet so im not sure how we could quote him...
:sneaky:


Really though, 64 cores in 1U is pretty sweet 🙂

 

[tweakboy]

Intel will lose IMO to BD has 16 core or 64 core .

Unless Intel gives us something new, like a 4 socket mobo for workstation put 4 CPU's in there with HT 64 core... Amazing

Like right now Im playing FEAR 3 everything maxed 4xAA 16xAF , @ 1080p with vsync on of course and tb.

When are these dual socket or 4 socket come out for Intel. We see what AMD is doing. It really want to win this race with more cores instead of actual CPU power. thx

 

[Soundmanred]

Intel will lose IMO to BD has 16 core or 64 core .

Based on what?

Unless Intel gives us something new, like a 4 socket mobo for workstation put 4 CPU's in there with HT 64 core... Amazing

Seriously?

Like right now Im playing FEAR 3 everything maxed 4xAA 16xAF , @ 1080p with vsync on of course and tb.

What does this have to do with anything? Most people here can do that, and you don't play games on blades/servers.

When are these dual socket or 4 socket come out for Intel. We see what AMD is doing. It really want to win this race with more cores instead of actual CPU power. thx

No, we hear what AMD is doing. I honestly can't read your last sentence so I can't even comment on that.

Ridiculous.

 

[Idontcare]

Ridiculous.

You just got tweaked 😛 :thumbsup:

 

[Terzo]

On the other hand the obvious intended upside is the prospect of garnering more visibility for Bulldozer amongst the consumers that currently associate AMD with second-best in CPU's. When the competition sells 5 cpus for your every 1 cpu, that's a lot of sales going to the competition that could be converted into the "sitting on their hands" club in a way that undermines the competition and not yourself.

Its a double-edged sword, but the edges are not equally sharp. One is slightly more dull compared to the other. (in a 5:1 ratio) So the key is to do your best to only cut yourself with the duller blade while wielding the thing.

What I'm gathering is not to read too much into AMD demoing Bulldozer (suggesting an "on time" release), since they could be trading current sales in the hopes that some consumers will see Bulldozer, then decide to wait for it rather than purchase Intel now.

Is that correct?

 

[Phynaz]

When are these dual socket or 4 socket come out for Intel.

Wow.

Just Wow.

 

[notty22]

What I'm gathering is not to read too much into AMD demoing Bulldozer (suggesting an "on time" release), since they could be trading current sales in the hopes that some consumers will see Bulldozer, then decide to wait for it rather than purchase Intel now.

Is that correct?

To give AMD investors some confidence also imo. Desktop is late *, with m/b's only available. Sales/ inventory #'s are eventually going to reflect the market waiting to purchase AMD's new 32nm offerings, forecast.

*Intel gets chip market share at AMD's expense

 

[Idontcare]

What I'm gathering is not to read too much into AMD demoing Bulldozer (suggesting an "on time" release), since they could be trading current sales in the hopes that some consumers will see Bulldozer, then decide to wait for it rather than purchase Intel now.

Is that correct?

Yep. That is a correct interpretation of my stated opinion, I don't know if my opinion is correct though 😛

One thing we should see from AMD if they are going to show BD at the [H] event is that they should show us the top-model. So that will be nice to get a better idea of just where the top-end of the spectrum is going to fall.

I'm expecting 2600K territory, definitely north of 2500K territory. Would be nice if it came out a little higher than 2600K and compelled Intel to release a 2700K.

 

[GammaLaser]

Yep. That is a correct interpretation of my stated opinion, I don't know if my opinion is correct though 😛

One thing we should see from AMD if they are going to show BD at the [H] event is that they should show us the top-model. So that will be nice to get a better idea of just where the top-end of the spectrum is going to fall.

I'm expecting 2600K territory, definitely north of 2500K territory. Would be nice if it came out a little higher than 2600K and compelled Intel to release a 2700K.

Any AMD release that results in Intel adjusting prices and/or releasing better SKUs is win-win for all (consumers, that is!) :thumbsup:

 

[Soundmanred]

You just got tweaked 😛 :thumbsup:

Someone has to point out how ridiculous he is once in a while, maybe someday he'll realize it.
Not holding my breath.