Printable View
Quote:
Originally Posted by Otis11
I am not sure about the specific model numbers (for G34) but historically
there were 3 different Opterons for each family:2000,8000 and 1000
The 2000 runs on 2P and the 8000 would run on 2P,4P and 8P but would
cost much,much more (then 2000).
The 1000 is another story all together because of the "consumer" socket/pin
count.
Does AMD "allow" companies (like Asus,DFI,MSI,Gbyte etc) to make a 1P (or 2P) G34 consumer board?
All bets are off as we move to 4000 and 6000.Quote:
Originally Posted by SocketMan
I will allow anyone to build anything that sells more Opteron processors.
However, these processors are focused on the enterprise market, not consumer. You will not see support for consumer features and things like overclocking.
I generally recommend Phenom parts for consumers because you will get a better experience.
I know and understand the thinking here but for what it's worth there is a pretty sizable market out there for us dual socket fans that also like to get the most out of our machines.Quote:
Originally Posted by JF-AMD
Just look at the interest there is in the new W555 dual socket 1366 board by EVGA and you'll understand.
Personally from what I've seen these Magny-Cours 12 core chips look to be very impressive and would love to see a board out there that could max them out.
Opteron 6xxx series pricing -- source
Type Clock L2 cache L3 cache Cons. (ACP - TDP) List price
Opteron 6128 1,5 GHz 8 x 512 kB 2 x 6 MB 75 - 115 W €254
Opteron 6134 1,7 GHz 8 x 512 kB 2 x 6 MB 75 - 115 W €489
Opteron 6136 2,4 GHz 8 x 512 kB 2 x 6 MB 75 - 115 W €692
Opteron 6168 1,9 GHz 12 x 512 kB 2 x 6 MB 75 - 115 W €692
Opteron 6172 2,1 GHz 12 x 512 kB 2 x 6 MB 75 - 115 W €917
Opteron 6174 2,2 GHz 12 x 512 kB 2 x 6 MB 75 - 115 W €1078
8 cores for €254 :shocked:
Even the highest-end 12-core seems more than reasonable at €1078
Well, that tells us a lot about it's relative performance to comparable Intel products. AMD would have surely asked for more if they held the upper hand.Quote:
Originally Posted by Frank M
So Opteron 6174 will more or less match a 6 core 3,33GHz Gulftown.
Or they just want to ruin Intel sales and gain far more market share...?Quote:
Originally Posted by savantu
I beg to differ on that one. Our pricing stacks are put together based on the market and customers, not the compeition.Quote:
Originally Posted by savantu
At any performance level you will see a much better price on AMD. Plain and simple.
In the intended workloads(throughput) for the 12 core Opteron,the 6 core Westmere at 3.3Ghz is going to get beat... You can't fight 12 10h cores @ 2.2Ghz with even a 3.3Ghz 6 core Nehalem with a bit more of cache.Plain and simple.Quote:
Originally Posted by savantu
oh rly? :rolleyes:Quote:
Originally Posted by JF-AMD
you better price X6 lower than U$280 then
I don't do client processors, I do server processors.
Quote:
Originally Posted by Behemot
One of the major cash cows for AMD was its server CPUs, more importantly the 4P segment where it held the performance lead over in 2004-2009. Selling the same CPUs you had on desktops for $99 for $1800 surely had a positive impact on margins.
Now they eliminate the segmentation. Good or bad ? Depends on who will have the performance crown and more features.
I believe it's mostly targeted at 2S price levels, because that's where their performance will be. Beckton has enough oomph and scalability features to ask for a premium at 4 and 8S.
Same number of threads, 50% more frequency, 50% more cache. Are you sure ? I think they will be really close.Quote:
Originally Posted by informal
There was a Prime benchmarks shown around here, 2P MC with 24 cores and 3GHz was 15% faster or something like that than a 2P Gulftown with 12 cores and 3.6Ghz. And we're talking best case scenario SW wise to load a CPU.
I hope you aren't saying that with a straight face; I still remember people paying $400 for A64 3800+ and $2200 for an 8xxx Opteron back in 2005.Quote:
Originally Posted by JF-AMD
Your pricing is where you can sell enough stuff to make some money and keep FAB utilization to an acceptable level, so fixed costs won't ruin you. And that level is mostly set by where the competition is performance wise.
If possible, you would gladly start pricing Magny Cours from $1800. Unfortunately, it overlaps quite a bit in performance with 2S Nehalems/Gulftowns, that is the $250-$1300 range, which is exactly where your price is btw.
Except Nehalem typically gets 50% or more throughput per core so Westmere with a 3:2 clockspeed advantage is going to be capable of matching or exceeding Magny Cours in throughput, except for things like Linpack.Quote:
Originally Posted by informal
AMD is not targeting a small wprime application with its 6100 series... In the intended workloads these chips will be great and the price will be even better. What's evident is that AMD is executing perfectly since Shanghai and this is an indication of what we will see with BD launch.
Make that a 60%,or wait 90%... Dream on man.Quote:
Originally Posted by accord99
MC is a DDR3 based platform and there will be more to come in terms of power efficiency and clocks . If AMD is doing this with its current 45nm cores what will happen when they launch 16 core Bulldozers next year,in the same infrastructure ?
So, current 4 core Nehalems don't already match or exceed the throughput of 6 core Opterons?Quote:
Originally Posted by informal
Underclocking and Undervolting isn't that hard of a game to play.Quote:
MC is a DDR3 based platform and there will be more to come in terms of power efficiency and clocks . If AMD is doing this with its current 45nm cores what will happen when they launch 16 core Bulldozers next year,in the same infrastructure ?
Quote:
Originally Posted by JF-AMD
What are you thoughts on a higher performance core versus core count? I'm not sure what part of amd you work in, but I'm just curious if the market these guys are targeted at truly have applications that can take advantage of the core count to make up for a slower cores.
Current Opterons are DDR2 only and work at lower frequency. Current single Opteron core is 15.6mm2 while current single Nehalem core is 24.5mm2,so it would be embarrassing for it to not perform (per core) better than 10h.Quote:
Originally Posted by accord99
Underclocking and Undervolting is not what I'm talking about.It's AMD's CTI.
Yes quad Nehalems do but 6 core AMD's are priced similarly and you will notice that in a whole platform the AMD system has a edge price wiseQuote:
Originally Posted by accord99
Underclocking and Undervolting is a very hard game to play in server world. Going away from specified perimeters voids the warranty on any server chip and worst than that insurance will not cover the chip anymore and you dont want to do that.
I know about examinations that are carried out by TATA insurance company "India" once a month to insure that the server is treated rite.
let magny IPC = 1, magny frequency = 1, let's say a westmere core with hyperthreading = 1.3 magnycour cores (?), and compare 3.33ghz westmere to 2.2ghz magny
magny performance = 12 (cores) * 1.0 (relative IPC) * 1.0 (relative frequency) = 12
westm performance = _6 (cores) * 1.3 (relative IPC) * 1.5 (relative frequency) = 11.7
?
hmm it seems as if that would be correct.
http://www.xbitlabs.com/news/cpu/dis...urs_Chips.html
That's about right, yeah, the IPC different isn't that big, though.Quote:
Originally Posted by hollo
Remember that Westmere pricing will probably be higher for 2-4 socket platforms.
Lord those clocks SUCK.
Many cores vs. better clocks. In the end, they will probably be on par in performance. And if not, I would suspect Intel's 32nm 6-cores have some headroom in clock frequency even at the server TDPs (my own experiences show this to be true).
AMD will offer 6 core for 35w
http://www.xbitlabs.com/news/cpu/dis..._Platform.html
Now since Thuban may share similar internals as this we can expect lower watts thuban also? That would be sweet when single core is used ramp it up otherwise when all cores are used return to stock.
Server customers look at total CPU throughput, so more cores is a better thing. We went from single to dual, dual to quad, quad to hex and nobody ever looked back.Quote:
Originally Posted by Wiggy McShades
Client, on the other hand, may be more predisposed to fewer cores at higher clocks. Many client aplications are not as threaded, so clocks are more important.
not sure what to make of this wording?Quote:
Originally Posted by ajaidev
Quote:
Originally Posted by xbitlabs
that 4160EE is on 45nm right? if so thats very impressive. would beat a 3ghz tri-core using laptop like power consumption, if every program scaled perfectly of coarse.
Quote:
Originally Posted by informal
:rofl::up:
I am waiting for people to start complaining about super pie numbers.....
Got to love some of our armchair "computer experts" ;)
This is actually, to a first approximation, a nice analysis ... it's hard to tell at this point so many variables, but it looks like to me MC will win some, WSTM will win some, and others will be a tie.Quote:
Originally Posted by hollo
To get a rough (very rough) idea, just take a 4S Istanbul SPEC score, I think WSTM will have an edge in FP, and MC may take a lead in INT.
Westmere will have an edge in FP? I will take any bet from anyone on that.
3 memory channels vs. 4 channels, westmere will not have an edge in FP. Remember also that there will be 12 FPUs in a Magny cours and only 6 in a Westmere.
DDR3 vs. DDR2?Quote:
Originally Posted by JF-AMD
4/5 instructions per cycle vs 3?
3.3GHz vs. 2.2GHz?
Mangy Cours might match Westmere in FP, but I would be surprised if it surpasses it. i7 currently spanks Opteron 2400/8400 series in FP workloads. I see AMD just fusing two Istanbul cores together as a stop-gap measure to compete at the high-end until Buldozer cores are available. And even then, Intel has Beckton to counter with (8 cores, 16 threads), due out in 1-2 months.
Oh, and for the record, I run a server farm with mostly Istanbul CPUs, so please don't brand me as an Intel "fanboi".
I think you will be impressed with the FP capabilities of MC. And I am not that concerned with Beckton. Intel is targeting that squarely to be "Itanium III." From what I have seen from a price and power perspective, they are going to be hard pressed to convince people that it is a wise choice for their data centers. The memory buffer design tells me that they didn't learn that much after all of their customers complained about FB DIMMs. And Beckton will only support 1066 memory from what I have read.
How much memory can you link to a socket for MC ?Quote:
Originally Posted by JF-AMD
For Beckton you can link up to 32 memory DIMMs without loss of speed.
DDR3 vs DDR3Quote:
Originally Posted by mstp2009
http://www.xtremesystems.org/forums/...24#post3995524
I think AMD slides also showed Socket G34 is DDR3.
I thing MC's can use around 64GB's of ram in a 8P setup '48core" i could be wrongQuote:
Originally Posted by savantu
You can use 3 dimms / MC so you end up with 12 dimms/socket. And if you compare the MC of Nehal against Magny you will see that the Magny is able to sustain a bit faster memory speeds when using 2 dimms/MC and 3 dimms/MC Nehalem will drop to 1066 with 2 and 3 dimms/MC in both SR/DR while Magny will only drop to 1066 when using 3 dimms DRQuote:
Originally Posted by savantu
again Beckton is a different price range.
Magny is only up to 4P so that is 48 cores and also 48 dimms :D so it should give a total of 384GB ram with DR 8GB dimms but those are very expensive......Quote:
Originally Posted by ajaidev
funny as usual a list from a certain source is going around as the final. Well I can tell you some speed allocations are wrong and are actually based on an old docQuote:
Originally Posted by Frank M
westmere x5680 is a 130W tdp part so you will have to compare against the 6176SE part so it's 2.3 ghz :cool:Quote:
Originally Posted by mstp2009
Magny match Westmere in FP :confused: is not fusing just 2 istanbul it will more then tripple the memory bandwith of an istanbul
Semi accurate says these are the prices:
8-core models
* 6124 HE, 1.8GHz, 65W TDP $529
* 6128 HE, 2.0GHz, 65W TDP $599
* 6128, 2.0GHz, 80W TPD, $309
* 6134, 2.3GHz, 80W TDP $599
* 6136 2.4GHz, 80W TDP, $849
12-core models
* 6164 HE, 1.7GHz, 67W TDP, $879
* 6168, 1.9GHz, 80W TDP, $849
* 6172, 2.1GHz, 80W TDP, $1,149
* 6174, 2.2GHz, 80W TDP, $1,349
* 6176 SE, 2.3GHz, 105W TDP, $1,599
http://www.semiaccurate.com/2010/02/...evealed-early/
if these are indeed true i am very very interested in a 6128 2P machine :)
Ya i meant 4P thats why i said 48 core :p:Quote:
Originally Posted by duploxxx
and i also meant quad 64 meaning 4*64= 256GB i do remmber reading somewhere that there will be 32 DIMMs so 256Gb will be the max that it supported
I do a lot of FP crunching, and most FP workloads are not memory bandwidth constrained. It's just not an issue for over 99% of workloads, and is a non-factor for us crunchers. It's just raw SSE computational power that matters.Quote:
Originally Posted by duploxxx
Do these CPUs use socket F / 1207 ?
They use a new socket, G34. It's much wider than the older one.Quote:
Originally Posted by Mech0z
So there is no motherboards out for it yet?Quote:
Originally Posted by N19h7m4r3
Not that I know of. They have been shown, but I haven't seen them sold yet.
In fp there will be no problems for MC Vs Westmere. Both 10h and i7 have same flops theoretically while i7 had an advantage of SMT (gives 0-30%) and DDR3. MC will answer this with much more cores ,2x more,at lower clocks but enough to exceed high clocked Westmere in throughput and with DDR3 support + qudchannel it will address the Westmere's 3-channel DDR3 controller. Simple yet very efficient.
edit: oh and the latest pricing of the MC posted at SA website is wrong too..
12 DIMMs per socket.Quote:
Originally Posted by savantu
I was unaware that there were 32 DIMM per socket configs for Beckton.
The most I have seen are 12 DIMMs per socket and max speed of 1066.
Do you have a link?
Quote:
Originally Posted by ajaidev
That information is not semi accurate, the accuracy rate is lower ;)
Anyone know if those are actually TDPs, as opposed to AMD "ACP" numbers?Quote:
Originally Posted by ajaidev
105W appears to be the SE "ACP" wattage for 6-core parts, which is why I ask.
The X5680 is a workstation only part, typically priced well above reasonable rates and consuming tons of power. We won't see it in the server world.Quote:
Originally Posted by duploxxx
29% of the 2P market is priced at $3K or below. How do you build a <$3K 2P server with $3200 worth of silicon.
62% of the 2P market is priced from $3-6K. Basically more than 90% of the market price points could never be achieved if you are using those procs.
The one quad Beckton board I've seen is 8 dimms per socket and correct on the 1066 memory.Quote:
Originally Posted by JF-AMD
I'd disagree with this, power consumption less than previous generation 45nm Gainestown chips.Quote:
Originally Posted by JF-AMD
$3200 for chips
$500 for board
$300 for memory
$4000.00 sub total
That leaves $2000.00 for video,hard drives,case,PSU..Tight yes but doable.;)
hehe don't worry, I wasn't going to compare at all high-end, it was just a remark to the " virtual performance calculation", there are however a few OEM that actually sell those X range in certain rack servers.Quote:
Originally Posted by JF-AMD
the next range of AMD G34 systems that we will ship this year will be based on the 6134 or 6136 will depend a bit on the final asp
FBDIMM has one mem buffer per slot while Nehalem-ex will have one mem buffer per channel so the power consumption of mem subsystem will be much lower.Quote:
Originally Posted by JF-AMD
In fact Intel claims up to 16 dimms per socket.Quote:
The most I have seen are 12 DIMMs per socket and max speed of 1066.
http://download.intel.com/pressroom/pdf/nehalem-ex.pdf
depends how much that buffer will consume (3xsmall against 1 big) and even more base cost.... yes your memory will be cheaper and less consuming but the base board will be way more expensive and i don't think you will be able to plugin those buffers each time you buy a cpu :)Quote:
Originally Posted by kl0012
It is 16 like another poster pointed out; apparently my math skills failed me, sorry.Quote:
Originally Posted by JF-AMD
2 IMCs x 2 Buffers/IMC x 2 Channels per buffer x 2 Dimms / channel = 16 Dimms / socket.
I believe that is ACP.Quote:
Originally Posted by terrace215
That means the SE part is 140w TDP while max power even higher.
In the case I am looking at (~960x 12 cores for a renderfarm), these cores will be maxed out 24/7. In 3D animation, raw horsepower is still king, and people will always want more power to keep up with demands for yet higher levels of "Wow!" that are being placed on even small production houses that simply cannot purchase this kind of firepower. Outsourcing is increasingly popular.Quote:
Originally Posted by Wiggy McShades
For the people comparing these to various desktop parts, That is the target audience that AMD is looking at. Although a few guys here will no doubt pick these up (XS Crunchers mainly, I would love too assuming I had the disposable to do so), do you really think that the whole of XS's purchases of these processors will equal even that single sale? And this is XS, the leading edge of the Enthusiast market.
These chips are not comparable to the desktop parts, and were never designed to be.
I can tell you honestly that MANY people on this forum would be shocked if they leanred the actual numbers of dual socket systems that the members here own.Quote:
Originally Posted by InSanCen
I have 4 duals and 2 singles at my house and I know a lot of guys here that have many more than I do..That's God's truth.:up:
AMD's TDP = max powerQuote:
Originally Posted by savantu
105W ACP = 137W TDP. It is the max TDP for the Opteron lineup.
i am assuming that you are talking about F@H. its not floating point that requires BW, its what you are doing. physics solvers cant fit in cache so more BW is important.Quote:
Originally Posted by mstp2009
theoretical performance means nothing. sse is not implemented the same way in both architectures(see prescott v. conroe).Quote:
Originally Posted by informal
Some older pics::shocked:
'raises hand' Dell Precision 650 SMP. Old? Yes! Dual cpu system? Yes!Quote:
Originally Posted by Movieman
yep I have a dual CPU system too. Dual Shanghai. Not the very latest, but still pretty beefy. I am probably going to try and get a dual socket system with these CPUs next year with my Christmas bonus. I just hope by then someone will have worked out a way to overclock the CPUs.
What about 1P but with a bit of ACC;);):shocked:Quote:
Originally Posted by EniGmA1987
I keep thinking (for some reason) mini ITX,DFI and socket G34........
Will not happen, there will be no access to ACC on these, they are targeted at server apps, not overclockers.Quote:
Originally Posted by SocketMan
I did have one of these in my hands today, had a bunch of customer meetings in our briefing center:
http://www.xtremesystems.org/forums/...1&d=1267146686
And we packaged up Magny Cours processors for all of the meeting attendees.
It's been done already. s7eph3n over in the AMD forum posted last year with some 3ghz+ MC goodness.Quote:
Originally Posted by EniGmA1987
His (ES) chip,has an unlocked multiplier,won't be the case for the "regular"Quote:
Originally Posted by Hawkeye4077
cpus.
Quote:
Originally Posted by JF-AMD
I think it's safe to say that we've established the bold part already :yepp: :p:
However the actual performance (of any hardware) would not change
with different "target markets".
A 2.2ghz cpu will perform exactly the same for "Joe Blow
Over clocker" and the "Big Shot Enterprise".
Also the money from the 2 is identical - if AMD sells 2 cpus to Joe at 1000$
(each) they get 2000$.When AMD sells 2 cpus at 1000$ to the Enteprise
they still get 2000$,but there are more Joes then enteprises.
A pic below shows the number of people participating in WCG,our (XS) team
(1 out of many) has over 1000 active members.Globally there are hundreds
of thousands of PCs and the number is growing on the daily basis.
Almost everyone out of that 1000+ overclocks their machines,stability
is not an issue,it's a given - because no one wants to lose a Work Unit.
We have a new "market segment" that basically developed over the last ~5 years.This segment is way above the "Ultra Enthusiast" "needs"
(4-6 cpu cores according to some charts),but obviously below the
"ënteprise".
It's money (a lot of it,we're talking tens of thousands of people) laying
on the road,all AMD has to do is to pick it up.
The cpus don't have to be targeted they just have to perform. :)
Also the term "Overclock/er" has become to general,there are at least 2 groups in it:
The first one is the "record breakers" -these guys use the LN2,Dry Ice and run benchmarks faster and faster under very,very high gigahertz (for a few minutes :p:)
That's were all the "fame" is.:rolleyes::p:
Stability here is only needed to complete the benchmark run and take a
screen shot.I do not think this group is even interested in M.cores.
The second group runs their overcloked PCs 24/7/365 at full load (without crashing btw),but of course the clocks are much lower then the "records".
This (I think) is one of the best descriptions as to "why" we need the over clocking ability.Let the motherboard manufacturers do what they do best -
give us motherboards with the features we need.
AMD has done their job already - the cpus are made.:up:
The philosophy has always been (something like):
"give users the ability to make a choice", and that is all we really need.No need to target us :p:
Quote:
Originally Posted by Movieman
I have to agree - think about good old 939 Opterons. There is huge market...and AMD should make money wherever possible.
I agree overclocking is a non-issue at 4p and above.
But 2p and especially 1p motherboards would bring MORE customers to that segment if some overclocking features are added. You would certainly not lose any customers due to overclocking functions added though.
This would not be a bad move on your part JF, I'm both a Master in Marketing student and a PC enthusiast and I am just one of the thousands and thousands of people who would be interested in a 2p MC system running at 2.8ghz. Even if it costs me 800$ more than a Phenom II platform.
Extreme overclockers do also enjoy the realm of server hardware. I remember back when the 1 & 2 series Opterons came out and they also targeted at these users, which AMD referred to as the "Workstation" segment. It was a HUGE success, Opteron 1 & 2 series interested many enthusiasts and I can testify to this as I have myself assembled 4 systems at that time for people who chose that platform over Athlons and spent a good 50-100% more just to have Server hardware with overclocking capabilities.
Please you must bring that back..
Do not simply discard our opinion about overclocking as a consumer hardware niche segment thing. This website represents a sample of the population of PC users that AMD marketing dept really needs to pay attention to. Not only because everyone posting on this site has a general good to very good understanding of hardware, but also because most are early adapters. The very segment which will test and approve first hand any new technology which you bring to the market. This would be one of the best places you could collect information about what the informed customer wants, and as a marketing guy, I have a very good hunch that this is exactly the reason why you post in XS Forum JF. If I were in your position, I would do exactly the same.
Let me try to put this in the right frame of reference.
We will make parts. They can be used anywhere you want. If you want to overclock, have at it, we won't stop you. If you want to use them for your desktop systems, fine with that. I use an Opteron 1389 in a cleint board (Gigabyte 780) and it works great.
The question is not "will it work" and more of "will you be marketing it that way."
The former is true, the latter is not.
It is a matter of focus and resources. Our main market is the server market. Our marketing efforts (positioning, messaging, benchmarking, advertising, promotion, customer engagement, etc.) will all be around the server market.
At the point where you start marketing "this is a great desktop", then your core customers start thinking "maybe that isn't what I want to run my enterprise database." Unfortunately part of marketing is targeting, and when you target, you are self-selecting the market down from everyone to a smaller group that is the target market. You can't be everything to everyone. There was an old Saturday Night Live skit about Shimmer "it's a floor wax AND a dessert topping."
When people try to be too much, you end up doing neither great. Camcorders also take still pictures, but they do a crappy job relative to a camera. And cameras can take video, but it is generally not great.
The markets that we are going at require us to focus and message at particular targets. We won't necessarily be a great selection for file/print sharing because you don't need 8 or 12 cores for that, but I can guarantee you that someone will buy us for that (based on the value that we will have). That to me, is a secondary market.
I am not saying that there will not be people using our systems for those tasks, but just that you won't see us making overt changes to the silicon to target those customers.
The real issue that I see is that we control the silicon, but it is the infrastructure vendors that will determine whether the product meets your needs, because the board choices will dictate whether you buy my processors. I don't make boards. I don't have influence on our partners (generally speaking.)
When we make silicon choices they will be focused on database, HPC, technical computing, virtualization, cloud/web, business apps (SAP, ERP, etc.) If those decisions also make my processor great for you, then, by all means buy it, because I think it is going to be a real winner.
thank you for your reply
I understand your point of view and for all its worth I agree with pretty much everything you said.
My previous argumentation was based on the possibility that you have some form of control over what features the new platform will support. From all the previous arguments in this thread, I concluded that motherboard manufacturers would not be able to add any desktop features on their server boards because AMD would not endorse it implying some form of control.
From your last post this seems not to be the case
Does this mean that regardless of what you will market and target MC as, that platform manufacturers will still have the freedom to produce a "hybrid" server/client platform i.e. g34 with overclocking options?Quote:
We will make parts. They can be used anywhere you want. If you want to overclock, have at it, we won't stop you. If you want to use them for your desktop systems, fine with that. I use an Opteron 1389 in a cleint board (Gigabyte 780) and it works great.
But otherwise, it is understandable you do not want to blend some particular target segments, and that's fair.
it is up to asus, gigabyte, supermicro, tyan, etc. to provide a G34 board with the features we like(overclocking).
if this sells more opterons, amd will NOT stop it, correct?
can amd give even just ONE board manufacturer a nudge/hint?
If someone wanted to create a board like that I would be fine with it. I want to fill sockets. I would not limit my partners from doing anything that they believe will drive more sales for me.Quote:
Originally Posted by Dimitriman
To that end we have seen servers with low end desktop parts in them from our partners. I'm not happy about that, but as long as AMD is getting the socket I support it.
I wonder who at AMD is trying to peddle false rumors to the trade press?
:D
http://www.techeye.net/chips/intels-...yed-says-rival
Here at TechEye, we hear odd rumours all the time. It’s par for the course in the dog-eat-dog world of billion dollar semis.
But once in a while we stumble across a rumour that is clearly the product of- if we’re being magnanimous - wishful thinking, or even more likely purposeful misinformation.
Of course, the firm involved chose to spread the info as “off the record” which leads to a rather interesting moral dilemma. Namely, if someone says something to a hack off-the-record and said hack discovers it's completely inaccurate, and maybe even purposeful misinformation, can one publish it anyway?
Our esteemed editor says yes. Drop AMD in it.
[...]
what was the point of that article and in what correlation is it with this thread??Quote:
Originally Posted by terrace215
:shrug:
As a former member of the failed QuadFX platform, I would say it was ASUS's fault the platform failed. BUT AMD could have helped keep it going. The CPUs were already there, just renamed to FX series. Asus launched a board that had TONS of upswing if it had been maintained. Only 3 processors released for it.. 3.. And we paid what.. 299 apiece for them (FX70s here). When the quad cores came out, asus just let it fall apart. There was an AMD based chipset designed and built (at least one anyways) It never ever saw the light of day.
After much complaining by its customers Asus finally gave us a Quad Core capable motherboard in the form of the WS /B. It will run barcelonas, but not Shanghai's. Tyan gave us the -E series motherboards that run Quad and Hex cores now. Once again, Asus had the opportunity to latch on to some more market share. Let's call that board the KFN32-D SL /SH. Shanghai capable. ASUS even has a section for it on their website with a bios release.. but guess what we don't have.. a motherboard.
Even MSI has a couple boards out there that run Quads. The Speedster series. But MSI has given very limited support on them. The only reason the Speedster2 is in the US is because a member of 2CPU went to MSI directly and got boards imported over. But it is already EOL'd the day after it came out.
If AMD wanted us to have an enthusiast level 2S platform, we would have one already, made by somebody...
Intel made their own ie Skulltrail.
I have owned every AMD multicpu capable platform since the 760MPX days. I fear that Socket F will be my last at least through the G34 series. I am just not seeing any motherboards out there that would fit my requirement of multiple video cards. And SLI is dead on AMD workstation boards. That does not help either.
Bottom line: Until AMD asks it's board makers to come up with something for us, they wont. The new AMD SR chipsets are a goldmine. CrossfireX boards in an ATX size are easily doable.
You might have to read the whole article, not just the excerpt. :shakes:Quote:
Originally Posted by Nedjo
Unless I'm mistaken, this thread has many posts about Beckton vs Mangy. :yepp:
And the correlation might be even stronger than that. :D
Thank you for this post. Well said and yes, I fall into this group..Quote:
Originally Posted by SocketMan
I'd love to see a MC dual socket board with OC options in the bios like the new EVGA W555 board has for the Intel Westmere's.
Is it a risk for a company to commit the resources to try this? Definately.
Will it pay off? I don't know as I'm not aware of the numbers needed to turn a profit on a board like this but in my heart I say Congrats to EVGA for having the balls to try.
I think realistically only Asus or possibly Gigabyte could make that kind of board for the MC platform. The key is top parts and forget the cost.
The market for a board of that type isn't money limited.
between Hexa-Core Gulftown and mission critical Tulsa (I hope I've hit codename), Beckton with it's special requirements for FB-DIMM 2, and complicated/expensive mobos is dead-end project of long ago departed Pat Gelsinger... only thing that's keeping him alive is the fact that Intel doesn't want to make another Larrabee announcement in such a short time-span!Quote:
Originally Posted by terrace215
http://www.59hardware.net/~benoit/Ap...Beckton_02.jpg
Investing so much resources in a complete spin-off product like Beckton for disproportionately big 4-socket market is nonsense, that might have worked before Economy breakdown, but not anymore!
but who would no... if Intel is so desperate for 4S product... they just might do something so unreasonable as Beckton launch!
point is what ever they do, I don't see a reason for AMD to talk about it, or spread rumors 'cos MC and Maranello as a unified long term platform that can enable 1/2/4S is far more appealing, and affordable for server builders
Beckton is DDR3-1066.. I have a pdf here showing some of the upcoming boards but would get my backside spanked if I showed it..:rofl:Quote:
Originally Posted by Nedjo
Well, AMD TDP is actually: TDP. Thermal Design Power. The thermal design power is the maximum power a processor canQuote:
Originally Posted by Oliverda
draw for a thermally significant period while running commercially useful software.
3. The processor thermal solution should be designed to accommodate thermal design power (TDP) at
Tcase,max. TDP is measured under the conditions of all cores operating at CPU COF, Tcase
Max, and VDD at the voltage requested by the processor. TDP includes all power dissipated on-die
from VDD, VDDNB, VDDIO, VLDT, VTT and VDDA. TDP is not the maximum power of the
processor.
http://support.amd.com/us/Processor_TechDocs/43374.pdf
Pure max power would be even higher.
And ACP... is just something marketing made up because they thought the TDPs didn't look so good. :shakes:
But by listing a number as just "wattage" or "power" in the usual places, they are getting sites to fall for it, and assume it is TDP. Bravo?
And you can bet that when beckton benchmarks are out, AMD *won't* talk about it. At all. :rofl:Quote:
Originally Posted by Nedjo
the asus board was the only 680a board ( really two 560a together) anyone with those got left hug out to dry.Quote:
Originally Posted by Hawkeye4077
the second was fasn8te board which was never made.
EVGA said they're waiting for new architecture from AMD, since nv780a was late to market and they don't have experience with am3 now.
someone needs to tell them socket G34 will support new architecture from AMD. (bulldozer)
In the work that I do my westmere is out putting a dual socket beckton system and from the benches I saw the MC at 2992 in S7's screen shot is so damned close to my westmere at 3458 that it's '6 to 5 and pick 'em"Quote:
Originally Posted by terrace215
Let's not be so sure of things until we see for ourselves ok?
My gut is telling me this MC system is better than people are thinking it will be.;)
Quote:
Originally Posted by terrace215
lol, I wonder if they added that "commercially useful" phrase after the Furmark Fail of 4800s? :D
thanks for pointing that out!Quote:
Originally Posted by Movieman
please correct me if I'm wrong, but wasn't there long ago some info about Becton using FB DIMMs?
Oh! thanx for Google:
http://www.ixbt.com/short/images/kaigai394_01l.gif
That I'm not excluding, but it sure has no interest to talk about it now, as would that article suggest ;)Quote:
Originally Posted by terrace215
Your correct, there was, key word is "was"Quote:
Originally Posted by Nedjo
Buffer is on the board, memory is DDR3-1066..
The pure max power is only needed for main board engineers to design circuits capable to provide that peak power within micro and milliseconds.Quote:
Originally Posted by terrace215
But in continuous run TDP means max power consumed: if processor dissipates 125 watt an hour - it consumes exactly same electrical power.
Intel is more focussed on releasing 6core 32nm at the same time AMD will launch G34 then beckton... perhaps they might tip that up as a paper launch. beckton sampling yes since nov 2009 but very limited amount and certainly not mass delivery for oem yet while MC already is and no they are not corelated unless you count the DP 65xx series 6510 1.7ghz 4core, 6540 2.0 6 core, 6550 2.0 8 core not that earth shattering is it, lets see how many OEM really are going to adopt that part .... that is creating another platform you know... 1P + 2P neh // 2P beck // 4p + 8P beckQuote:
Originally Posted by terrace215
anyhow 1.5 weeks from now and i know exact when we can expect becky, what flavors and positioning
Oh I think my previous post was not clear enough. So:
TDP is maximal thermal power that process can emit.
And that means, that in continuous run TDP is maximum electrical power that processor can consume.
Yes, it is DDR3 1066, but every memory channel has a buffer chip ON THE BOARD! That means all the mess of FB DIMM with more power and more cost, except the cost comes at the board level, not the DIMM level. But, I guess if you are going to pay that much for a processor, you'd have no problem dropping the extra coin on the board.Quote:
Originally Posted by Movieman
I heard someone refer to beckton as "Itanium III" this past week. Feels fitting.
No, that is not correct.Quote:
Originally Posted by terrace215
And if you want to put a Magny Cours up against a Beckton plugged into power meters to see who pulls more real power at the wall, just let me know where you want me to bring the server.
We just did a power demo for the industry analysts this week and one of the best things I heard was that they were blown away at how little power 12 cores can consume.
Here is the rough equivalent.Quote:
Originally Posted by SEA
Intel:
Average power = TDP
Max power = Maximum power
AMD:
Average power = ACP
Max power = TDP
To give you an idea, the 95W TDP of the Nehalem X5570 is overshadowed by the 155W max power: http://www.cpu-world.com/CPUs/Xeon/I...602X5570).html
I am being gracious by not calling out the 197W peak.
We'll talk about our two favorite benchmarks:Quote:
Originally Posted by terrace215
SPEC Power
SPEC Wallet
I think Beckton might score some new records there. But not in the way they were hoping.
Correct on all points. I heard that also.. Not so sure if it's a true analogy but personally I always called Itanium , Itanic..:DQuote:
Originally Posted by JF-AMD
Never understood where they were going with that line over all those years but assumed they knew more than I did..:rofl:
Quote:
Originally Posted by Movieman
Is that because you know something or because you forgot to take your fiber? :hm:
:rofl: :sofa:
Quote:
Originally Posted by Movieman
It (the "Unobtanium" program) keeps HP "on board", and "aligned" with the "program" among many other "wonders".:eek:
JF one simple but very important question:Quote:
Originally Posted by JF-AMD
Why isnt AMD adopting the same TDP definition as Intel?
Marketing wise you seem more power hungry and sorry for putting it this way, but compared to the blue company you dont have as many resources to explain the difference to every company and potential customer.
Furthermore the same problem crosses over to desktop CPU's where consumers are even less aware of the difference between AMD 125 Wat TDP and Intels 95 Wat TDP. The second one seems simply as the better solution given the current market situation ... :confused:
Excuse me folks while I get out my hammer and send my friend Otis here on a long vacation!:rofl:Quote:
Originally Posted by Otis11
It's a feeling based on numbers I've seen.
MC looks to be the real deal.
Better than Westmere in some ways, a tiny and I mean tiny bit behind in others.
Thats why I said 6 to 5 and pick'em.
Now if I could get S7 to answer his damned cell phone maybe I'd have more info!:rofl:
Quote:
Originally Posted by Shadov
OK, let me try to walk through the history of power.
Intel has throttling, we do not. If their chip gets too hot, it throttles back. So, to some degree, power and performance is linked. We will have that feature in the future.
In the old days, Intel used a derate of max power to determine TDP. That was ~80% (so, theoretically, a 100W max power part would have a TDP of 80W). This is exactly how it was described by Intel engineers when I worked for an Intel-based OEM (so don't flame me on that description.)
When we brought out first parts out, they had a TDP of 95W. However, the problem was that in the worst case scenario, you'd never see more than 50W of power (heavily FP-incluenced HPC.) The way to measure max power is to run something called a "thermal virus" that basically fires up all transistors at the same time, stressing the chip in a way no real workload could.
The problem that we faced was that we were 95W and customers were provisioning their racks expecting that level of power, and it was generally in the low to mid 40's (half the power consumption.) So their racks were running inefficiently with more headroom and less density - that is bad because it eats up floor space.
So, after working with customers for a long time, the overwhelming demand was "tell us what the REAL power is, not the 'design to' power."
ACP was born.
We take several server benchmarks and run them at 100% utilization and measure power at the CPU, that becomes the ACP.
So a 115 TDP, nets a 75W ACP. And the typical customer is seeing power consumption in the low 50's. So ACP is still conservative, but closer to reality.
The real problem was that Intel had roughly 110W TDP parts in that timeframe and they were getting beaten, bad, by AMD in the power area. So, when you have a TDP of 110W and your competitor is at 95W, what do you do? You change the derate from ~80% to ~65%. Voila, lower TDPs.
The reality is, at the wall, until Nehalem, intel was soundly behind AMD in power. In today's world, in the "at the wall" benchmarking that we have seen, the standard power parts are about equal at idle, and they are higher at full load.
But the problem for them is that the power ramp is huge. If we go from idle to ~3% utilization, we jump a watt or two. They jump 10% (they have spent more time working on idle and that can cause the big ramp up.)
So why don't we just use the same measurement as Intel? Two reasons:
1. The architectures are different, and that is not a fair comparison, it favors them.
2. They have a history of changing the definition of things to suit their needs, so if we went to match exactly what they were doing, they could just change their measurement to suit their needs.
I always tell customers to ignore TDP and ACP. What really matters is TOTAL PLATFORM POWER at the wall. I believe that we have the advantage here.
I just came out of the lab last week and saw the new Magny Cours is pulling basically the same power as a 6-core istanbul. Twice the cores, massive performance increase and rougly the same power. Actually, at some utilization levels it is actually lower.
I can't speak to client power, I am a server guy.
I'll take the counter on that argument.Quote:
Originally Posted by SocketMan
HP has Intel by the nose on this one. They are 85%+ of all of the itanium sales and they have the control in this situation.
I would guess that it is more likely that HP can pressure Intel to keep itanium alive by holding its x86 business over Intel's head.
I have no idea what happens inside either of their walls, but if HP dropped itanium, intel could save a ton of money on keeping that product alive.
Why is intel pushing so hard on beckton, which will sit right on top of itanium? Probably because they want out of itanium and need an escape path. If Itanium was so strategic, it would be on their leading edge technology, but it is later than everything else, the cadence has slowed down (how many years was tukwila late), they use their old process (65nm) and they don't really pick up volume on that platform. It is life support a this point.
http://upload.wikimedia.org/wikipedi...casts_edit.png
The reality is the market has spoken on 64-bit volume technnology. It will be x86, not EPIC.