Actually, I was asking you that question as you're the one that somehow framed the original statement as a valid technology opinion. If I wasn't using a modicum of respect I would have used the name calling and generalized statements everyone else seems to resort to. I simply mentioned that the logic used made me laugh...seemed pretty preposterous to call the statement but I hope in the near future Larrabee comes in and blows them both out of the water any substantive type of opinion on the technology itself. Everyone knows that's just going to lead to flaming. In fact, many posters here seem to purposefully bait others with these types of statements.
Yes, whomever posted that shows some true wisdom, heh.....But it grows oh so tiresome to visit forums looking for information only to find half of the content filled with nothing but people sniping back and forth like 3rd graders.
If you find that ironic, I find it doubly ironic that you're responding to me at all.
Well, I think many would prefer there were no iceberg at all. And why would we need blatant examples of acceptable behavior? Simply keep those posts that are acceptable (subtly -or- blatantly?) and remove those that aren't.
No, if everything was 'swept under the rug'/simply removed then people would be more polite to one another, more respectful of what others said, and would actually consider the content of what they wrote before posting. They wouldn't post without thinking.
On top of that, all posters, those with 1 post, or those with 2800 posts, would all know the score....intuitively understand the board's philosophy, and see that the board was filled with intelligent discussion of technology only...with no other feces to muddy the water.
I'm sorry, but the thread-crapping started before I ever entered the picture. I guess you just choose to ignore that however based on your iceberg statement.
In any case, yes, I'll contribute to the actual thread.
1. I think a 20% increase from Shanghai would be fantastic if true, but obviously from the initial claims made regarding the last major AMD technology release (K10/Phenom) everyone will be skeptical of that. Of course, I think those initial claims re: K10 were probably based on simulations or perhaps just simple optimism. Question is, in what true context was the statement being made for Shanghai and in regards to what particular performance metric or set of metrics?
2. When it comes down to it, we'll not know the true story until an actual product release, but it is interesting to hear these tidbits of information as they come out.
3. I hope both Intel and AMD are successful with their products as I believe there is room in the marketplace for two major x86 companies. They are obviously making each other better, pressing technology to limits, and ultimately benefitting the rest of us in the end. To hope for poor performance and/or poor market reception for a product which could ultimately spell the end of one of these companies shows ignorance as to the big picture.
But if SHANGHAI A.K.A Deneb has 35 % performance improvement overall doesn't it put Deneb very close to Core i7 in terms of performance.
At 35% performance improvement it should be said that Deneb will definitely beat penryn.
At 35% performance improvement it should be said that Deneb will definitely beat penryn.
cusideabelincoln
Diamond Member
- Aug 3, 2008
- 3,275
- 46
- 91
Cookie Monster
Diamond Member
- May 7, 2005
- 5,161
- 32
- 86
Since the initial wave of shanghai CPUs will support HT1.0 only, i think these CPUs wont perform to their full potential until next year when HT3.0 supporting shanghai/deneb start rolling out. Im not sure why there couldnt get HT3.0 support in there.
Its based on the same design. Thats like saying an Opteron is 'different' to a Phenom/X2, or a Xeon is 'different; to a C2D/C2Q. They essentially the same silicon, rebadged for different markets.
It's based on the same core, not the same design. For example, a socket F (1207) chip has a peak of 24GB/s bandwidth on the cHT links, and the AM2 version (Phenoms) has 16GB/s peak bandwidth...that is bandwidth per chip BTW.
Understandable given the bandwith requirements of server apps, but are there any other major differences? At the core level it is the same chip.
Originally posted by: harpoon84
Which makes the numbers that much harder to believe.
Sorry double post.
Read below.
Originally posted by: harpoon84
Which makes the numbers that much harder to believe.
I kinda agree with you.
I think 35% performance gain must have been peak gain achieved when compared against Barcelona in some particular scenario's , but if on average 20% performance improvement is achieved compared to Barcelona THAT'S AWESOME FOR A DIE SHRINK.:thumbsup:
But just in case lets say AMD manages to PULL 35% performance increase on average , wouldn't that be some kind of world record for performance gain achieved just from a die shrink , because I think 35% performance increase is only achievable from a new architecture , RIGHT PPL ? . :shocked:
I believe the 35% figure includes the additional clockspeed, ie. a 3GHz Shanghai will be 35% faster than a 2.6GHz Barcelona for instance, so 20% higher IPC + 15% higher clockspeed.
Still, I think 20% IPC gains from a die shrink sounds extremely optimistic, it would certainly be a 'world record' if it turned out to be true, no die shrink I can think of has increased performance by more than 10% per clock, and that was way back during the P4 Willamette -> Northwood shrink. In more recent times Penryn gained 5 - 6% per clock against Conroe and that was considered impressive.
I'll remain skeptical until AMD provides hard numbers, they pulled a similar stunt with Barcelona claiming it would be '40% faster than Clovertown over a wide range of workloads' - once bitten, twice shy.
Still, I think 20% IPC gains from a die shrink sounds extremely optimistic, it would certainly be a 'world record' if it turned out to be true, no die shrink I can think of has increased performance by more than 10% per clock, and that was way back during the P4 Willamette -> Northwood shrink. In more recent times Penryn gained 5 - 6% per clock against Conroe and that was considered impressive.
I'll remain skeptical until AMD provides hard numbers, they pulled a similar stunt with Barcelona claiming it would be '40% faster than Clovertown over a wide range of workloads' - once bitten, twice shy.
Certainly the design of the cores is the same, but the design of the chips is slightly different (as the max clockspeeds often are). The Opterons and Xeons have quite a bit more circuitry for virtualization, RAS enhancements, bandwidth, etc...
Certainly they aren't just re-badged chips...
The 35% increase does indeed include the additional clockspeed...(which is what AMD said in the first place).
So it's 20% faster at the same clockspeed, 35% faster overall...
Edit: BTW, it's not just an optical die shrink (though I don't yet know all the specifics), so it's nowhere near a world record...
But according to AMD 35% is clock-to-clock increase , i.e A Barcelona cpu at 1ghz would under perform by 35% compared to a shanghai cpu at 1ghz.
May be I didnt understand the interview properly , MY BAD.
Originally posted by: BLaber
But according to AMD 35% is clock-to-clock increase , i.e A Barcelona cpu at 1ghz would under perform by 35% compared to a shanghai cpu at 1ghz.
No, what they said (according to Phynaz's post) was:
1. According to Patla, Shanghai will offer 35 percent more performance and 35 percent better power efficiency when compared to Barcelona products.
2. At the same frequency (speed), Shanghai will outperform Barcelona by about 20 percent
Note that point one is comparing 2 product lines in total, and point 2 compares at the same clockspeed...
So Viditor, do you consider the 20% IPC gains realistic? I know you've been very bullish about AMD's prospects in the past, what about these latest claims?
AMD Shanghai chip gets thumbs-up from OEMs
Chip maker ships out new 45nm quad-core processor to vendors for early look-see
An AMD spokesman confirmed today that the company has started shipping production-quality Shanghai chips to OEMs so they can make final validations.
Industry analysts said that shipping the quad-core server processor -- the company's first 45-nanometer chip -- to OEMs for an early look-see is a very good sign for AMD. The new Shanghai chip was originally scheduled for release in the first quarter of 2009, but AMD has pushed the ship date ahead to the fourth quarter of this year.
Rob Enderle, an analyst at Enderle Group, said that when Barcelona was in the works, AMD executives were distracted by the chip maker's acquisition of ATI Technologies Inc. during 2006. Now, with the acquisition settling in and new leadership at the helm of AMD, the focus is back on processors.
"The combination of the higher level of focus and the less risk that they're taking is resulting in them beating their own date," said Enderle, noting that the Shanghai chips may be cast on a new 45nm die but architecturally are quite similar to Barcelona. "Right now, the OEMs seemed to be pretty happy with it. One, it's showing up ahead of schedule, and it seems to be performing well against expectations. We're still at the front end of sampling, so we won't know how well it's performing for a few weeks yet, but so far, the OEMs like what they've seen," he added.
Dan Olds, principal analyst at Gabriel Consulting Group, said an OEM told him that after an evaluation, it plans to use the chip in systems. "I think we're beyond the 'looking/evaluation' stage and into the 'let's build the systems that will use it' stage," he added. Olds noted that the vendor, which he declined to identify, plans to ship Shanghai-based systems as early as the fourth quarter of this year, but possibly in the first quarter of 2009.
Source : http://www.computerworld.com/a...omyId=162&pageNumber=1
Looks like AMD's claim of performance improvement has some weight to it , if not OEM's would not like it , would they... :sun:
Chip maker ships out new 45nm quad-core processor to vendors for early look-see
An AMD spokesman confirmed today that the company has started shipping production-quality Shanghai chips to OEMs so they can make final validations.
Industry analysts said that shipping the quad-core server processor -- the company's first 45-nanometer chip -- to OEMs for an early look-see is a very good sign for AMD. The new Shanghai chip was originally scheduled for release in the first quarter of 2009, but AMD has pushed the ship date ahead to the fourth quarter of this year.
Rob Enderle, an analyst at Enderle Group, said that when Barcelona was in the works, AMD executives were distracted by the chip maker's acquisition of ATI Technologies Inc. during 2006. Now, with the acquisition settling in and new leadership at the helm of AMD, the focus is back on processors.
"The combination of the higher level of focus and the less risk that they're taking is resulting in them beating their own date," said Enderle, noting that the Shanghai chips may be cast on a new 45nm die but architecturally are quite similar to Barcelona. "Right now, the OEMs seemed to be pretty happy with it. One, it's showing up ahead of schedule, and it seems to be performing well against expectations. We're still at the front end of sampling, so we won't know how well it's performing for a few weeks yet, but so far, the OEMs like what they've seen," he added.
Dan Olds, principal analyst at Gabriel Consulting Group, said an OEM told him that after an evaluation, it plans to use the chip in systems. "I think we're beyond the 'looking/evaluation' stage and into the 'let's build the systems that will use it' stage," he added. Olds noted that the vendor, which he declined to identify, plans to ship Shanghai-based systems as early as the fourth quarter of this year, but possibly in the first quarter of 2009.
Source : http://www.computerworld.com/a...omyId=162&pageNumber=1
Looks like AMD's claim of performance improvement has some weight to it , if not OEM's would not like it , would they... :sun:
I certainly consider it possible...
We should know pretty soon, as they've started shipping samples to OEMs already.
Beyond that, I really have no idea...
Well both the analyst and the OEM have considerable conflicts of interest when it comes to this.
Neither the analyst nor the OEM are the end-customer of the chip, they are both engaging in as much of a PR machine as they foresee themselves making a paycheck from.
When a customer, end-user, starts to say "yeah, we likes, we buys" and isn't ashamed of saying it (i.e. doesn't require they remain anonymous like the OEM's did here) then we got ourselves some weight behind the claims.
There were no shortage of analysts or OEM's touting the wonderful attributes of Prescott.
Buyer beware when you are on the receiving end of unsolicited free information from an analyst. You are getting their advice for free for one reason and only one reason.
I'd say the same thing if this Analyst and his anonymous OEM's were bashing Shanghai versus lauding it. And no, you don't get a free iPOD for being the 1,000,000 person to click the link
Originally posted by: Idontcare
Well both the analyst and the OEM have considerable conflicts of interest when it comes to this.
Neither the analyst nor the OEM are the end-customer of the chip, they are both engaging in as much of a PR machine as they foresee themselves making a paycheck from.
When a customer, end-user, starts to say "yeah, we likes, we buys" and isn't ashamed of saying it (i.e. doesn't require they remain anonymous like the OEM's did here) then we got ourselves some weight behind the claims.
There were no shortage of analysts or OEM's touting the wonderful attributes of Prescott.
Buyer beware when you are on the receiving end of unsolicited free information from an analyst. You are getting their advice for free for one reason and only one reason.
I'd say the same thing if this Analyst and his anonymous OEM's were bashing Shanghai versus lauding it. And no, you don't get a free iPOD for being the 1,000,000 person to click the link
Yup Just like build-up before Barcelona's release , build-up to shanghai's release is looking ALL BUT GOOD. :roll:
Just hope AMD really lives up to expectation on this one.
Made the L3 cache run faster than 1.8 ghz. Upping the cache clock rates to be synchronous with the cores would do it.
But then the chip would use an enormous asspile more power.
So, in conclusion, yeah. Seems like one of those claims is fairy dust. We wait, we see, and in most likelyhood a certain forum member gets a fresher quote for his sig.
It's going to be really sad if it does in fact turn out that Barcelona had it's IPC castrated by 20% solely because the L3$ was underclocked and 4M too small.
Seems like one of those things that would have been kinda obvious in the very earliest of simulation passes way way way before tapeout. Missing a 20% IPC bottleneck is huge, too huge for me to believe.
In my opinion, if 4M of faster L3$ was all that Barcelona needed in order to have 20% higher IPC then you would have either seen a larger die (despite higher cost, it would not have been 20% higher) and higher TDP (4M of faster L3$ would not have increased TDP by 20%).
I'd believe a 5-10% IPC improvement comes from the larger L3$. In-line with the die-size increase and TDP increase it would have brought to the table, thus making it more of an acceptable loss by excluding from Barcelona.
There must have been something in the core architecture itself that was included in Shanghai but left out of Barcelona, perhaps for timeline considerations and validation purposes.
I'm sure Anand will have a great writeup on it, if not then realworldtech will have one for sure.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter