Charlie at INQ reckons Intel in deep dodo
Dude, it's the platform that matters. Charlie commented about the sucking INTEL graphics without noting the fundamental principle I talked about a long long time ago.
It doesn't even matter if your CPU is 10% faster, when your graphics and memory subsystem suck, the whole computing experience is degraded back to 486 levels. Charlie was talking about gaming, it wasn't just those DOOM3 FPS games, even kids games today demand high level of 3D graphics--something Intel doesn't have at all.
Not just games. Even Windows itself requires high level of 3D support. Otherwise, your PC/notebook looks like a poor 3rd world kid's toy without the modern bells and whistles of Vista. As I wrote here, the Microsoft + AMD alliance will kill off Intel with Vista.
Going forward, 90% of Intel's IGP graphics become total junk, that will be at least $1 billion loss of quarterly revenue. Nvidia doesn't have an IGP solution for Intel, AMD won't create a IGP solution for Intel, Intel is stuck with itself -- and Intel doesn't have a discrete graphics either. Even if Nvidia is so forgiving of Intel and starts creating an IGP for Intel, it will take months to finish and will eat Intel's own IGP business alive.
Now, you understand why AMD is smart and AMD+=ATI is the killer enterprise I declared before the merger was announced.
AMD's CTO was onto something here. He called it killer micro. Hector Ruiz hinted something before too, he said AMD will introduce an architecture that will be a real killer. The timing is 2008. The increase of computing power is two orders of magnitude, or 100X.
The question is, can Intel catch that in time? I doubt it. AMD had a 20% lead over Intel since 2003, but it lacked the OEM support to grow the business. Today, the situation is different, everyone is on AMD bandwagon. In 2008, both FAB36 and FAB38 will be producing 45nm at a total of 45000wspm. 100x performance, DELL, HP, IBM, SUN, GATEWAY, LENOVO, plus a bunch of japanese firms.... But, I told you Intel will BK by 2Q08 ....
posted by Sharikou, Ph. D. @ 11/07/2006 08:19:00 AM
19 Comments:
two orders of magnitude, or 100X
shouldn't that be 100%
two orders of magnitude, or 100X
shouldn't that be 100%
No. 100x. As Hester said, GPUs are 50x faster than CPUs now, in 2008, GPU frequency will get a 4x boost from Fusion
It's amusing to me that you have tacitly admitted Intel CPUs crush AMD CPUs and have moved on to your ridiculous crusade about integrated graphics.
Integrated graphics are for cheap machines, especially home web browsing and business PCs. 3D power isn't that important to these users, and 965G is, contrary to your blatant and boringly predictable lies, perfectly adequate to run the Aero interface at full performance.
Still, it's nice to see you grasping at straws like this. It shows how pathetic you are. What's next, ramblings about how the USB controller on AMD machines will be slightly faster and nobody wants "Intel ultra low end USB controllers"?
"GPU frequency will get a 4x boost from Fusion"
Well, G80 technically runs at 1300MHz already.
I would think Fusion becomes something similar to Cell but instead of PPE and SPE's there is one or more CPU cores and one or more quads or what ever unit will be used in that GPU.
I hope that kind of thing can be used as a massively parallel (and not all that flexible) SIMD unit by the CPU. Hello x87 addon, only this time it is added to the CPU.
Though I wonder what would they be using to deliver enough memory bandwidth to that monster. Dualchannel DDR2/3 is not nearly enough. Even my several years old rather low-end FX5600 had much more bandwidth than today's top of the line CPU's. If you add insanely powerful SIMD units to a CPU it will be constantly starved unless something like XDR or FBDIMM is used
Integrated graphics are for cheap machines, especially home web browsing and business PCs. 3D power isn't that important to these users, and 965G is, contrary to your blatant and boringly predictable lies, perfectly adequate to run the Aero interface at full performance.
You are missing the whole picture. Why do we need faster PCes? We need them for those demanding tasks, such as kids 3D game and HDTV stuff. We don't need a faster PC for MS Word, for that a 1GHZ CPU+2D graphics is more than enough.
When we do need some PC performance, it's the whole system performance that matters. Graphics and memory bandwidth is the key for overall performance. We don't want a crappy Intel IGP that can't play anything meaningfully 3D, and you don't want an Intel IGP that chokes on doing Aero. Choosing Intel would invite ridicule and resentment from your kids, who find their RollerCoaster Tycoon can't run smoothly on Intel ViiV.
So, the wise thing is to choose AMD. You know with AMD, you either get a Nvidia or AMD graphics, both are DOOM3 capable and more than enough to handle Vista+RollerCoaster...
The choice is a no brainer.
Many of you are missing the point.
70% of the market uses integrated graphics. Intel locks its OEMs and channel partners into its crap (G965) graphics with Centrini, ViiV, and now VPro.
With Vista, the public will finally realize how badly Intel has shafted them. The only real way to test graphics is a side by side demo.
Finally calling it Doom 3 huh? I guess you got ahead of the game with DOOM4!
Now if you could only get your headlines straight: 'Doodoo', not 'dodo', dippydoo.
You do have a point thought - integrated graphics are important for people who don't know better.
No one wants to buy a PC off the shelf and find it doesn't work with new software.
However, the current shovel of IGP's from Intel are just fine for those people.
Any kid who wants high performance gaming will surely let his parents know that he wants a video card as well. Any salesperson worth their salt will mention that an add-in video card makes good sense when showing off a machine. No one of these people will settle for onboard IGP from AMD either. So your point is pretty much nullified.
It doesn't make sense that you pick and pick and pick on this little insignificant (and moderately untrue) point. You have nothing left to criticise Intel about?
Recently you've criticised Intel for:
1) Inadequate IGP performance.
Proven untrue - low end onboard Intel works just fine for Vista.
2) Exploding batteries.
Also proven untrue, and industry accepted as battery fault, not the processor. You know nothing about electricity or batteries if you say otherwise. I can directly short circuit any properly functioning battery and it will not explode. Even IF Intel's CPU's are short circuiting completely, a battery won't blow. No, you twit, it's the battery creating gases, case closed. Give it up.
3) Poor Conroe performance.
Budget Conroes outperform top of the line AMD solutions in most cases. Dual Core AMD cpu's do not perform as well as a dual core Intel. Simple. Maybe in one odd test out of ten they will outperform a similar Conroe, but across the board Conroes are the best performing - especially at equivalent clock speeds.
Also, your prophetic and deep understanding of the PC industry is overstated:
-The idea that Dell might sell AMD products was a 50/50 call. Either they would or they wouldn't. You picked that they would. If you asked someone off the street who didn't ever heard of Dell or AMD they would have just as much chance of getting it right as you had.
-You claim industry reviewers to be biased and paid pumpers, yet it has been proven that their data is accurate.
-You pick and choose benchmarks and when they are valid. Yet when confronted, you claim that a benchmark is worthless.
-You continue to say Intel's impending bankruptcy will happen sooner and sooner because of this and that, yet your prediction date advances further in the future. This spring you were saying 5-7 quarters, and you are now still saying 5-7 quarters. At the same time, every third or fourth post by you is some sort of catalyst that will hasten Intel's bankruptcy. How long will you keep saying this?
What HAVE you said or predicted that is provable?
Er, if you want to play new 3D games all integrated stuff sucks so you'll end up buying a separate card anyway. Gaming on a Geforce 6150 or Radeon Xpress is hardly an improvement on Intels stuff; all get soundly beaten by the cheapest X1300 or 7300 card you can find.
--You are stupid for the following reasons:
1) Live and Viiv are not meant for game playing. They aren't DOOM3 consoles. They are media centers and/or home theater PCs. Repeat after me, Live/Viiv boxes are not for games.
2) HDTV "tasks" have nothing to do with 3D graphics chipset performance. The CPU does most of the work in these applications.
Sharikou said...
"In 2008, both FAB36 and FAB38 will be producing 45nm at a total of 45000wspm."
Still no link?
Well, I like Charlie, but my thoughts are Intel and AMD both have problems.
1.Intel selling of divisions to cut costs, some of these divisions like the X-Scale could have been turned around IMHO.
2. AMD does not have a 65nm chip out and their capacity is constrained very badly atm.
3. Intel stuck with the NB FSB instead of an integrated memory controller.
4. Intel does not have a DX9 or 10 IGP or discrete graphics chip.
5. AMD should have forgone 4x4 and produced a non-native quadcore MCM.
6. Intel still will be developing the money loosing abomination that is the Itanium.
7. AMD still has the Anti-Trust lawsuit against Intel,though they could afford to buy ATI, a top 10 semiconductor company.
8. AMD still refuses to market worth a darn, how can they compete if they don't market?
9. AMD is becoming quite an attractive target for a PE Firm, don't think there are not ones with the cash to buy it. A PE firm could bankroll a couple of fabs and grow AMD faster than a cash strapped AMD can on its own.
"In 2008, both FAB36 and FAB38 will be producing 45nm at a total of 45000wspm."
Still no link?
Dude, your memory is failing. I can't keep posting AMD's tech analyst slide all the time...
Sharikou said...
"In 2008, both FAB36 and FAB38 will be producing 45nm at a total of 45000wspm."
So are your suggesting that both FABS will be fully ramped at 45nm and only 45nm in 2008 and allowing them to have 45,000 wspm at 45nm?
I really just want to be clear here, because the slides say...
----------------------------------
FAB 38 - "20K wafers/month expected by Q4 2008"... Page 4.
"On track for volume 45nm production by mid-2008"... Page 6.
"Expectmid-2008 45nm introduction"... Page 26.
----------------------------------
So please clarify is it leaving 2008 or starting 2008?
In either case they won't have 45,000 wspm until the 4th quarter 2008.
Well, I heard somewhere that Intel's having something like 10-20 (forget the exact number) multi core projects going on right now. It doesn't seem something a company in deep dodo would do (or maybe it does?).
"It's amusing to me that you have tacitly admitted Intel CPUs crush AMD CPUs and have moved on to your ridiculous crusade about integrated graphics."
No, it's the thinking like yours that is amusing. You are just like the numerous clueless amateurs who think the way Intel markets.
One thing that Intel Core 2 does perform better than AMD K8 is SSE. Heck, this is even true with P4. Another thing is complex pointer-memory access instructions (due to fusion and better memory dis-ambiguation), which probably contributed to the good SuperPi results.
Other than these, Core 2 has about 5-10% advantage in average over K8 at the same clock rate. For business applications, Core 2 and K8 are about even; for memory intensive applications such as web serving, Core 2 hardly wins over K8; for many cryptographic algorithms, Core 2 is even slower than K8 at the same clock rate.
Note that performance varies from one app to another even in the same category. There are hundreds of thousands mission-critical applications that require better performance, and I assure you SuperPi is one of the least significant, among which are some other "benchmarks" performed by those online review sites over and over.
So while Core 2 has a good core and works well with SSE, it is nowhere crushing K8. If you had to run a high-performance security server, or a high-bandwidth game server, or a high-traffic web server, K8 would be the better chocie, since Core 2's good SSE, large cache, or better memory dis-ambiguatious doesn't help very much here. OTOH, if you had to transcode video/audio with SSE or do complex AI (where some complex memory accesses take place, e.g.), then Core 2 would be the choice.
The conclusion is, Core 2 is nowhere crushing anybody. To believe otherwise is most definitely due to ignorance of computer performance other than the (same set of) benchmarks performed over and over on those online "review" sites.
PhD Pretender wet dream
""In 2008, both FAB36 and FAB38 will be producing 45nm at a total of 45000wspm."
How will they do that. They aren't even on 65nm yet. Simply not possible to retool for 65nm then comeback a year later and do 45nm. Unless that 45nm is one crappy 45nm process. I expect at 45nm the leading process companies will be making big leaps to provide continued performance.
AMD simply doesn't have the money to fund R&D while also running their factory. Simply doesn't work. They are too small, not enough people, not enough money, not enough technology know how. Man they don't even develop their own technology. They pay a bunch of skanks in East Fishkill to pick the tools, set the design rules, develope the recipes. How can you compete like that. It the same people that bowed out of DRAMs, HD, PCs, OS2...
Your BK prediction is still absurd just as the BK predictions by Intel fans for DEll and AMD are.
I've had a nearly perfect prediction record for the past 3 1/2 years. A more accurate prediction is that sometime during 2008 Intel's monopoly will break. I know that some like to think that Intel is taking back share from AMD but I haven't seen one single statistic to show that Intel has take back even 1/10 of 1% yet. I would say that by the end of 2008 AMD will be in the 3rd-5th ranking worldwide. This will be quite a bit of growth from the current 13th place. I would also imagine that AMD will announce a third FAB toward the end of 2007 and should have about 40% share by 2010. At this point Intel and AMD would essentially be equal.
There is no way that Intel will go bankrupt in 2008. AMD cannot grow quickly enough to capture that much market share unless there is a worldwide recession during 2007. Intel is fully capable of paying its bills with 70% of the market. The only thing this will effect is its plans for future expansion. You may see FAB upgrades shelved and existing FABs becoming idle. I doubt it will have any effect on its current production FABs or the two new 45nm FABs. However, it is possible that one of the current production FABs may not be upgraded to 45nm. This is not a bankruptcy scenario.
It's now too late for Intel to put the AMD genie back into the bottle. AMD should be able to outpace Intel's growth all during 2007 and 2008. I do really wonder if the glowing reports about Intel's roadmaps are true: desktop processors at 4.0Ghz while drawing 30% less power. Quad core chips at 3.73Ghz. All arriving the second half of 2007. I guess we'll see.
It doesn't even matter if your CPU is 10% faster, when your graphics and memory subsystem suck, the whole computing experience is degraded back to 486 levels.
Comments like this are just further evidence that Sharikou lacks any intelligent thought. All your predictions will turn out to be wrong and you will be completely insignificant. My prediction, Sharikou's credibility will BK this quarter.
So, the wise thing is to choose AMD. You know with AMD, you either get a Nvidia or AMD graphics, both are DOOM3 capable and more than enough to handle Vista+RollerCoaster...
The choice is a no brainer.
Totally agree, people without brains make this choice. Intel Core2 Duo and NVidia SLI kick crap out of AMDs best offering today.
Post a Comment
<< Home