SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  For example, here is how to disable FireFox ad content blocking while on Silicon Investor.

   Technology StocksAdvanced Micro vs Intel (AMD / INTC)


Previous 10 Next 10 
To: Yousef who wrote (2560)2/8/2008 7:56:08 PM
From: Jim McMannis
   of 2581
 
What the heck are they doing at AMD? Sitting around all day? It's like the keystone cops.

Share RecommendKeepReplyMark as Last ReadRead Replies (1)


To: Jim McMannis who wrote (2561)2/9/2008 12:56:49 PM
From: mas_
   of 2581
 
Price of not executing at 65nm in either process or design.

Information from Power 6's power breakdown shows a very leaky 65nm process.

realworldtech.com

To fix this for normal and low-power desktop use AMD had to increase the gate thickness on their variant which limited their scaling.

Message 24023068

The TLB bug should not have escaped validation or pre-OEM testing either.

45nm sounds a lot better of the bat, probably due to the addition of immersion technology, and it needs to be if they are going to hang on until Bulldozer arrives.

siliconinvestor.com
siliconinvestor.com

Share RecommendKeepReplyMark as Last Read


To: Yousef who wrote (2560)2/10/2008 4:44:40 PM
From: tecate78732
   of 2581
 
Mas is Jimmy, how quaint.

Share RecommendKeepReplyMark as Last ReadRead Replies (1)


To: tecate78732 who wrote (2563)2/11/2008 12:24:52 AM
From: Jim McMannis
   of 2581
 
Nope

Share RecommendKeepReplyMark as Last Read


From: pgerassi8/13/2008 6:59:01 PM
   of 2581
 
RE: Message 24843496

Classmate hasn't sold them yet. 300K XOs were sold by 4/08 and more after that.

By 7/08 the total was 667,000 per:

en.wikipedia.org

Classmate OTOH has sold to date less than 100K per:

pocket-lint.co.uk

Wbmw should read that Portugal may sell up to 500K Classmates. When (or if) those sales will be completed is anyones guess. But one shouldn't count them until they are really sold to end users. But even then, the whole amount is less than XOs (OLPC) above 667K which is still greater than 600K (however long it takes Portugal to sell them).

Pete

Share RecommendKeepReplyMark as Last Read


From: pgerassi8/13/2008 8:36:07 PM
   of 2581
 
RE: Message 24843540

Wbmw you are wrong even more. The TDP quoted includes the Geode 5536. Unless you are using the maximum TDP which Intel does not state for Atom (or any of its other CPUs). And apples to oranges comparisons are worth less than the screens they paint. To compare to Atom or any non ODMC Intel CPU, you must use the core only. Which the Geode LX 800 CPU core uses 0.9W per the whitepaper I linked (which you evidently didn't read or you wouldn't make such an outrageous claim). The Geode LX 700 (400MHz CPU clock) uses 0.8W and 2W typical with the Geode 5536 chipset straight from the XO technical briefs. Thats the actual use during Winbench 99 with Windows XP (MS sells for $3 for use on XOs).

BTW the Geode 5536 has USB 2.0 ports, IDE interface, AC97 and much more. The Geode GX and LX dies have a PCI interface, which have more stuff able to be attached than PCIe does, even now, especially in the embedded world. They also have a GPU (2D only, but weren't you saying that is all that is needed for T&Ls?) with TFT outputs and CRT outputs (much more likely in the low cost 3rd world). PCIe isn't needed especially in the low end. You do see the 5W typical power budget for XO don't you? Atom CPU alone uses too much damn power for it to be considered, 2.5W typical compared to Geode LX 700 plus Geode 5536 combo's 2W typical. Add to that all of its chipset power hogs like memory, IGP, PCIe, etc. and you use 6W more than budgetted.

If you want to compare 8W that the Atom chipset combo uses, that it about what a Geode 5536 like mobile NB/SB uses with a Geode NX1500 (an Athlon XP-M at 1GHz). Which runs circles around Atom. Case in point, POV-Ray completes in less than 73 minutes compared to 294 minutes for a 733MHz P3 Celeron, a 3 issue OOE processor, likely twice as fast as a dual issue in order CPU. For the P3 Celeron would have to clock at least (assume linear which is highly doubtful it would be that good) 3Ghz to render in that small a time. 3.3x the performance for perhaps 2W TDP more than a 1.8GHz Atom, both with chipsets.

Then there is the 1GHz Sempron 2100+ S1 (90nm SOI Keene) which uses 9W TDP max (65nm SOI Sherman uses 8W TDPmax) and typically about 5W including a dual channel DDR2 controller. With all the typical stuff on a mini ITX MB with 2 channels of DDR2-667 memory, 2 GbE, 4 SATA and 690G chipset, power usage is 16W TDP max. Far more than is included with Atom and its chipset (2GB DDR2-400/533/667 2-200pin SO-DIMMs memory, 3D Radeon Express 1250 IGP, 2 GbE ports, 4 SATA, PCI slot, PS2 Mouse & KB ports, 4 serial I/O ports, 8 USB 2.0 ports, DVI, VGA and more). POV-ray (64 bit mode) took only 44 minutes. That means the P3 Celeron would have to run 4.9GHz, so that would have about 5.4 times the performance for less than twice the power.

So more power gets you lots more performance.

As for the CPU, although the Geode LX 700 can go 433MHz, the OLPC uses it at 400MHz. So the Atom equivalent would be about 1.52GHz. Still quite a bit of performance for only 2W typical with the Geode 5536 chipset included. And it can still be made with foundries like TSMC or Chartered. For a whole lot less than Atom can be made for.

Geode GX and LX are normally found in embedded applications that go for about $45 for the entire system. Geode NX series is vastly more powerful than Atom, yet is used in systems of about $70 to 100. Low power Keene and Sherman K8 Semprons are used in systems of $100 to $150 including the M690G and MB600. Soon that will change to M740G/80G/90GX with MB700/50. They are way beyond Atom in performance, able to run 64bit Vista Ultimate with all the trimmings.

Pete

Share RecommendKeepReplyMark as Last Read


From: pgerassi8/13/2008 11:58:47 PM
   of 2581
 
Wbmw:

And the XO isn't a PC either. It has only 128MB of RAM. So how does Atom with 2GB makes for an apples to apples comparison? Try it with 128MB of slow DDR 200MHz and then compare speeds. Besides, the Atom is dead slow. As to your contention that it matches a Banias at 1.2Ghz according to "my reviews" and not back that up with links. BTW a 945G chipset with a 1.6GHz Atom that uses 60W only gets 46 minutes on the a POV-Ray test (not the standard test used in the link you provided).

A 2.13GHz Core 2 Duo T2600 got 330 seconds on a POV ray test (another one it seems) while the 1.6GHz Atom got 1660 seconds. That puts the Atom's performance at 20% of the C2D 2133MHz on the exact same MB using 75W. Or the equal of a SC Banias (Pentium M) of 850MHz assuming the best linear clock performance scaling on the link you provided (which obviously isn't seen on those benchmarks). That matches about a 600MHz K7 (Duron). All at the same 512MB to 1GB RAM. Of course that is well beaten by a 1GHz Geode NX1500 (Athlon AXP-M). Its no comparison to a 1GHz Sempron at 8W (65nm SOI Sherman) on a ITX M690G/MB600 using only 15W, 1/4th what the Atom/945G above uses with 1GB DDR2. There is more good stuff on the Sempron ITX board including the far better 690G IGP and video decoder.

An ITX MB using the higher performing M740G/M780G/M790GX & MB700/MB750 could throw a monkey wrench by using the RV610 IGP's GPGPU to speed up POV-Ray renders. 400/500/700Mhz 40 FPMADD's per clock is vastly larger than that of a 1GHz SC K8, 32/40/56 SP GFlops versus 4 SP Gflops.

As for the 733MHz Xbox, the small amount of RAM in an Xbox matches what the Geode LX 800 has to deal with. That is likely the source of the large render time difference as the render has to swap to disk (Xbox) or flash (XO).

Thus your contention that a 1.6GHz Atom would have 4-5 times that of Geode LX 700 at 400MHz vastly overstates the real case. Even using conservative scaling, Atom delivers less than 850MHz Banias performance which is fairly close to what a 850MHz P3 did given much more memory. What any review of yours likely forgot to use just 128MB of slow memory to do the performance comparison to Geode LX 800. The Xbox render probably does a good job of that. All of a sudden Atom is less than 50% faster assuming linear scaling. With poorer scaling factors, it may not even beat Geode LX 700 (400MHz).

One benchmark that Intel would avoid with passion is anything using AES 128 de/encryption as Geode GX and LX have hardware acceleration built in. Tests show that it is faster than a 3GHz P4 which likely would blow Atom completely out of the water and keep up with slow Turions in AMD64 mode.

Pete

Share RecommendKeepReplyMark as Last ReadRead Replies (1)


To: pgerassi who wrote (2567)8/14/2008 8:46:05 AM
From: dr_elis
   of 2581
 
The Atom maybe slow but consumers like Atom-based netbooks/mini-notebooks a lot. Models like the MSI Wind, the Asus Eee PC, or the Acer Aspire One, all based on the 1.6 Ghz Atom CPU, are selling like hotcakes. Overall, Atom-based netbook sales could reach a figure north of 10 million this year. That´s a nice additional revenue stream for INTC, which was basically non-existent last year. And this is just the beginning.

Share RecommendKeepReplyMark as Last ReadRead Replies (2)


To: dr_elis who wrote (2568)8/15/2008 6:29:34 PM
From: pgerassi
   of 2581
 
Dr_elis:

Unfortunately for you, a third party tester has found that Atom systems use more power than a 1GHz Athlon 64 2000+ system while being much slower. And Atom can't even do HDTV because its chipset is buggy as hell.

I see big lawsuits against Intel in the near future.

Pete

Share RecommendKeepReplyMark as Last ReadRead Replies (2)


To: pgerassi who wrote (2569)8/15/2008 9:54:28 PM
From: combjelly
   of 2581
 
"Unfortunately for you, a third party tester has found that Atom systems use more power than a 1GHz Athlon 64 2000+ system while being much slower."

To be fair, that is only for the ones intended for desktop/netbook/embedded use. The one intended for MIDs is better on power consumption.

AMD needs to do the BGA version of that processor. That, and the eventual move to 45nm and a lower voltage, could mean a very compelling product.

Share RecommendKeepReplyMark as Last ReadRead Replies (2)
Previous 10 Next 10