Story 30,000 Afraid of retail? Us?
INTEL ALMOST ALWAYS puts out pretty high quality documents and benchmarks, and if anything, gives its competitors a fair shake - the graphics world, this is not. That is why I found a presentation entitled 'Intel Platform Competitive Guide_Feb 15 2006_Rev1' such a shock, because it is so beyond bad and slanted it is embarrassing. The document, a 32 page presentation, is obviously aimed at the sales set, and has 'Approved for NonNDA (sic) Use' all over it. It takes severe pot shots at AMD the whole time in a very un-Intel way.
Anyone even casually following the market will know that the current K8s simply abuse the P4 offerings in every respect, they are usually faster, cheaper and more efficient. P4s will win here and there, but overall, it is not a fair fight. Now reading the IPCG, you would come to the conclusion that it is not a fair fight, just with a different winner.
The document has five sections, 1) Platform Architecture Comparison 2) Intel's Dual-Core Advantage, 3) Platform and Ecosystem Advantage 4) Mainstream Competitive Matrix and 5) Conclusions. The first part, Platform Architecture is embarrassing, it starts out with Intel having a larger cache, and then goes on to point out in big yellow balloons that AMD has 2.5X more cache misses, but it conveniently leaves out latency. The huge win it quotes is SPECint_base2000, where Intel wins with a whopping score of 1713 to 1629. Now if you go to an independent source like
Ace's Hardware, you see that Intel gets its ass kicked in an unfriendly way, and if you use peak scores, it gets worse. 0:1, advantage AMD.
The next three pages go on to tout how much of an advantage an external memory controller is, and how AMD is 'chained' to the older memory technologies. This one is true, AMD is locked to DDR for a bit longer, but until recently, this was a decided plus. Either way, it only quote bandwidths, never latency. Very selectively picked illustrations, but some are right, advantage neither.
The last slide is the only one I don't have a technical problem with. Intel points out that any graphics controller is two hops to memory, chipset to CPU, on AMD where it is one hop on Intel. While that is indeed the case, AMD seems to pound Intel into the ground on almost every gaming benchmark there is, so I am not sure about how relevant it is in the real world. Again a tie.
The second section, Intel's Dual-Core Advantage is where it gets laughable. It starts out with a classic listing of its own code names as 'good' and since its competitors don't have the same names, they must be bad. Call me crazy, but I'll take an SLI chipset with Nv Raid over a single slot Intel with Matrix Storage any day, thank you. The one that I will dock Intel the most points for is that it claims that a GMA 950 based (i945G) integrated graphics chip is better than 'limited support 3rd party support' on the AMD side. Oh yeah, it wants you to believe it will run modern games too. Hands up anyone who thinks ATI and NV graphics are inferior to the Intel GPU machine. Advantage AMD here, 0:2 total so far.
The next one goes on to show how an FX-60 using Intel's own Intel Digital Home Capability Assesment Tool fails three of the four Intel designed tests, while strangely the i955 does real well. I won't dignify this one with an explanation, but I did just replace my i840XE gaming box with an FX-60 earlier this week.
Intel goes on to show how an 920 CPU with a i945G chipset handily trounces both a 520/i915G and a 631/i945, but there is no AMD listed there. Anyone guess why? Then on to listing 'smoothness' when working on Intel, and a performance advantage to boot, naturally on everything. The pinnacle of all this is a five per cent win when running Razor Lame and Norton AV and Trillian and Firefox and iTunes and Acrobat and Excel while importing a 355MB Outlook file. Sounds like my average workload, how about you. Bad Intel, no cookie.
Part three goes on to show the Intel 'Platform Ecosystem Advantage', and this is the one place where Intel is better for the most part. It features the aforementioned Matrix Storage, XD bit and iAMT. The Intel equivalents are more broadly supported, have you ever seen anything supporting AMD's ASF 2.0? How about its standard image platform? Intel closes the gap here 1:2, but cheapens it with unnecessary potshots.
P22 is all about Viiv versus AMD Live! but since Live! is more or less a slide presentation right now, not much can be said here. Oh wait, Viiv is a slide presentation still too, along with a bunch of logos, and to top it off, Live! has no DRM infection, and does not cut Linux out of the market, so far anyway. Another tie.
One of the most egregious slides is the socket comparison, and it quoted the INQUIRER on this one. It somehow tries to show that the 754, 939 and 940 sockets are a problem and confuse the market. In the same time however, Intel has had 423/478/479/480/775/603/604/771 and probably others I am forgetting. Toss in that most chipsets are not compatible with all CPUs that physically fit in Intel sockets, and the advantage is clearly AMD here. 1:3.
Intel closes the section on a low note, making vague FUD references to AMD platform reliability. They seem to think that an all Intel top to bottom solution is better than using third party chipsets, but conveniently ignore the fact it is using ATI chipsets on Intel mobos. Oops.
Part Four is all about comparing CPUs. This puts Intel and AMD CPUs head to head on SYSmark 2004 SE and PCMark 05. From this it concludes that a 2.8GHz PD (French slang for nonce mind you) ties an X2 3800 in one and beats it in the other test. It gets funnier though, the not-on-the-roadmaps 805 @ 2.66GHz does not have a counterpart on SYSmark, but 'beats' an X2 3800 in PCMark. If you buy this one for a second, I have a bridge to sell you. The retail market tell a story that directly contradicts this page. 1:4 AMD. The last part, Conclusions, ties this happy little world together. Intel wins everything, is better in every way and has no weaknesses. Sad, and Intel should not stoop to this level. One look at the retail segment shows why Intel has to resort to low blows, but really should be ashamed at sinking this low.
The INQuirer
Linear Mode
