SuperMicro gets some more courage
From selling their motherboards under Monarch brand, to publishing their own press release on their Opteron boards for Google, now, SuperMicro puts Opteron solutions on their front page.
In some sense, Intel and AMD are treated equally on SuperMicro home page. Neither Intel nor AMD appears there, only product names such as Xeon and Opteron. But there are more Intel code names (such as Blackford, Glenwood, Mukturd...), as Intel dudes are very productive in making new names -- unfortunately, naming seems to be the only thing they really excel at.
posted by Sharikou, Ph. D. @ 8/31/2006 07:23:00 AM
16 Comments:
Must be a slow anti-Intel news week if this is the best blog you could come up with after your CPU exploding fiasco blog from Monday.
On a related note, there is now an AM@2 motherboard ad with AMD64 logo on tghe Anandtech mainpage - rather remarkable for a "paid Intel pumper".
Intel probably asked SuperMicro to put some AMD stuff on the homepage.
This is because of the antitrust lawsuit.
SuperMicro has had Opteron products on the "/aplus" page for a long time.
Though it is Tyan that went early to Opteron and has profited greatly. Even though the early boards were fraught with issues, due to both Tyan and AMD, Tyan deserves some credit for being there and trying to make the early AMD stuff work.
Quote from GamePC.com
"The X6800 beats AMD’s top of the line FX-62 processor in every CPU benchmark, even when the FX-62 is running at overclocked speeds. In order to truly compete with the X6800, AMD would have to put out dual-core chips in the 3.6 – 4.0 GHz range, as Intel’s performance per clock is quite a lot better compared to the Athlon64 X2 product line. With AMD’s current manufacturing process technology, these kinds of clock speeds aren’t going to come about any time soon, which means Intel will likely hold the performance crown for the foreseeable future."
My only question now... how will Sharifraud spin this bad news for AMD into a diatribe against Intel and it's "stupid management" and "incompetent engineers"? If they can wipe the floow with AMD like this imagine what Intel could do if they weren't handicapped with a bunch of "morons" as Sharifraud would have us all believe.
Alright they have the faster processor, but come on the memory that they use is just stupid. "CAS 5,5,5-15 Latency at DDR2-800 Speeds with 2T Timing" everyone knows that that hurts the fx or x2 performance.
So yeah the 6800 is still faster, but not with the margin they tell everyone.
Game pc just doesn't know how to execute a good benchmarkingset.
Quote from GamePC.com
Even handicapped with 5-5-5-12 RAM, FX62 is only slightly slower than Con XE 6800 in most tests (except SuperPi). In Cinebench 9.5, Con XE 6800 is 16.6% faster. In 7-zip, Con XE is 21.7% faster. In WinRAR, Con XE 6800 is 8% faster. In Windows Media Encoder, Con XE 6800 is 12% faster. In HL2, XE 6800 is 15.9% faster. In FEAR, XE 6800 is 10% faster.
This is consistent with our previous analysis that Con XE6800 (2.93GHZ) is about 10% faster than FX62 (2.8GHZ) when using 4-4-4-12 RAM.
This also indicates once Rev G is out, AMD will have a 50% performance lead over Conroe on integer and 200% lead over Conroe on FP.
Intel's BK is inevitable.
Some anonymous coward wrote:
Alright they have the faster processor, but come on the memory that they use is just stupid. "CAS 5,5,5-15 Latency at DDR2-800 Speeds with 2T Timing" everyone knows that that hurts the fx or x2 performance.
When will AMD fanbois stop with the CAS latency meme? They bitched about it when Anandtech benchmarked Conroes after Spring IDF and then when AMD fanbois bitched that they cooked the benchmarks to be unfair to AMD and when they made the fixes the fanbois wanted, guess what? There was almost no difference.
I can already predict responses to this post.... "YEah, but Anandtech is paid Intel pumper, blah, blah, blah". Guess what, GamePC.com sells more AMD systems than Intel so they are a paid AMD pumper but since their BUSINESS depends on selling to the performance market, they don't care who is faster. They just sell whatever is faster. So now they will likely start selling more Intel systems. That's just how it goes. AMD was on top for a while, now there back down where they belong.
Sharikou also says the 6800 is only slightly faster. I think a 20% improvement is pretty good at the top end and the E6700 is also faster than FX-62 overclocked and costs a lot less. You can pin your hopes on rev. G but I would be surprised if AMD comes anywhere close to 100% faster integer and 200% faster FP. Not to mention, Intel has a lot of headroom on the Core 2 Duo chips as evidenced by their amazing overclocking capability. This means the manufacturing process for Core 2 Duo is pretty damn good and they are binning the CPUs for future growth to counter such moves by AMD.
Intel Layoffs soon.
"stupid management" and "incompetent engineers"?
Does smart management have to do belated evaluations of the whole company and plan to overhaul its business and need the rumored massive layoff?
Do competent engineers keep following the messy Netburst roadmap only to end up with product deadends, detours, U-turns, Y-turns?
Do competent engineers get ridiculed by its counterparts in Intel Israel for riding the "Donkey" till it hit the wall?
All the "stupid management" and "incomptent engineers" are in the Intel U.S operation.
Just watch. None of the layoff will be from the good old boys network. People who's made wrong decisions will still have jobs at Intel.
Sharikou said...
"This also indicates once Rev G is out, AMD will have a 50% performance lead over Conroe on integer and 200% lead over Conroe on FP."
I am not saying anything about "K8L", but wouldn't it make sense that Rev G is the shrink to 65nm, and H is the quadcore?
I ask because Rev F was for DDRII, right?
So, Sharikou, in other words, X6800 'frags' FX-62?
I bet you wouldn't utter those words with a gun to your head, LOL!
Must be a slow anti-Intel news week if this is the best blog you could come up with after your CPU exploding fiasco blog from Monday.
I got one.
Intel Mac Randomly Shutdown
Wow, I just saw the inventory of the Athlon X2 3800+ that was mentioned in Sharikou's log on Aug. 3th drop by about 11,000 units in 1 day; yesterday the counter stood at 12,000 now at 1,686, either the counter is wrong, or someone is taking big bites...
(http://www.shopblt.com/cgi-bin/shop/shop.cgi?ACTION=thispage&thispage=011003000501_BJ70871P.shtml&ORDER_ID=354317267)
What I find interesting is that FB-DIMM is just about the same price as plain DDR2.
So while FB-DIMM does use a bit more power, there is no meaningful price penalty (about $20/DIMM for 2GB ECC 667Mhz DIMMs).
The only exception to this is everyone's favorite dominatrix darling, Apple, whose mega-heatsink FB-DIMMs run way above the cost of normal FB-DIMMs. Last I checked, it was a $200 premium for an Apple-compatible heatsink! Ah, noting compares to the sweet green cash flow of Apple bondage.
All of this is to say that Intel is in a good position on the server side for the next 2-3 years with FB-DIMM.
It is far easier to support large memory configurations with FB-DIMM vs. DDR2 due to the various issues with "ranked" memory in DDR1/DDR2 land.
All in all, it looks like Intel made a good choice going with FB-DIMM and AMD made the wrong choice going with plain DDR2.
An FX62 with 3-3-3-9 1T is within 5% of the X6800. Bump the FX clock to 3.2Ghz and the FX actually wins the majority of the benchmarks. The 65nm process will bring this possibility just in time when the Conroe is available - I mean REALLY available.
AMD is not too worried and they have good reasons not to be. Many will say that 3-3-3-9 RAM is
rare and expensive - it is - but it is getting cheaper and more available every day and at a better rate than the trickling Conroes. The Conroe is a great chip but a little too little a little too late.
Anonymous said...
"Bump the FX clock to 3.2Ghz and the FX actually wins the majority of the benchmarks. The 65nm process will bring this possibility just in time..."
The problems with your statement is that the 65nm will be limited to 2.4GHz (initially), and FX will stay at 90nm.
So what kind of TDP will there be at 3.2GHz at 90nm?
Can they even reach 3.2GHz on air?
Anonymous said:"It is far easier to support large memory configurations with FB-DIMM vs. DDR2 due to the various issues with "ranked" memory in DDR1/DDR2 land."
That's the thing. With integrated memory controllers, each socket drives its own bank of memory.
This
a. doubles the available memory bandwidth and
b. doubles up the max memory "ranks" of DDR2, without resorting to higher latency, slower FBDIMMs.
Win-win situation for all.
Intel only has one memory controller, located outside the CPU, separated by the much slower FSB, so has to resort to FB-DIMMs to support the same memory configurations.
Just as an example, HP's DL380G4 (pre Woodcrest) only supported 12GB max memory DDR2
http://h10010.www1.hp.com/wwpc/us/en/ss/WF06a/15351-241434-241475-241475-f79-397634.html
HP's DL385 using DDR (which is less stackable than DDR2) could support 32GB!!
http://h10010.www1.hp.com/wwpc/us/en/ss/WF06a/15351-241434-241475-241475-f79-428936.html
With FB-DIMMs, the DL380 with Woodcrest can finally support 32GB
http://h10010.www1.hp.com/wwpc/us/en/ss/WF06a/15351-241434-241475-241475-f79-1121516.html
But by then, the Opterons with DDR2 would be able to support far higher memory configurations (128GB if I recall correctly).
Post a Comment
<< Home