Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Mozilla dlls
Yes I'm still using 1.7.5 and they are part of the quality reporting system. They were bumping other dlls probably because they were being loaded early in the boot up phase - to check for errors.
Did you check to see if any other dlls are being relocated?
I use McAfee anti-virus and their developers seem to know nothing about where to locate their dlls so there are a lot of conflicts. Many can be rebased but there are a core group that McAfee checks on startup and complains if they have been touched so you're left with no option - by default McAfee sets all their dlls at 0x10000000!
Many Mozilla plugins also try to load at default.
Snappier Firefox
Mmoy, take care that the components of firfox are based properly. I use Mozilla and had to rebase a couple of the dlls so they weren't relocated and also to prevent them shoving some unrebaseable system dlls into the pagefile (2 of them I moved were fullsoft.dll and qfaservices.dll, both had been left at 0x10000000).
After moving these around in virtual I found the difference in performance quite obvious.
Multi processes under Windows
I keep seeing all these references to multi-threading being necessary to get the most out of dual core but surely this misses the point of the way Windows works.
Let's take a browser for example. Even when its in the foreground it has to call services to actually get the data stream. Surely, in a dual processor system, those services can run on the 2nd core. When I look at my system just after startup there are > 20 processes running. One of the heaviest uses is antivirus and a lot of systems have that.
So, IMHO, except for some unique situations like games I see that a multicore makes lots of sense. No?
Throttling?
If its o/c and running lots of SSE code it may well be thermal throttling that causing the issue.
If its malfunctioning because its being run so far out of spec I would expect a blue screen rather than just slowing.
Has he blown out the chip fans recently?
However in that famous Stanford talk the Intel engineer did allude to unexplained states in the P4 where it would just go 000s of cycles without doing anything useful and the designers did not know why.
There may be some (confidential) Intel notes on specific combinations of instructions that can cause recovery logic to set in.
Customers demand Xeon
Yup that's what happens when you don't put any effort behind the brand.
Brand money comes from product pricing, just like I said in my last post.
Athlon64 prices are too low.
I agree. AMD's marketing is as technically challenged as their R&D is advanced.
Every management game I have ever played hammered in the idea that you do NOT sell your product too cheap. Its OK to drop prices for a short time to grab market share, like the japanese car industry, but longer term you must get those prices up to finance the corporate and product brand.
Over the years I have followed this company I have concluded that AMD has no clue how to sell CPUs. It does a good job selling flash because that's not a consumer product. AMD has to realize that it IS a consumer product company and act like one.
Put UP the prices and give discount coupons but make sure that P4s are cheaper in list price than Athlon64s else you have an uphill battle to convince buyers, corporate as well as private, that your product is worth anything. Right now even I question the value of the Athlon64 - WHY IS IT SO CHEAP? WHT IS THE SEMPRON NOT PRICED AGAINST THE P4.
I think its time shareholders addressed this issue and threatened to withhold votes until we get some answers.
Excellent post chipdesigner.
Right to the point, well done.
NSG & SSG at Sun
I would think that Newton is the manager, probably a VP while Bechtolsheim is a technologist, maybe CTO. I think I asked that question and CTO was the answer. It was a full evening of conversation including several beers at the Rose & Crown.
For me the important thing was realizing why there is this schizophrenia between sparc and opteron.
I don't see Bechtosheim reporting to Newton.
Galaxy is ready now.
See my earlier post on why they are not officially lauching it until late this year.
Well what I am reporting is not sales hype, the guy who told me isn't in sales and is well connected in Sun in the SF main offices.
".... bullish on Galaxy, business is good"
Of course that doesn't mean that Galaxy business is good but rather that their overall business is good - tho' Scott McNealy did say recently that he didn't expect the stock price to move much before the end of year.
Sun's business model
Actually Sun's business model for their major customers is not based on selling hardware. Sun will price up a solution either on a per seat basis or a per employee basis. It doesn't matter whether they sell you x86-64 or sparc from a sales price point of view.
Their channel sellers do sell hardware but to support the software stack so the actual item sold is more often software, with supporting hardware.
Waiting for Galaxy - why the delay?
In a word validation. I gathered that they just need months to fully validate the systems - like having a baby you just can't rush it.
Sun info
Had dinner with senior Sun guy last week. Here's what I learned:
There are 2 hardware groups: SSG (scalable systems group) & NSG (network services group)
SSG is sparc & niagara
NSG is opteron, galaxy headed up by Newton
Galaxy is ready now, suggestion that Galaxy is being used for offsite installations (i.e where the customer doesn't have the hardware onsite and the whole shebang is operated by Sun)
Galaxy is awesome and "will change the server paradigm"
Newton will give away some Galaxy units to seed the market before the market launch late this year.
Suggested that SSG guys are probably polishing their resumes.
NSG is a lean smaller unit.
Scott is very bullish on Galaxy, business is very good. Sun see weakness in several competitors especially HP & Dell.
Lots of other stuff, what do you want to know?
This is pathetic.
If you don't understand even the basics of physical vs virtual I suggest you go invest somewhere else. Toys R Us perhaps?
Isn't the assembler in the SDK?
I'm sure you've seen this but just in case:
"Building 64-Bit AMD64 Applications With Visual Studio"
http://www.devx.com/amd/Article/21313
Anandtech's SSE3 test
Anandtech did its test on old xp32, not Windows XP64. The software was then 32-bit and only had 8 SSE regs to work with.
I've profiled SSE sequences and its clear that its register limitations that stall execution. There may be a small improvment due to less instructions using some SSE3 sequences, but a profile will shows that most of the time the instruction can't be retired because its waiting on some earlier result in one of the 8 registers.
This may not be the case with 64-bit code since there are 16 rather than 8 SSE registers.
The results from Anandtech show nothing about 64-bit code. Now that will be interesting.
E0 steppings, interesting post
This at Anandtech, comment posted by someone who sounds like they know:
"I don't think the BIOS of the test machine was adapted to Revision E Opterons.
I adapted LinuxBIOS to the Rev E stepping last week and the 1 GHz support is really the easiest thing (was already present in revision D processors). Changing the HT speed while the operating system is running is _very_ difficult. It requires a reset or LDTSTOP on both CPUs for the new frequency to be effective, so this is normally done a boot time in the BIOS. I guess ntune does not really change the HT frequency.
In addition Revision E has a number of errata fixed which result in improved performance (for example Errata 94).
The most important point is the new memory controller mode that reduces the DRAM bank conflicts. It improves STREAM benchmarks scores around 30%. This modes has to be automatically enabled by the BIOS, so please rerun the benchmark on a mainboard that supports Rev E processors."
Time to buy
I was just waiting for the E-stepping
Makes for a good weekend.
Acer doing well (they sell AMD)
http://news.com.com/Acer+pulls+ahead+in+global+PC+race/2100-1047_3-5571402.html?tag=nefd.top
Sun EOLs Intel kit
http://www.theregister.co.uk/2005/02/10/sun_kills_intel/
Love the URL
Is that in 4Q2004 dollars?
I know that I'd much rather have 4Q2003 dollars.
Before this little-boots was elected I could actually fill my car up with gas for $20.
Looking at GLD
CPU usage - Firefox
Just to give you an idea, on a 450MHz w/ dialup connection at 49.2Kb/sec I pulled up Santa Barbara on Google maps.
The CPU usage was avg 5% and never more than 10%
Also running Outlook 2000 and Word 2000 plus mcafee firewall & antivirus oh and Java 1.4.2 (and some other stuff)
When I scroll the map (at 1280x1028) using the mouse I do manage to hit 100% CPU usage at spikes though its averaging about 70%.
Those K6-3s are amazing.
Yes, dual core is what's needed for HD Video. Hexus.net had a review of (trying) to edit HD and demonstrated that dual Xeons was >> single Xeon. Still it was not great.
HD editing under AMD64 is a natural (just map the video files into virtual) and dual cores should make it workable. You may need 2x2 tho'
"Why does AMD need something faster than their current chips?"
Editing HD video (720p, 1080i, 1080p) we need it NOW. Even the fastest current setups choke on this. I'm hoping someone will hang a DSP off the HTT so we can encode/decode w/o the CPU.
Firefox disk writes
I think you just indicated the answer when you talked about cache objects being partially written to disk.
Do you know if firefox opens any of its cache-related files as Non-buffered? Looks to me as though it might do, just to reduce problems. That's fair - you need the most robust code possible and there's lots of people who do dumb things.
Yup I know about about:config. I have a couple of changes in there, mainly so that Outlook can handle the mailto's. Its not paging that's the issue - I can monitor that fine. The program is disk bound for me on the 5 machines I use.
Firefox MOOX builds
Well I generally use Mozilla. Its the browser installed on my main office machine which is a lowly 450MHz. I installed a MOOX firefox build on my home XP2600.
I noted no difference between the 2600 and the 450. In fact running the CPU% performance monitor on the 450 it is pretty rare that CPU usage ever gets anywhere near 70%.
I do have a nice disk setup on the 450.
So firefox, as I see it, is not bound by the CPU. Rather its limited by disk access (especially & suprisingly - disk writes). What's the point of optimizing the builds for flavors of ISA then?
Damning review of Pentium M "Dothan" as desktop chip
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2342
Its rendering
The Athlon64 seems to be doing rather well in video rendering. The cost of an animator/illustrator is so much more than an FX chip that it only makes sense to buy the fastest kit.
I think they are being snapped up by industry and few are left for gamers.
BTW if you watch Sky Captain & the World of Tomorrow on DVD there's a bit about rendering/hardware/software costs in the addional features - not a big analysis just them talking a bit about rendering times and how it was an issue. Pretty good little movie too, just like the Saturday matinees.
"The P4 designers used the strategy of making the common
case fast."
Duh. And I suppose all the other processor engineers optimized ARPL?
EVERYONE optimizes for the common case.
Shifts & rotates are NOT that uncommon. The P4 suffers badly as a consequence of its design.
The P4 engineers screwed up and now their product is history. The fact is that they wanted to optimize shift/rotates but the Netburst core forced them not to do it. Just a poor design.
AMD and Microsoft Partner to Move Enterprise Customers to 64-Bit Computing
I believe this to be the most important news since Sun adopted Opteron.
I will sleep with a smile.
Twas me!
And the results are?
And you should insert asm{} etc in the old code to direct & see what's really going on - hmmph, who trusts compilers!
Wireless driver for XP64
According to the newsgroup posts, Win Xp64 v.1289 includes a Broadcom wireless driver.
"Belkin uses Broadcom chips..so does Linksys. Belkin
should work correctly with the in-the-box (1289) Broadcom
wireless driver.
I have A Linksys wireless adapter that works perfectly
with the in-the-box Broadcom driver" - Microsoft's XP64 newsgroup, yesterday
New anti-virus engine for AMD64
According to the notes, McAfee's new engine for its widely used anti-virus product now supports Win XP64 for AMD64.
http://a64.g.akamai.net/7/64/2015/2004-11-03-6-13-51-300/download.nai.com/products/licensed/superdat...
I hadn't seen this, maybe others knew.
Thinking about rgb->bgr
the best way is probably the simplest:
;esi ->src & dst
;ecx -> # triplets
;*************************************************
top:
mov al,[esi+0]
xchg al,[esi+2]
mov [esi+0],al
add si,3
dec ecx
jnz top
Only problem is that xchg locks the bus, so you may wish to do:
top:
mov al,[esi+0]
mov ah,[esi+2]
mov [esi+2],al
mov [esi+0],ah
add si,3
dec ecx
jnz top
Isn't that easier? And write allocation takes care of the memory accesses for you.
rgb -> bgr
Here's my first attempt, x32
;assume esi points to source
;output bgr's in ebx,edi,eax,edx
;x=don't care
;**********************************************
mov ebx,[esi+0] ;r0,g0,b0,r1
mov eax,[esi+4] ;g1,b1,r2,g2
mov edx,[esi+8] ;b2,r3,g3,b3
mov edi,eax
shld eax,edx,8 ;b1,r2,g2,b2
shrd edi,ebx,16 ;b0,r1,g1,b1
shr ebx,8 ;x,r0,g0,b0
bswap eax ;b2,g2,r2,x
bswap edi ;b1,g1,r1,x
bswap edx ;b3,g3,r3,x
bswap ebx ;b0,g0,r0,x
I'll leave the boundary conditions to you!
HT 2.0 Performance
New article.
http://www.devx.com/amd/Article/27012
And some details on Sun's Honeycomb/Opteron system
http://news.com.com/Sun+hopes+for+better+storage+with+Honeycomb/2100-1015_3-5553913.html?tag=nefd.to...
Solaris 10 blog
http://blogs.sun.com/roller/page/bmc/20050125#solaris_10_revealed
Makes for some interesting reading and mentions Opteron
Are you ignoring compression?
I'd have to dig a bit to give you a complete answer but I do know that our analog cable company squeezes more than 100 analog channels down their coax and uses heavy compression to do it. DTV signals are already compressed and while they can be compressed a little (which the companies do) its nowhere like analog.
Our TV allows > 100 channels on analog CATV.
I suspect that the bandwidth is less than 600MHz throught the network and there are unusable parts (joins and distribution boxes trim the available spectrum).
I see you're right about cable HD - they use 16-VSB rather than the 8-VSB used for OTA and thus can get 2xHD in one band on digital cable. I hadn't realized this.
Well the choice of fiber seems redundant for this purpose doesn't it.
Thanks - I learned something!
OTA HD content.
I don't know what town you are in but here on the edge of LA County we get 20+ HD channels with most of them multicast.
The major channels: CBS, NBC, ABC, Fox, PBS (several) are all in HD. Only UPN is broadcasting in digital that is not HD (480i). We also get some of the new "local" DTV channels, like LATV
You may want to check www.checkhd.com to see what is available in your area. Note that their listings understate the content available as they don't list all the HD programs, uncheck the "HD only" box at the top of the listing.
That's good news that they have done fiber to the door. Here is would be outrageously expensive as its quite rural. I hope that WiMax might be a solution. At least fiber would cut down on the interference we see - on analog - from leaky coax.
Less channels on HDTV cable
We've had HDTV since Jan 2004. We don't have cable - we get all the channels we need over a 12-year old antenna, most of them in picture-perfect 720p or 1080i. Watching West Wing on 1080i is a real pleasure! We never watch on analog any more. We use a 32" Zenith integrated HDTV set, $899 at Best Buy.
Problem with HDTV cable is that there's only so much bandwidth and cable companies have sold the # of channels as a figure of merit. So if you go to HD content you get less channels. I can't see that customers are going to be all that happy at paying more for perceived less, even if the channels they lose are crap content.
You can see the ridiculous level of compression that the cable companies are using now. The quality of the picture is awful, even on analog. Over the air (OTA) is much, much better.
My friend is thinking about going HD cable for one reason - Playboy porn on HD! I think he will go to Zoom. I mean - he wants warts and all.
I would like a PVR but it has to be HD and I don't want to end up with an out-of-date system, so much is yet to come in HD - especially the HD DVD format. I think I'll set up a dedicated small box for doing it and I'll use an Athlon64 for the purpose. I think I can do it with a DVI out for the picture and use the Picture-In-Picture to handle the system functions.