SM will have boards with pci and PCI-X..;)Quote:
Originally Posted by chuckbam
Printable View
SM will have boards with pci and PCI-X..;)Quote:
Originally Posted by chuckbam
I asked Dr.Who? ( Francois Piednoel from Intel) that same question and his response was as said above here:Quote:
Originally Posted by highoctane
"The expiration of the NDA is under NDA"
Then laughed. Good to see these people with a sense of humor.
My guess is that they are still refining these and don't want any 'official" numbers released until they are happy with the results.
Also possible that they still may have small issues to deal with.
The MB manufacturers seemed to be tossing fits over the tri channel memory and making it work right.
Then also the still up in the air issues on memory standardisation on these boards..DDR3-1066,1333?????
would be nice to hear more today, well at least we can hope. ;)Quote:
Originally Posted by Donnie27
http://forum.desktopreview.com/showthread.php?p=2256457Quote:
Originally Posted by Hornet331
We're getting close when pics start to show up like this. We know the server folks will get them first. Like Movieman said about the memory issues. Servers don't have to run the RAM balls-to-the-wall. Geesh, wouldn't 3 channels of 1066 be enough for now?
At that link they show the older 4 core chip for the Tigerton board(quad core,quad socket, older 604 pinned socket) but they are talking about the new 6 core chip that is supposed to be a drop in replacement for that.Quote:
Originally Posted by Donnie27
This is the one I was talking about that I may get to review..
24 cores on one board..:rofl:
:eek: you will be able to review the Intel's 6-core Dunnington I am imagining you counting the days till it comes to your door step :yepp:
i wonder how dunnington performs with the 3rd lvl cache, since its shared between the 6 cores, so the quad fsb only is there for inter socket communication. :)
I misunderstood, I thought you said compile time wasn't important. And yes, i do know the difference.Quote:
Originally Posted by gosh
What does that have to do with the Nehalem Launch?Quote:
Originally Posted by JumpingJack
Nothing definate yet but when I inquired on doing a review on a Gainstown(dual socket Nehalem) they asked if I would be interested in doing a review on a 24 core Dunnington board first.Quote:
Originally Posted by @@@@
At first I said no as I was thinking I'd have to provide the cpu's,memory,etc..
That's close to $11,000.00...:rofl:
so my question is. What did you do to convince them that your part of the press?Quote:
Originally Posted by Movieman
I didn't. I told them I was a moderator on XS and a devouted "dualie" person and wanted to show all these crazed gamers that there were options out there that they might want to look at.Quote:
Originally Posted by Blauhung
Now the fact that when I posted the review it was grabbed by a ton of sites all over Europe didn't exactly hurt either.:D
Nehalem and its plafform is a very complex change, we are busy making sure it is rock solid. More than 800Millions transistors, new Bus interface, new IO subsystem, performance validation and functional validation are exponentially complexe. It does increase as a permutation of combination.Quote:
Originally Posted by Movieman
People rarely understand the scale of validating a processor, it is looking for a niddle in 800millions parts ... with a motherboard team, a CPU team, a BUS team, a performance team, a Quality team, and several version of the CPU, for mobile, desktop, server multi processor ...
I am lucky enough to live this from the inside, right now, there are thousands of people busy, working like ants, and I admire the General manager for making all of this happen in parallele. Attention to details is critical.
We are paranoid, as Andy Groove taught us, so, even if it looks ready, we got to have the CPU and motherboard good enough to face the hightest requirements. This is almost an impossible task that we are after, we need time for this. We are workig hard on it, very hard, week end and nights are too short :) :rofl:
aka, nehalem isn't ready.
It's very complex so it's understandable.Intel can now see how hard it is to validate such a complex chip (and still make a mistake in the process->Barcelona launch and TLB issues).Deneb is having a bit over 700mil. transistors,although the platform is already validated(AM2+),hence easier time for competitor from that pov at the moment.
Intel is having a tough job as DrWho said,with all the platform changes that were done with Nehalem.But it's a huge company with a lot of awesome engineers,they'll probably make it on time.
Hehehe! Yes but that is a big sum-bee-otch!Quote:
Originally Posted by Movieman
aka, we are QAing it. no need to spin it negatively, it is a normal process, nothing wrong. It is ready when we launch it, not before.Quote:
Originally Posted by PetNorth
Its normal. I´m sure things like the Itanium gets even more validated than a K10 ever was or Nehalem.Quote:
Originally Posted by informal
And its all on time.
Not sure what you try to spin out of it. That ohh its so hard to validate now? Along with its so hard AMD failed on K10? :rofl:
Shintai,what part of the above is hard for you to grasp?:rolleyes:Quote:
More than 800Millions transistors, new Bus interface, new IO subsystem, performance validation and functional validation are exponentially complexe. It does increase as a permutation of combination.
Montecito took 2 years to validate after tape out; the biggest problem was Foxton which was eventually killed.Btw , Itaniums are tested in a heavy radiation environment at Los Alamos were it was exposed to neutron beams.Quote:
Originally Posted by Shintai
Page 22 http://h20427.www2.hp.com/event/kr/k...0solutions.pdf
Well having paid those long nights myself on our own projects, I know there has to be a ton of work going into this final stage. And we all know that it only takes one bad part deep in the chip to spoil our fun too. So patience is a good thing for both sides in this process.
I think sometimes it's like watching a fancy car come off the line. You are itching to drive it, but you have to let QA and all the checks go on first, so you know it's really GTG when you get the keys. I feel like it won't be a lot longer, but I also know I'm appreciative of the thoroughness that Intel puts into getting the chip right. Isn't that the very reason we love Intel chipsets for example? Because they work, solidly, the way we like.
I7 is coming. Just needs a bit more time for the finishing labor and touches and then we'll be enjoying it. :)
Absolutely nothing.Quote:
Originally Posted by chuckbam
Sure and of most us are glad you guys don't have to rush out a product with TLB bugs and etc... Then be desperate enough to sell and not at least offer an exchange. Well almost, see below?*Quote:
Originally Posted by Drwho?
The good!
To the other guys, give Intel props for choosing something doable in the first place. I remember certain folks here at XtremeSystems mocking Intel, making the wrong choice by going with MCM they said. Then went on about how Native Dual Core would crush them. At least admit Intel was right? It's not like Conroe was easy. Give Intel credit for not making the same mistake AMD did.
The bad!
Why I'll still just watch the Nehalem launch without buying one.
*My Wolfdale C0 is a fine processor but seems broke and parts missing compared to the E0 that has more features, runs cooler (I'm talking stock, not overclocking) and has less errata. Sure its faster than Conroe and even my bud's A64 X2 at 3.1GHz, but it is slower, hotter, and etc... than the newer E8400 E0 he just replaced that AA64@3.1GHz with.
I'll wait for a couple of steppings before I touch Nehalem! I honestly feel I didn't get the best deal with this Wolfdale! Don't want to make that mistake again. It's not a TLB bug but IMHO, it ain't far from it:rolleyes:
I am going to hold for a while. I don't know if I need the fancy retail board. DDR3 tri kits should come down. I want a second Video to Cross-fire. And, a TB Seagate. I am thinking like next April. I do want the Bloomfield w/ the tri channel and 12 GB of mem.
Do the vanilla Asus P35 come with an OC bios?
Yes!Quote:
Originally Posted by chuckbam
Well if one comes out like that, with a good phase Power, I will get an vanilla i7 board. One problem, I need the Intel Raid.Quote:
Originally Posted by Donnie27
Of course. Some boards have better support than others.Quote:
Originally Posted by chuckbam
Note I'll believe a Budget or even Mainstream board for Bloomfield when I see it. Anything coming from or meant for Intel with the letter "X" means eXpensive.;) X58 is more than likely just that. But everything I've heard or been told is that it will be worth every cent.:up:
Well if I need to buy the box an all the cables, I will. I think you are 1000% about no bland boards at first. But, I don't know if that will be true after everything settles down. That is why I am thinking April or even out to July. Also, I have jumped early many times before, just to buy rev. 1.1 a few months later :rolleyes: I only want to build my new system one time next year.Quote:
Originally Posted by Donnie27
God, please give me the strength to wait a bit.
I'm thinking more like Aug or Sept 09. Performance gains will have to be pretty dramatic for me to jump the gun.Quote:
Originally Posted by chuckbam
hehe obviously. And I feel that it is later than Intel thought. Some guy filtered a september launch, but this thing isn't ready yet (IMO due to some problem not expected).Quote:
Originally Posted by Drwho?
Who, fuad? :p:Quote:
Originally Posted by PetNorth
September would be Q3. And its always been a Q4 product.
Have you seen theis chinese test already? http://diy.pconline.com.cn/cpu/revie...426985_10.html
The numbers are quite a mixed bag it seems, I hope I got the (quick and dirty) clock for clock comparison with penryn more or less right:
SuperPI MOD 1.4 --- 19,3%
WinRar 3.71 --- 77,3%
Fritz Chess --- 29%
Cinebench R10, singlethread --- 6,7%
Cinebench R10, 4CPU --- 25,7%
TMPGEnc 4.5 --- (-8%)
3dMark06, CPU --- 13.9%
Vantage, CPU --- 50.5%
Call of Duty DX9 --- (-8,6%)
Half Life 2: EP2 --- 2.3%
Wold in Conflict --- 6.6%
Company of Heroes --- (-7.2%)
Devil May Cry --- 5.1%
Nope,not posted before afaik.How come Far East guys always get the best toys first? :)
Is there any mention of the Turbo mode and whether it was active during their test?If it wasn't on during testing,according to your numbers it's on average 16% faster per clock,very good.
I have 2 of your clovertown systems here and 2 harpertowns.Quote:
Originally Posted by Drwho?
Both have exceeded expectations and the two clovertowns have been running at 100% load since Jan 2007 with an uptime exceeding 98% at over 3000mhz.
As good as the clovers are, the Harpertowns almost make them look slow.
You take your time.
I'd rather have it "right" than fast.:up:
I guarantee you that I'll have a dual 3200 gainstown here as soon as the chips and boards are ready.
Can't wait to see what that will do in our Aids and cancer research!:up:
That's no good.Quote:
Originally Posted by Jacky
Check the revision: Stepping 2, Rev B0.
Big improvements since then.
really? like what? I know this design isn't really meant to increase gaming performance but those numbers aren't that great, especially if the 2.66ghz version doesn't oc as well as a Q9550.Quote:
Originally Posted by terrace215
Nehalem has definitely been over-hyped by enthusiasts. Although, I have the feeling the average consumer will first be disappointed with a) single-threaded and general gaming performance, and then b) the expense of the entire platform. I also still have reservations about the overclockability of anything other than Extreme processors. Any delay in shipping dates certainly won't help quell this possible disappointment.
Although, I respect any company that doesn't rush buggy incomplete products to the market. Good things come to those who wait.
Have to agree!:yepp:Quote:
Originally Posted by User9498
what delay in shiping dates? is Q408 already over... damn timewarps... :p:Quote:
Originally Posted by User9498
Geeks over hype everything though. In their zeal and excitement, they/we always drool over new stuff! Look at the Barcelona/Phenom Hype last year?Quote:
Originally Posted by User9498
Reservations? Hell, folks like Jumpingjack, savantu and Hornet133 gave excellent posts on why these folks should stop over hyping Barcelona, didn't help there LOL! You can add Video Cards and even Newer Raptor Hard drives as well. Then, don't even get me started about 3D LCD's:D
@Informal
16% is too simple. That average hides too much. I'd not even bother wasting time on 3D Marks other than something to demo on a computer I'm trying to sale.
Call of Duty DX9 --- (-8,6%)
Half Life 2: EP2 --- 2.3%
Wold in Conflict --- 6.6%
Company of Heroes --- (-7.2%)
Devil May Cry --- 5.1%
That's just short of 6% on those games. If you're a gamer that's not that impressive at all. But if games are an after thought and those other apps are you main use, those same games and useless 3D Marks is hiding some pretty awesome performance improvements.
SuperPI MOD 1.4 --- 19,3%
WinRar 3.71 --- 77,3%
Fritz Chess --- 29%
Cinebench R10, singlethread --- 6,7%
Cinebench R10, 4CPU --- 25,7%
TMPGEnc 4.5 --- (-8%)
25% on these 6 and 31.66% on those in bold! Big time improvement on an apps AMD has always given Intel a whipping on and even after Core, gave Intel a run for the money. Never mind that it is an older steppings that better not show up on the open market.
To the other guy, good luck even finding a lot of single threaded apps. Most folks buying Quads are using them for Multi-threaded apps. Hell, even I don't use Quads. I'd love to see a 3.2GHz cheap Dual Core Nehalem rescued from the scrap heap right now, but I'm going Quad next year.
The B stepping had performance-impacting bugs, so analyzing these numbers is a waste of time.Quote:
Originally Posted by shiznit93
They already have C-steppings (likely to launch, per Blauhung), and are even working on a D-stepping for the inevitable follow-up.
With C-stepping sample results now starting to appear, it is odd that this site used an old sample, but there you go.
Have we seen any numbers of the new stepping? What's your source, did I miss it in this thread and what exactly was broken? AFAIK we have only performance numbers for B stepping or earlier, which are damn impressive for everything multithreaded anyway.Quote:
Originally Posted by terrace215
http://www.xtremesystems.org/Forums/...postcount=1288Quote:
Originally Posted by Jacky
Not sure exactly what was fubar'd in the earlier steppings.
Read through Dr. Who? 's posts, as well.
Are you referring to Everest read/write/copy scores from pconline test?The scores are 15.3/14.7/16.5 MB/s from 2.93Ghz B0 versus 17.1/13.8/19.2 MB/s from 3.2Ghz XE Cx.They are lower on B0 due to a 1066Mhz memory that was used by pconline instead of 1600Mhz DDR3 used by hardspell and by higher clock of the NB in the XE 3.2Ghz Nehalem model... 50% difference in clock of the memory and lower CPU/NB clocks are going to affect the memory scores for sure.Quote:
Originally Posted by terrace215
:shrug:
I think everybody should relax and stop speculating on old steppings, it is like taking a brand new prototype of Nissan GTR skyline, and test it with the version of 6 months before the release ... performance tuning did not happen yet on it! :rofl::rofl::ROTF:Quote:
Originally Posted by informal
Be patient :clap::clap::clap:
Well i still don't understand why pconline even used that B0 CPU to begin with...
I think any highly anticipated release of any product receives the potential for an inordinate amount of hype. My take is that Nehalem will show overall mixed results, in some cases the results will be stellar (highly multithreaded of course), in other cases mediocre or less impressive (single threaded circumstance). On the whole, it will probably we well received but there will be a good number of fanboys who will put down the product and will be 'unimpressed or disappointed' ... even if it quadrupled performance there will be a select group that will chant -- 'was expecting more, extremely disappointed' ... :)Quote:
Originally Posted by User9498
It is as predictable as predicting night and day.
Deneb will be the same way to and extent, however, AMD has been much lower key on Deneb/Shanghai than they were with Agena/Barcelona -- probably to deflate the hype factor.
On your last statement--- one big Amen!
its basically multithreading oriented, so all it needs to show is improvement over penryn quad cores.
i guess if cheaper n's dont clock very well, penryns will still sell more...the imc and monolithic core features are the same as amd phenoms. (feature-wise) so from that basic laypersons perspective..thats that, but phenoms will remain a fair bit cheaper i suppose, due to, presumeably, lower performance.
I like all us geeks cannot wait....:p:
In the areas where they wanted to improve on conroe/penryn like multi-threading, memory bandwidth and latency, smt, qpi, they appear to have done a great job. AMD's server market share is about to disappear unless they come up with a miracle. If most of the applications I use every day would take advantage of 4-8 threads then I would be all over it (some do but I don't use them enough to justify the cost). But since I mostly use my pc for gaming I can't help but wish for an even better penryn with an IMC and 8mb+ of L2.
Expect this Winter top games to come out threaded :)Quote:
Originally Posted by shiznit93
If you don't have a 2500x1600 screen, you will see benefits. :up:
For those with 2500x1600 , well SLI or Xfire can help :ROTF: to see the benefit of Quad cores.
The new gen of game is threaded, I got the betas, it looks good!
I fly Flight Simulators...combat flight sims mostly in the jet age, and nothing will bring a CPU to it's knees faster. Everyone talks about GPU power with games, but with sims it's very different. The majority of the load is on the CPU. There is a load on the GPU but the majority of stuff is done on the CPU.
The normal deal with flights sims is "More"...more CPU, more RAM, faster HDD's, more GPU Power in that order.
There is a new title being worked on now that will probably be worked on for a few more years called Fighter Ops. it's in development and it will bring realism to a whole new level. I currently fly a sim called Falcon 4. These sims are real, and F4 has a dynamic campaign engine. There is literally a real war going on and what you do in sim has a direct impact on what happens in the final outcome of the war. It massively taxes the CPU. Earlier versions taxed the CPU even more becasue the grpahiocs engines were not designed like they are today. More was done on the GPU.
In F4 we don't fly for the pretty graphaics although there is constant work being done on that end of it. We fly for the realism. A 10 year old sim with a DX7 engine is still the very best combat flight sim ever to be put out there. It has the love of everybody who flies it.
Even F4 that is based on some older coding performs better on multi core CPU setups. It does responmd to GPU upgrades but not as much as CPU upgrades.
This new Nehalem is not gonna hurt performance over the Penryns. It will actually increase performance so for the person upgrading now there is really no reason to go with the older 775 based stuff. The 775 based stuff will only get obsolete faster. Even with older titles like I fly it will increase performance. Flight sims are something that takes many years to develop. They are not like games. People who develop them do lots of research and some of it is hard to get. These are what we call "study sims". In any event, it takes alot of CPU to do the things I do and this new Nehalem is just what I've been looking for.
actually, did you try Flysimulator with FS Service Pack 1, because it is one of the only hightly threaded program. Nehalem will scream on this one especially :welcome:Quote:
Originally Posted by T_Flight
http://www.tgdaily.com/content/view/35547/135/
" At this year’s Consumer Electronics Show, Intel’s tech guru Francois Piednoel showed off a prototype Skulltrail-enabled Alienware system by blazing through several benchmarks. He also virtually flew an Airbus 380 (he’s French after all) in Flight Simulator, one of the very few games that can take advantage of eight cores.
"
many pc games are threaded yes, but the usage on cores 3 and 4 is small enough that a 4.0ghz dual core will beat a mid 3.x ghz quad most of the time. There are some exceptions like UE3 but UT3 sucked and Gears was infested with GFWL. Source engine can use 4 cores very well with mat_queue_mode 2 but unfortunately CS:S hasn't been upgraded to the Orange Box engine and mat_queue_mode 2 isn't very stable anyway. Maybe Left 4 Dead will fix it.Quote:
Originally Posted by Drwho?
Quote:
Originally Posted by Drwho?
Are you talking about FSX? Flight Simulator 10 from Microsoft? If so, I don't currently have the computer for it...not even close. :( I am familiar enough with it and have freinds that fly it. Currently the technology still does not exist to properly fly that sim with everything maxed. It does do well with lots of memory on Vista with the high end machines, but you have to chill on the sliders a bit.
I'd actually like to see some of these reviews done with Flight Simulators. They do put a real hurting on stuff in ways that current games don't. They are a bit different. FSX could be used as a benchmarking program. Even the most expensive system imaginable would not run it maxed. If Bill Gates has a PC running it maxed, he has something we can't get yet. I've seen it tested with a 8 core setup with an obscene amount of RAM on Velociraptors in RAID, with Quad SLi and it still can't be run maxed out. A system like that cost upwards of 10 grand. That's way more than I can spend on a computer.
More multi threaded DVD making programs, please...
:)
Quote:
Originally Posted by XS Janus
www.handbrake.fr does very well, if you have an ipod video or iphone, using 8 threads :)
there are now many application using 8 threads for video ...
and if you want to rotate the video of few degree, to mess up the video, you can use the new super useful features from nvidia, blur and rotation ... lol :rofl::rofl::rofl: who wants to blur his video? :shrug:
http://www.theinquirer.net/gb/inquir...o-editing-roll
:ROTF::ROTF:
sorry, i had to take this out of my system ...
Quote:
Originally Posted by T_Flight
Fighters Anthology FTW! :woot:
As for Video, google badaboom. Encoding on yer NVIDIA card. :)
:shocked:
Yes! please try this! and tell us what you think ... The file will be 4 times bigger than Handbrake, and the quality will be worst, about double the PSNR (Signal noise ratio), and you will not have ANY deblocking ... a 2.6MB/s bit rate is a little too high for my taste, it is suppose to be around 700KB/s for an iPhone ... :ROTF: in Codec, the "C" is for "Compression"... somebody forgot about it :clap::shrug::shrug::shrug:Quote:
Originally Posted by [XC] riptide
Nice tech demo, but if you look at it closely, it is not ready at all, like most of the CUDA demos, if you digg, it is all a joke :down:.
PS: This is my very personal opinion, as a Codec expert (I program codec for 7 years with a VERY famous codec team :up:)
Feel free to react, just stay polite :)
Which one? I'm interested :DQuote:
Originally Posted by Drwho?
I could not agree more. I am working on my current build with a 3.2 Ghz Core i7, 4 15K SAS drives in RAID 0, 2 GTX 280's, 24GB of RAM (assuming DDR3 4GB sticks by the end of Novemeber. My goal is to run this program at high, not the highest, settings. Everything is going to be watercooled. Look up project Crazyness and you will see what I am talking about.
I am fairly certain even this system will not run FSX at full settings. Tech will not be around for another few years that will run that game at max settings. Finally, when a pC can drive that game at Max settings FS XI will be out and it starts all over again.
Quote:
Originally Posted by T_Flight
Hey man. :up: Thanks for the advice. I tried it already a while ago anyway. I wasn't interested in using it for Ipod/ Apple type formats.Quote:
Originally Posted by Drwho?
Take a look at post 4. http://www.xtremesystems.org/forums/...d.php?t=198524 :)
Drwho?, I wonder if this poster means DVD generating programs (via video from a digital camcorder) like Pinnacle Studio, Ulead MovieFactory, Sony Vegas, Adobe Premiere, etc. rather than recoding/compressing existing DVDs.Quote:
Originally Posted by XS Janus
I have an older version of Sony Vega Movie Studio (version 6 I believe) from 2004 that could generate four (4) threads. I am really looking forward to the Nehalem launch to build a new rig (and get the current version of Vegas Movie Studio).
Yes, I love Handbrake!
I wrote my post under influence of Ashampoo DVD burning software. I tried making a DVD from an avi file that day and noticed it used ONLY ONE CORE!
Helloooo it's the 21st century!!!
:yepp:Quote:
Originally Posted by sdsdv10
I was demonstrating at Computex, Sony Vegas 8 uses SSE4.1 and 8 threads ... Nehalem is a screamer on it. I was able to manipulate HD video that I grabbed from an HDV cam, and I was converting it at around 25FPS , with good Blueray compression, using Main Concept codec (MPEG2).
You go a little down if you do AVC or MPEG4, because the codec is doing BI directional Motion block search, that cost memory Bandwidth, making Nehalem the dream CPU for it. The GPU guys just can't do anything like this, because if you want to archive good compression, after you are done encoding the frame you have to re decode it, and check your level of quality (PSNR) before you decide if you have enough bandwidth. The GPU guys have to go back to main memory in this case, that means they have to use PCIexpress in the "wrong" direction, at 1X PCIex speed ... and that will collapse any performance advantage they could have gain with the parralelizme of the shaders. An other issue with parrallele encoding, if you have n slices of the picture to encode, you ll get n matched motion block buffers, and if you want to compress well, you have to check for duplicates between all of those n buffers full of motion matched block ... this is a very costy operation, and this is why "software x86" codec alway does better than any hardware encode. :shrug:
So, you ll get a lot of dance from CUDA apps, but they just can t do sorting, or they can't do a special treatment for one of their threads. Codecs today already look for faces and add bits where your eyes pay more attention, making smarter choice for bandwidth constrained encoding (iPhone, Zune, and internet Divx files , youtube ...)
Cuda is very similar to the MPEG2 cards competing with the Pentium II for DVD play back ... we all know how it finished: Pentium III :ROTF:
Nobody uses complexe pipeline to decode video anymore, just a little motion compensation part in every integrated graphics and GPU ... but the big jobs is done by the CPU today, and playing a DVD today is 3% of a nehalem .... :up:
"It is a matter of time, for every generic algorythm before the processor has enough processing power to do it by itself, this is what drives the PC industry since day 1."
Have a good week end guys, I am going under my Jaguar XJS 1986 to change to Catalizer, and make it a little more nature friendly ...:D