I will definitely be waiting until they get all the bugs ironed out. It seems it was even too early to launch today...
Printable View
I will definitely be waiting until they get all the bugs ironed out. It seems it was even too early to launch today...
Sure not at these prices...Quote:
Originally Posted by BeardyMan
They could add worlds worst performing quadcore aswell. True means nothing.
No, cache dont benefit K8 or K10 arch much. Also the main reason for Core 2 to have so much cache is to counter the much lower memory bandwidth and hide the latency.Quote:
Originally Posted by JVguest
Again, check K8 512Kb vs 1024kb. And use 90nms for both to compare with.
I love these 'it is the cache' explanations.... it's like people don't understand what an integrate memory controller really does.Quote:
Originally Posted by JVguest
While I sort of agree, I think that's old news in the server market. Barcelona availaibility was poor post official launch, production cost per 1000 CPUs is high when factoring in yield and the die size, much worse than 45nm Penryn, but it started to appear in good volumes and sold out repeatedly after around 25th October. Before that it was hardly available to enterprise customers AFAIK, especially the 2350. Not in ones and twos but in volume orders. However, Penryn server is hardly being sold and supported by major MFGs because of bad platform support and instabilities, which gives AMD room for much leway. HP is one and Rahul stated on the 14th Nov why (his site):Quote:
Originally Posted by JumpingJack
Quote:
We haven't launched Intel's 45 Nanometer processor as planned. We, like many, were hoping that it would work flawlessly on certain chipsets - and well, unfortunately it doesn't - not yet anyways. Even though we were getting close to qualifying it - last week we received some really bad news. The bottom line is we're working on a solution for Nvidia SLI, but at the moment there isn't one.
Interestingly enough there are those out there configuring this processor in configurations which we *know* are unstable. I think their customers will be in for major disappointment based on current issues. ..although there are certain configurations which do work, the question is whether they deliver a decent value to the customer.
That said, it's not an issue of Intel's chip reliability, it's an issue of platform stability on certain current non-Intel platforms.
I don't want to get into the details, it's not a pretty situation. There is much confusion surrounding this launch -- it's somewhat unbelievable.
OEMs still cant ship Barcelona servers.... IBM dont even exopect to be able before december. Thats why they had to accept the humilliation of the NON COMPLIANT on spec.org for their Barcelona server(s).Quote:
Originally Posted by KTE
So no good volumes yet. There is a few popping up here and there. But lets face it. Even EE CPUs are in bigger volume.
Also your penryn server statement is BS. I already have RUNNING Harpertown servers from Dell! And they sell them like any other.
On the OEM side, yeah... the problem with this is in a third party chipset. There are plenty of other OEMs that will launch with a P35 chipset and no issues. Unfortunately, Intel will not be in mainstream with 45 nm yet, as they are not threatened by anything AMD has to put a cheaper 45 nm CPU into the market.Quote:
Originally Posted by KTE
Rahul is a boutique maker, so high octane gear (which means SLI) is what he will deal in... it is unfortunate, but the problem is not the CPU but the chipset.
I don't think AMD has any room, thier top bin Quad is slower than Intel's lowest bin quad... dual core is still the mainstay for the moment, and the K8 is still at a huge deficit to 65 nm C2D.
This just isn't pretty.
maybe i should have been more specific... i was thinking that the quadcores are pretty much limited to the highend market which means that most quadcore buyers have a big monitor, big monitor = high resolution = gpu limited
with gpu limitation the difference in games is small, in my opinion small enough that phenom could be an option for gamers if priced right.
but, yes, in general the Q6600 is the better deal, especially considering even my "crappy" B3 goes easily over 3Ghz
See this is fair... however, at 250-300 price points, these are more mainstream... heck Gateway and a few others put out Quad retail units to the market not too long ago.Quote:
Originally Posted by naokaji
The problem is this... if you are asking the question -- which is computational the better CPU, then you must ensure that you are stressing the CPU and not some other component of the system. Granted, and I will agree with you, at 'realworld' conditions, at resolutions I would enjoy a Phenom or Q6600 will be the same FPS wise; however, if games is all I want I would rather throw 300 or 400 bucks at a console rather than 1000-1200 at a new computer.
Nope.... I enjoy gaming, as do most people, but a computer is more than just a glorified Xbox or PS3. I also like to encode home movies to DVD, photoshop my photos, rip music, etc. etc. Furthermore, if a good GPU does come out that does make a huge leap, I want to know whether or not my CPU will bottleneck it or not... hence, I would want the CPU that runs gaming code the fastest at the time I decide....
Jack
there wont be much of a performance difference between the q6600 and the phenom, at least unless you monitor your fps like a hawk and write it down ever couple of sec you wont notice it, phenom is a nice cpu, it offers a option now, plus its compatible with the older am2 boards which makes it nice and easy to upgrade, intel all though there are a variety of options the still cost alot and if you want the most performance you need that new board.
its a good option to those who have a am2 board and don't want to get a new one, as well as mainstream makers will being producing rigs with these chips which will sell, cause out of all the people at my work at least they don't do any research.
I don't think you would want to upgrade a 6000+ or higher for a Phenom though, as it is getting beat by that CPU in most benchmarks to.Quote:
Originally Posted by knightwolf654
However, if you are in for a complete new build then it will take some real thinking -- do you buy into a platform with a promise of getting better? It is really up to the purchaser, but for me as an example -- I will want performance even at the expense of a few bucks. For various reasons, if it comes to gaming then again... even if I won't notice the FPS now, there is the future where a new GPU or a new game comes that may want or need that extra oomph... again it is preference, it is clear you are made up on Phenom though it is a low end quad.
Here is an example: http://www.matbe.com/articles/lire/5...500/page16.php
I mean look at that disparity....
which is the reason i am waiting for a later stepping, plus that comparing a 2.4 ghz phenom up against a 3.2 windsor, clock them the same and will see,
a later stepping would hopefully mean higher o/c if i remember correctly the x6800 was faster then the qe6700 or something of that sort. intel has had alot more time to correct the problems with their possess.
Huh ?Quote:
Originally Posted by KTE
Have you ever seen a server ? Wtf does Nvidia desktop chipsets have to do with Xeon servers ? :down:
How much is Nvidia chipset out of the desktop volume ? 2-3% ?
The remaining 97% are Intel chipsets and Penryn is rock solid and shipping on volume with them.
Everybody is selling 45nm Xeons , just as Intel can't make enough $1k Extreme Editions.
http://www.theinquirer.net/gb/inquir.../11/13/penrynsQuote:
The chips are out, in fact sold out, and the only complaints that the partners on the floor have is that they can't get enough high-end parts, as Tier 1 gorillas are grabbing everything despite the $1K+ price tag per piece.
http://www.eetimes.com/news/semi/sho...3102713&pgno=2Quote:
We are raising our Q4 revenue and EPS estimates on Intel to reflect continued positive data points on PC unit growth and dominance of most segments of the market," said analyst Doug Freedman of American Technology Research, in a recent report.
"We believe Q4 is setting up for another solid period of PC unit growth, as evidenced by strong late Q3 chipset demand at Intel and data points out of companies including Asus, Acer, and Compal," he said, referring to the Taiwan ODM giants. "As a result, we believe the current quarter is likely tracking ahead of plan for PC component companies with Intel's fabs full and demand outstripping supply."
There's bad news for Intel's rivals. "We believe the company's server and notebook parts continue to dominate the OEM space as Barcelona is late to ramp and Santa Rosa continues to benefit from record notebook shipments," he said, referring to Advanced Micro Devices Inc.'s latest processor, dubbed Barcelona.
"We believe [Intel] will continue to dominate the high-end market as AMD has been late to deliver Barcelona/Phenom, both of which are struggling to deliver performance boosts due to a bad marriage of process technology and design," he added.
Frankly, based on the news and data we have seen, this is the advice I would give any AMD seeking buyer, wait until B3 comes out... it is clear they need another spin.Quote:
Originally Posted by knightwolf654
To fix what ? Errata ?Quote:
Originally Posted by JumpingJack
Performance won't increase even if they spin it for 10 times more.The difference between B1,BA and B2 was solely in some people's heads.
Moreover , the wonder stepping B2 , which caused AMD engineers to "dance in the isles" is proven faulty when going over 2.3GHz.
And some said it could reach 3GHz.Even if you OC one , what's the chance to get random errors and lockups ?
By the time B3 is released ( late Q1 according to AMD , mid-late Q2 by our standards ) you're a few months away from Nehalem/45nm K10.
Since the 45nm K10 is a K10 shrink , I see no reason whatsoever why it would outperform a current Kentsfield per clock.I doubt it will reach higher frequencies too.
As most review sites note : there is no reason to buy a K10.Slower , hotter , same prices as yesterday's Kentsfield.
Quote:
Originally Posted by savantu
:) There is enough rumblings of problems this and problems that, that if someone was bent on AMD and Phenom and asking me should I buy now or wait, I would advise wait. Not because it would get any better, but man... if 2.4 GHz induces a bug (for what ever odd reason, from circuit logic stand point it does not make sense) .. then the 2.3 GHz is running near a cliff, and frankly I have been burned in the past with AMD pushing the limit of the spec so much so that the CPU I was using literally gave out, hence my advice to someone asking ;)
If someone were asking me, what should I get in general, right now that nod goes to Intel, of course -- runs cooler, more overclocking headroom, better chipsets, their slowest quad is faster than AMD's fastest quad, superior process technology-better made etc. etc.
What I am not able to do and what I am not going to do is try to convince knightwolf654 that Intel is a better choice, he most obviously has his mind set on a Phenom, plain and simple. That is his choice, if he wants to spend 280 bucks on a low end CPU it is his money he can do that... but he stated he will wait.. I simply agreed and that would be my advice :)
EDIT: Let me add to this... it is kinda funny to watch people argue that a new stepping will fix some performance ridden errata that will make a huge difference. I have followed the industry for 17 years now, and the point at which an release stepping showed a certain performance, no where have I seen subsequent steppings fix anything that has given major overhaul to the intrinsic per clock performance. At this stage, the architectural capability of K10 is fixed where it is, there is little more or less that AMD can do that will improve it with just a simple respin, other than may be getting L3 latency down. But 10 or 20% for a errata fix??? Yeah, right -- pull the other leg and it will whistle 'jingle bells'.
Yeah, we "threw" ours together in about three days....which is why it is a preview. I all reality we had NO help from AMD in any way, shape or form which is why we are not using many of AMD'd canned benchmarks. We weren't invited to Tahoe, we didn't talk to AMD....yadda yadda yadda...
Also note that our preview is done with a sample directly from the retail channel so this is the performance a consumer will most likely get. Miso banged this one out extremely quickly. Also, I personally have an Phenom 9500 which will be reviewed with the entire Spider platform in the coming weeks.
There are also a few other things that really MUST be said about the Phenom:
- Personally, I think this will mostly be used at a drop-in solution for AM2 users out there WHEN motherboard manufacturers get their BIOS out.
- Without a massive speed bump, AMD is on a wing and a prayer right now
- Speaking about performance....with these figures quad Crossfire is almost dead before it is officially released. Imagine the processor bottleneck we will see with four HD3870 cards teamed up with the Phenom? Ouch...
not totally unexpected
the only bad division AMD has is its cpu division
they have great chipsit
they have good graphics cards (which do make money....,only 2m$ previous quarter,this one will be better)
and they have mediocre cpu's
if they would have at least matched kentsfield they would have a tremendously strong platform,too bad it didnt happen
For suckers only...
Pretty sad to see this. Its like the runaway R600, but many times worse in comparison.
They had to release now because any longer would mean head on competition with midpriced Yorkfields and Wolfdales, which would just be slaughter. The price advantage is pretty much offset by what you have to pay for the 790FX boards, so its only really practical for those who already have AM2 systems.
Looks like it was a good idea to purchase ATI after all :)
what are the 790/770/780 boards going to cost?Quote:
Originally Posted by Miwo
I also think the cache is the problem, but a different one. The L3 cache is probably what holding back the clockspeed on the K10. AMD's 65nm tech is not that bad, the Brisbane 5000 BEs can hit over 3ghz easily. I think what AMD needs now is not another respin, but a revised design. They should scrap the L3 cache and just increase the L2 cache to 1mb per core. They did it with the K8, so we could reasonably assume they would be able to do that to the K10. Then we should be seeing at least 2.8+ ghz K10s.Quote:
Originally Posted by JVguest
gigabyte DQ6 790fx is ~200€ but there are cheaper 790fx ranging from 130-180€
Wrong on the first part. K10 has twice the SSE throughput of K8 and improved IPC (well, somewhat, because it's doesn't have large enough L2 cache), the two aren't comparable. P3 (support 4 way SIMD SSE) wouldn't gain much from going from 1 meg to 8 meg L2 either. The purpose of L2 cache is to feed data to the CPU at a high rate and at low latency <8 ns). Even with a hypothetical triple channel DDR2, AMD's IMC + small L2 cache will be far slower than a large L2 cache, that's connected with a 256 bit bus to the CPU. ;) The most important thing is low latency under <8 ns on the working data sets used in typical software, although you are right that the prefetchers and large L2 do reduce the amount of time core 2 spends accessing memory.Quote:
Originally Posted by Shintai
If that was true, we would see whole other SSE scenarios on 1 and 2MB Core 2s. In short, you are wrong. And I´m pretty sure AMDs engineers would have done this already if there was a valid reason. But there aint. I know it would be nice if there was an easy killer solution. But it doesnt exist, period!Quote:
Originally Posted by JVguest