When quad-core processors were released, most automatically assumed
that once they hit a competitive price point, they would become as
ubiquitous as dual-core processors are today. It’s been over a year
since the first quad-cores were released, and while quad-core
processors have certainly dropped in price and become more popular,
they have not come close to overtaking the market which dual-core
processors currently occupy. Dual-cores still offer great value for the
money, and there has not been a “killer” piece of software or operating
system which has required the use of more than two processing cores.
There’s also been backlash of sorts against quad-cores which was
rather unexpected. People are beginning to realize that quad-core chips
really aren’t necessary yet in order to have an excellent performing
system, whether it be a standard desktop, gaming rig, or a workstation.
Granted, some markets make use of the technology more than others, but
quad-cores are not the end-all, be-all in terms of having a
high-performance system. While they do provide some benefit in the
majority of cases, they also have significant caveats, such as larger
price tags, increased power consumption and increased heat production,
which can lead to a noisier system.
With a quality heatsink and some proper airflow routing,
keeping a quad-core processor cool isn’t difficult. However, when you
try to shove them in a small form factor or notebook chassis, things
get tricky. Such, Intel has revamped their Core 2 lineup with top-notch
45nm manufacturing technology, which should help their chips produce
less heat and consume less power. They’ve also outfitted them with 50%
more cache per core (3 MB per core, 6 MB per dual-core, 12 MB per
quad-core), which helps performance as well. Not to mention, they’ve
also thrown in a bunch of SSE enhancements and upped the front side bus
speeds as well. This architecture, as many people know, is called
“Penryn”, and it will encompass Intel’s product lineup throughout 2008.
However, there is one offset design of Penryn which has us excited, and
that’s “Wolfdale”, which we’ll be looking at today.
Read the whole review. (8 Pages)
With the somewhat dissappointing raw performance and lackluster
overclockability of AMD’s Phenom processor, along with the delayed release of
Intel’s 45nm Penryn quad-cores, many are re-considering if they really need
four cores. With the release of Intel’s
“Wolfdale” 45nm dual-core chip, it’s getting even more confusing.
While I’ll be writing up a lot about Wolfdale recently, here’s the
basics in a nutshell. It’s a damn fine chip. It’s pricey, especially on
the high-end for the Core 2 Duo E8500 processor - exceeding the price tags of
AMD’s Phenom quad-cores and close to that of Intel’s own Core 2 Quad
Q6600. However, it’s a mean little processing engine. Performance per clock is excellent, as is heat, power consumption, and overclockability.
The retail boxed Core 2 Duo E8400 which I grabbed from Newegg is up and
running at a brisk 4 GHz on low-noise air-cooling, with minimal voltage
tweaks. All that was needed was a
motherboard which was flexible with front side bus speeds (for this project, I
grabbed XFX’s new nForce 780i SLI platform), and kicked the FSB speed up from
1333 MHz FSB stock to 1778 MHz. Voltage
was kicked up from 1.2V to 1.4V, and well, there you go. Easy as pie. Let me tell you, a 4 GHz dual-core is definitely
competitive against lower-clocked quad-cores, and Windows Vista feels very
snappy with this new chip in place.
Even more exciting is that I’ve got an E8500 chip (likely an
unlocked engineering sample) on the way, which should see what Wolfdale is
truly capable of. A top-grade binned
chip without any kind of locks or limitations?
It’s certainly exciting – and I’m looking forward to seeing how fast
Intel’s new architecture can scale to, given the proper environment. With exotic cooling, others are hitting
close to 6 GHz on this chip. If I could
hit a 4.5 GHz on air, I’ll be happy.
With Phenom under-performing and under-whelming, it’s nice to see a
processor which is exceeding its basic specifications.
Of course, when Intel releases their cheap 45nm quad-cores,
Wolfdale won’t look nearly exciting.
Apparently, I have a soft spot in my heart for this chip, as I just can’t
stop talking about it. Once I get these
Wolfdale articles out the door, I swear I’ll shut up about it – at least until
the next speed grades are released.
Several months ago – I had just had it. Windows Vista had me so frustrated that I
formatted my system and re-swore my allegiance to Microsoft’s aging, but still
terrific, Windows XP operating system.
Vista, in my eyes, was a failure, and Windows XP was the only real
way to work on a Windows system.
Windows XP is a very forgiving operating system, and fares well on
modern day hardware. Everything just
runs super quick at this point, and application compatibility is nearly
unbeatable. I’d bet there are more
applications which will work on XP and not work on Vista than vice-versa. Of
course, a lot of newer applications are being built Vista-only, and that’s unfortunate,
but it’s also the sign of developer support for Microsoft’s new operating
system.
There are good things about Vista. The driver and security layers are greatly
improved, as is overall stability. It’s
got a massive driver base to pull from, supporting much more hardware, and has the ability to do nifty graphical things with hardware acceleration. Why the hatred? Vista takes at least 2 GB
of memory to really run properly, costs quite a bit more than Vista, and has a
questionably popular new interface. It
basically does the same core thing as XP, but requires much more hardware and
doesn’t seem to run as fast.
As it turns out, my biggest problem with Windows Vista was that
disk I/O speed. Everything just took so
long to do, even with 10,000 RPM hard disks, multi-core CPU’s and tons of
memory. It always felt like you were
doing things with weights on your shoulders, and that no matter how much
hardware (also, $) you threw at the problem, you wouldn’t be able to get those weights
off.
However, with the release of Windows Vista SP1, which is available
online if you search for it, things have changed. Vista SP1 is still in release candidate form
at this time and won’t be truly available until March in its final form.
However, SP1 fixes my #1 issue with Vista, slow disk performance, quite effectively. On the same hardware with SP1, Vista is noticeably
snappier in initial loads, start menu loads, and raw application loads. Work habits improved greatly, as I finally
felt I could “trust” the speeds of my machine to not slow me down.
Some of the Control Panel areas were renovated, network transfer
speeds are vastly improved, and the operating system just “feels” more
solid. Since its release, I’m feeling
pretty confident in saying that Windows Vista is now a superior operating
system to XP. Before it was
questionable, but now there is a solid difference between
these two operating systems. Granted,
the difference isn’t huge, but it’s there. Most tasks are just easier on Vista – and with
SP1, performance is just as good as an XP system. Windows XP was supposed to see another speed
boost with SP3, which I’ve tested, and didn’t notice any day to day performance
increases. XP with SP3 still is a very
solid operating system, in both performance and usability – and its
market is still huge.
However, I'm hesitant to move my testbeds over to Vista. Even though there are tons of
users who are upgrading to Vista, the majority of users are still XP. Even enthusiasts and gamers are split on which is the
better operating system. Vista is a bit
more difficult to test with, as well, as
I’ve never been sure about the disk I/O’s performance on benchmarks. With Vista SP1 out, I’m beginning to trust
Vista more as a reliable test platform, which will provide accurate numbers. Of course, there is the inevitable Vista
+ SP1 to OSX Leopard argument, although sadly, I don’t know enough about the
topic to weigh in. It’s possible that I’ll
have a Leopard setup in a few days to give some proper thoughts on the matter.
For now though. Vista+SP1.
Thumbs up! Honestly – not bad!
Unfortunately, this means I'll never get to make an AGEIA icon. As of yesterday, our favorite physics processor producers AGEIA has
been purchased and will merge as one with the great green semiconductor firm of
Silicon Valley, Nvidia. Of the possible
suitors for AGEIA (who designs physics processing units and a physics API),
Nvidia seems to be the best bet right now.
They have huge resources, relationships with all the fabrication plants,
and a genuine desire to drive faster processing.
As AGEIA is private, terms weren’t disclosed. More information is expected this week. However, I’m guessing this wasn’t some huge
blockbuster deal. AGEIA is still a
smallish company in Mountain View who don’t actually manufacture anything. They design the chips then send them off overseas
to be fabricated. Then the chips come back and are assembled onto cards by a
third-party board maker.
AGEIA is also responsible for the PhysX API and for creation of
the drivers – and given how much of their products have relied on software, I’m
guessing a good majority of their team are pure software guys. Nvidia gets the teams, the knowledge, the
PhysX chip designs and brand names. The PhysX API is likely the most prized
possession, as games developed for PhysX hardware will now likely to plug
directly into future GeForce GPU’s for enhanced physics performance.
Now, it’s easy to look at it as if Nvidia is just going to somehow
pair it up with a graphics processor to deliver a superior gaming graphics card
to whatATI can offer. However, I don’t
think that was the motivation here. I
feel that Nvidia is gearing up for a fight against Intel. Intel recently bought Havok, the company
which makes the dominant physics engine on the market for gaming, and paired
together with their long-awaited (codename) “Larabee” graphics processor, Intel
might be working on a game-changing product.
Nvidia is heavy on GPGPU computing as well, and processing on GPU and
PPU architectures isn’t that incredibly different – so Nvidia will get a team
of guys who know hardware physics really well – which seemingly can do nothing
but help out their GPU physics efforts.
AGEIA always felt like a company which was fighting an uphill battle against
a very hostile industry, both from competition and from consumer
acceptance. People who bought AGEIA cards were actually ridiculed,
since they could only be used in a few games, created heat and noise, and their
long term viability was questionable at best. Teaming up with
Nvidia is likely good for this very solid technology to grow.
We know that AGEIA's technology can be really impressive from a demo point of
view, and we think Nvidia is likely the best partner at this time to really
bring it to fruition.
On a secondary note – this is the second big business move to come from
Silicon Valley in the last week. The other being Microsoft trying to take
over Yahoo! Perhaps all this rain in the valley has made the valley
a bit stir crazy. C'mon Intel, go on and buy someone!
Everyone else is doing it!
AMD. Say what you will about their decision making process lately – they’re a
plucky bunch. They’ve fought off Intel several times before, and they've been in much tougher situations. Many people have
claimed that they were “done” throughout the history of the company, but AMD has
always come back and hit a home-run now and then, and the whole cycle starts
again. Their last home-run, in my opinion, was the Athlon64 X2 4800+, which was
quicker and cheaper than most other dual-core solutions on the market.
After that, Intel started to eat their market share with the Core 2
architecture, and have been able to maintain that lead for the last several
years. In their struggle to hit their next home-run, AMD acquired the additional
skill (and weight) of Canadian graphics chip dynamo ATI. The corporate merge
wasn’t exactly pretty, as both AMD and ATI shed some pretty high-level people,
and products were noticeably delayed. It’s only now that both companies seem to
be recovering to normal and are getting back in line, slowly, in both company’s
(ATI and AMD) road back to enthusiast relevance.
Read the whole review. (8 Pages)
Whilst the newly christened Wolfdale testbed gets up and running, let’s
talk about this whole Yahoo!/MSFT thing.
Some news outlets are reporting it as “Microsoft Buys Yahoo!”, which
simply isn’t the case. Microsoft has
filed a letter with an intent to takeover the company. Yahoo! hasn’t been bought yet – but this is
certainly the closest a company has come to taking over our exclamation! point!
happy! neighbors in Sunnyvale. Microsoft
got close about nine months ago, but Yahoo! rebuffed the offer.
Basically, Yahoo! is struggling against Google and against
itself. Yahoo! is a huge company now,
and such, isn’t the quickest moving company out there lately. Their ad sales have never been as strong as
Google’s, and their new savior ad platform didn’t seem to generate much beyond
initial buzz. They’ve had a ton of
managerial shakeups lately, including a founder coming back to run it. It’s been quite a long time since Jerry Yang
went back to work as CEO, and there doesn’t seem to be much of a change. Combined with a massive stock drop which
apparently hit the magic “we won’t pay more than x amount for Yahoo!” amount –
Microsoft made its move. Yahoo! could possibly be known as the company
which Windows Vista purchased.
So, Microsoft offered 44 billion dollars for this company. Even for Microsoft, this is a huge chunk of
change and one which takes a lot of effort in order to pull off. Share holders were offered twice their value
for the stock to sell to Microsoft, which honestly, is likely a pretty solid
move at this point. Yahoo! Hasn’t had “the
spark” in years, and perhaps even a collaboration with Microsoft might be a
good thing. People complain about
mergers a lot and look at the negative effects on competition. However, there are a ton of players in
this market, and these two companies might actually need each other to make a
product to formally compete with Google.
Of course, mergers aren’t easy.
There are a LOT of products between the two companies which overlap,
which can’t make a lot of Yahoo! employees very happy. Many simply won’t want to work for a company
like Microsoft. Many will be fine with
it, but might just not give the effort to a monolith company like Microsoft
rather than a slightly smaller one in Yahoo.
It might be a year before the companies get in sync, and a slow-moving
year against Google is something which neither company can afford right now.
If nothing else, this bid shows how big of a threat Microsoft
considers Google. It’s willing to drop
44 billion dollars on getting the technology and people just for a chance to
fight them. It’s a fairly large gamble –
and most people who hear about it are immediately skeptical that it could
work. It doesn’t seem like an entirely
natural fit. There are just some mergers
which “fit”. Say, for example, Pizza Hut
and Wing Stop. Complimentary products
which were having difficulty surviving on their own. This is no Pizza Hut and Wing Stop merger
here.
Yahoo! has some smart guys – and they still have the #1 website in
terms of raw traffic on the entire Internet.
That’s nothing to sneeze at, in its own.
They have a huge base of users for their mail client, and their portals
are huge traffic funnels between their properties. They have their own eco-system. They are by no means a small and fragile company. They did seem to grow a bit too big, too
fast, and seemed to lose their fire for innovation – or the innovation got
bogged down in additional layers of management.
However – despite all the weirdness, I think it’s going to
happen. The industry needs a great big
merger to shake things up a bit and get people thinking again. It’s definitely a strange turn in the pathway
which the Internet has taken thus far, and likely one of those forks in the
road which will define a large segment of technology in the future. Just think!
Children born in 2009 might not even exist in a world where we have
Yahoo! anymore! That is tragic!
Yahoo! will always hold a
special place in my heart, as it was the first website which I saw through a
browser, through Mosaic in Scottsdale, Arizona.
Good times.
Perhaps you might have seen it on a little site called Engadget, but in case you didn't see, I wrote a cool review. Did some tests which I doubt the rest of the Internet has had the ability to test on their own.
The second part is submitted to the system and should be online any day now. It's been fun working with all these cards, and hopefully the article can help out some potential workstation buyers find the right card to buy.
Really looking forward to seeing Nvidia's new QuadroFX 3700 card in action against all the cards I tested with. Anyways. It was a lot of work, so check it out.
Despite the fact that I have absolutely no need for one of these
processors, I’ve gone ahead and ordered an Intel Core 2 Duo E8400. Little bugger should be up in the testbed
tomorrow night. This is Penryn’s
first chance to shine in a dual-core product, emphasizing the low-performance,
low-heat attributes of this 45nm architecture. To
date, Penryn has only been seen in a quad-core design, which (of course) is
going to be significantly heftier in terms of power consumption.
The E8400 runs with two cores at 3.0 GHz each with a shared 6 MB of L2 cache, which is
still insanely powerful for a desktop processor. I honestly am curious to see how Wolfdale
Core 2 Duo processors, running at stock speeds, really have a chance to compete
directly in terms of CPU performance to AMD’s quad-core. Most processor-specific benchmarks show a
pretty large delta between dual and quad-cores – although I have a feeling that
with dual-core’s adding additional power, we might see only see a very small
gap between high-end dual-cores and low-end quad-cores – impressive as dual-core processors are significantly less expensive than
quad-cores at this time. The E8400 runs
at $225 currently, about $50 less than Intel’s slowest quad-core. If
performance was close between the two, while running at half the power
consumption level, dual-core chips might not to look so aged - and in fact - may see a resurgance.
Folks purchasing from the same source as I are claiming excellent
overclocking numbers with little work on very simple air cooling. I’m honestly hoping to see this little 3.0
GHz dual-core hit 4.0 GHz clock speeds on silent cooling methods, and be able
to match (or beat) the speeds of their stock quad-cores across the majority of
tests. Seeing that much performance
from such a low-power chip would be an impressive technology jump.
Also, it’s been a while since I’ve run a home system which would
run at clock speeds well beyond anything which is commercially available on the
market. That’s always a slight thrill
when starting up one’s system. There
is something oddly exciting about testing out a new processor architecture
though. Seeing what it can and can’t do, as
the system “feel” changes with the usage of different programs – and being
impressed by something becoming visibly faster than you’d ever seen it go
before. I’m hoping that Penryn will give
me that feeling – especially when clocked exceedingly high. The last Core 2 architecture ran Vista great
at clock speeds of 3.0 GHz and higher – and everything with Wolfdale seems
improved, assuming we don't see any unexpected jumps in latencies.
I, uh.. also picked up one of these. Ah, the Coolermaster Cosmos. It's a beast. Yes, it will be huge, but Extended ATX cases are really nice to work in over long periods of time. Really nice modern feature set, design is somewhat tasteful, and a heavy focus on low-noise with decent airflow as well.
While some of me thinks it may be coincidence, the eternal cynic
in me remains skeptical. On the day of
AMD/ATI’s coming out party for their fancy new Radeon 3870 X2 graphics card, an
anonymous leaker in China releases pictures of Nvidia’s upcoming 9800 GX2 card,
which will be the direct rival to the 3870 X2.
Of course, viewers who were reading information about the new ATI card
would also just happen to see news bits about this new Nvidia product, and
such, comparative discussions roared throughout the day. That’s not to say ATI’s day was ruined, but
whoever was responsible for leaking these pictures certainly did a good job of
distracting potential buyers.
Of course, I looked at them, and no doubt many others did as
well. The reason why these images are so
interesting is not that this is a simple “hey, we can do it too!” product –
Nvidia is doing some really bizarre and interesting stuff with this product. While the end result will be the same
final product (dual GPU’s on a single PCI Express board with all of the
interconnects done on-board), ATI and Nvidia are going at it in very different
ways. While ATI is going with a single
PCB solution, Nvidia’s board actually uses two PCB’s, sandwiched together and
connected via a mini-SLI adapter.
In essence, what Nvidia is doing is making two GeForce 8800 GTS
512MB cards in a short factor format, and including an SLI bridge chip on the primary
card to handle PCI Express communication between the chips. Each PCB (basically, each card) requires
its own 6-pin PCI power connector,
although this sandwiched together multi-card only connects through a single PCI
Express slot (presumably PCI Express 2.0).
The REALLY interesting thing here is the cooling system. With two physical PCBs, Nvidia is looking to
cool both boards with a single block of metal between them. Nvidia’s prior attempts at a multi-PCB card
(GeForce 7950 GX2) featured super-slim individual coolers for each board –
which had to include high-speed fans to keep the “inner” board cool. These cards had a tendency to overheat a lot –
and Nvidia is not having that situation again.
This time around, one of the PCB’s is flipped, so the two GPU’s
are actually facing each other. Sitting
in-between them is a huge chunk of copper, surrounded by a highly customized
aluminum alloy block which covers the memory chips and voltage regulation components. Sitting at the end of the board is a blower
fan which will pull air from the insides of the case, blowing air over the
copper cooling center and outside of the case.
Nvidia has made cutouts in the PCB itself to allow for airflow to go
through it and reach the ultra-hot center of this card. It will be very interesting to see how hot
this thing gets, with two hot GPU’s pushing heat into one chunk of metal under
heavy stress tests. For those who don’t
realize how hot these things get – I seem to recall a comparison way back when
(don’t quote me on this) which claimed that modern CPU/GPU’s create more heat
per square millimeter compared to a nuclear reactor. Granted, we’re talking about vastly
different spaces – but it would not be surprising to see these chips running at
levels of 100C (212F) and up.
Other interesting pieces?
DisplayPort output (Win for Nvidia) along with an optical S/PDIF audio
connector. Presumably, this is an
optical input connector to allow the board to connect to HDMI systems and push
audio and video signals in a single cable.
However, so much is up in the air at this point, it’s hard to be
sure. Also interesting is that the GPU which Nvidia
is using is a newer version of the 8800 GTS, which may include improvements in
power consumption and/or performance. It’s
definitely an interesting product – and again – the rumor mills are pointing
towards the March timeframe.
As expected, AMD/ATI’s new Radeon 3870 X2 graphics card has now
launched, and AMD now is producing the fastest single-PCB graphics card on the market. There's an excellent review over at the Tech Report. Of course, that single-PCB tag in there brings up a lot of debate. Can this card count as a single graphics card
when in reality it’s just two graphics cards glued together onto a single PCB? ATI and Nvidia fans will debate this point
until Nvidia releases their own version of a similar product, at which point
the topic will be dropped. Let’s face
it folks, this is probably the future. The ultra high-end GPU, as we know it,
is probably on its way out. Multiple
mid-range GPU’s connected together will likely dominate the high-end market in
the future.
ATI Radeon 3870 X2 1GB. The graphics card which silently judges you.
Ultra high-end GPU’s are simply too expensive to develop and
fabricate, whereas mid-range GPU’s are much less of an undertaking and can be
manufactured with much better yields.
If we’re at a point to where the hardware and the software for multi-GPU
processing have reached a point to where it’s seamless for the end user to use
one of these cards as it would be for a single GPU card, this would seem like a
great move for both ATI and Nvidia. They
can take one GPU design, make tons of them, and sell it in both mid-range and
high-end scenarios, depending on how many GPUs you want in your system.
The CPU market went through this exact same growing phase several
years ago. When dual-core processors
first hit the market, there was the same debate as to whether a system with two
CPU’s connected together could be compared directly to the performance of a
single CPU. These graphics cards, are
in essence, dual-core. One graphics
card, two graphics cores. This is not
that different from a single CPU package with two processor cores underneath. It took a little while for software developers
to start coding for multi-core CPU’s, but now it’s standardized for the most
part. GPU’s will get that way too – most
game developers already code with SLI and Crossfire multi-GPU environments in
mind. All the pieces lined up, and AMD
realized this and have executed with a fantastic product. Kudos to them. Good timing!
Things will just get better from here on out. Eventually, ATI will figure out a way to
lose the PCI Express bridge chip which connects their two GPU’s, and will build
this logic into their next generation GPU’s.
That will cut down on the board
size and will make for a much more “tolerable” solution. We’ll see improvements in power consumption,
as GPU makers learn how to switch on and off their GPUs when not being used to
save power. If they can cram a pair of
GPUs onboard with a single PCI Express power connector and have the card width
no longer than a standard ATX motherboard, we will start to see real movement
towards these products.
In any case – 3870X2 cards.
$449. Plenty for sale on Newegg. Comparable to the
performance of $630 8800 Ultra board – roughly the same size and power
consumption. Works on any motherboard,
doesn’t have to be Crossfire ready.
Noise level appears to be a little higher than an 8800 Ultra board, but
not enough to really be considered a hassle.
Yep, ATI has a winner. Of course,
everyone knows that Nvidia is right on their tail with a similar product based
on dual G92’s, and based on what we know of this GPU in single and SLI setups,
it’s pretty clear that if Nvidia clocks them at the same level, it will beat
the 3870 X2. Now, the waiting game for
this product begins. How will Nvidia
price it? Will it be huge and loud? Will their single-card SLI solution work in
non-SLI motherboards? All good questions
– and it looks like we’ll find out in the March timeframe.