First Deneb preview!

"four nuclear processor samples"
"ultra frequency result and the radiation"

Good thing I just got my lead pants back from the dry cleaners.

"four nucleus Phenom"

I've got a fever, and the only cure is more Anti-Phenom!

"Seagate Barracuda (hopes the Czechoslovakia cool fish)"

http://www.dopefish.com/images/swimfish.gif


Babelfish joking aside:
"45nm Phenom may save near 40W under the full-load conditions compared to 65nm Phenom"

The benches all look great. Performance improvements across the board versus 65nm parts. Great power usage drop as well. Two less pins than 65nm Phenom?? The main things I'd like to see in a review now is a comparison to X2's as well so I can see how they stack up and testing on new SB750 boards to see if ACC is useful on Deneb cores.
 
LOL yeah the translations are quite funny!

Hey, nuclear power is very efficient, it could be a reason why the power consumption is so low. :D
 
These new phenoms will work in motherboards that support the old ones right, like my m2n32-sli deluxe wifi 590 motherboard? :\

EDIT: Nevermind. Asus updated the page for it, it lists Phenom support. w00t. I know what my next chip will be. :D
 
These new phenoms will work in motherboards that support the old ones right, like my m2n32-sli deluxe wifi 590 motherboard? :\

EDIT: Nevermind. Asus updated the page for it, it lists Phenom support. w00t. I know what my next chip will be. :D

Does it specifically say 45nm Phenom support though?
 
Boy is that a horrible translation...

Anywho... hope to see these chips out soon! I want 4 worthy cores already (without going intel)
 
Have to wonder how much of the performance increase is coming from the cache increase alone. Haven't heard much about the lower cache versions of the 45nm cores lately, but those could certainly be interesting for the budget crowd if they still offer most of that IPC improvement without the cache.

I'm also itching to see how this thing looks with DDR3 support (those numbers are with DDR2-800). The Phenom platform seems to dislike higher latency memories, I wonder if the added bandwidth DDR3 provides will make up for the increased latency. I don't expect to see a large performance increase going to DDR3 yet, especially compared to the price of those sticks now compared to DDR2. A year down the road things should look a bit rosier, though. I've always liked how AMD waited for memory tech to mature before it adopted it for themselves, but the Intel pushers are still necessary in order to get the technology rolling in the first place.

Regardless of performance changes, those power numbers look absolutely outstanding! These new CPUs paired with a 790GX or FX motherboard would make for one damn fine HTPC or a low power media-centered enthusiast PC. I certainly like the idea of running off an IGP on the desktop and using discrete cards for games. I was thinking it would be an obvious choice next year to go to the Nahalem, but this could certainly make things interesting for my wallet unless the Intel platforms drop in price significantly and still offer all the tweaking goodies.
 
To save several clicks, here's a summary of the results:

Clock for clock power improvements @ 2.3GHz 65nm/45nm/% improvement
idle (W)
39.4 / 32.4 / 18%

load (W)
104.1 / 57.3 / 45% <-- nice


Clock for clock performance improvements @ 2.3GHz 65nm/45nm/% improvement

FritzChess
5403 / 5655 / 4.7%

Wprime (sec)
18.9 / 17.9 / 5.3%

POV-Ray
1309.3 / 1534.1 / 17%

MainConcept h.264 encoder (sec)
110.6 / 106.8 / 3.4%

3DMark Vantage CPU score
7557 / 8044 / 6.4%

Quake Wars (fps)
80.2 / 86.7 / 8.1%

Crysis (fps)
36.8 / 40.1 / 9.0%

1080p h.264 playback Pirates of the Carribbean (CPU %)
28.6 / 26.9 / 5.9%

1080p h.264 playback Casino Royale (CPU %)
36.9 / 35.2 / 4.6%

Solid 7% average clock for clock improvement. Not bad for a shrink/tweak and a couple of % better than I had (guess)timated. The load power figure is pretty impressive, hinting at some headroom in the current wattage grades. Unfortunately the improvements clock for clock don't seem to catch up to 65nm Core 2 Quad (Kentsfield), let alone 45nm Core 2 Quad (Yorkfield). AMD is really going to need to crank up the clock speeds at 45nm. AMD will probably need a 2.6-2.7GHz Deneb to compete with the Q9300, and that's just the low end.
 
It did for some Intel when they went 45nm.

And most AMD CPUs IIRC. The main thing is that the motherboard must ID the chip correctly so it supplies the correct voltage and whatnot. The fact that these have 2 fewer pins than the 65nm Phenom's might also cause the motherboards some grief. It's not that the motherboard can't support it, it's just that the BIOS needs to know what the hell is sitting in its socket first. Should require only a simple BIOS update if anything but it's up to the manufacturer to do that.
 
It might be possible that the NB clock is only 1.8 GHz or so and will likely be increased on production models - the power usage might go up because of this, but the performance should as well. Who knows; I'm probably wrong. :p
 
Clock for clock power improvements @ 2.3GHz 65nm/45nm/% improvement
idle (W)
39.4 / 32.4 / 18%

load (W)
104.1 / 57.3 / 45% <-- nice
:eek::eek::eek: I can has Phenom nao? When is this getting released?

Also, big thanks to PXC for the summary.
 
Not bad for a die shrink. The drop in power consumption should help AMD, both in power senstive applications (server stuff) and clock speeds. 140W was kind of ridiculous.

Do we have any idea what the initial clocks are going to be?
 
it seems to scale better also the 3.2ghz phenom shows a good lead over 2.3, I remember the 3ghz didn't really do much for the 65nm
 
Looks like I'll upgrade to one of those instead of a X4 9850.
 
So, to reiterate my earlier question, when are these going to be availible to purchase?
 
Well, the reduction in power consumption is certainly impressive, and just what AMD needed.
The performance gains are nice, but not too big of a deal. I wonder how much is due to the increased cache size though, as AMD users generally were in denial about larger cache sizes. I guess we'll know as soon as the L3-less benchmarks come in.

I suppose it all comes down to how high they can scale the frequency up now. Can they become competitive in a larger range of quadcores now, rather than just touching the Q6600? Because the real money is made with the faster models, not the bargain basement quadcores.
 
Because the real money is made with the faster models, not the bargain basement quadcores.

I've always imagined the real money being made on the bargain bin processors which they make in bulk and sell in large quantities if only for a small markup.
 
I've always imagined the real money being made on the bargain bin processors which they make in bulk and sell in large quantities if only for a small markup.

Ditto. No wonder they haven't discontinued the X2 range yet.. they make for fairly powerful HTPC/home-usage machines, especially with the 780g chipset.
 
I've always imagined the real money being made on the bargain bin processors which they make in bulk and sell in large quantities if only for a small markup.

Well, quadcores aren't absolute bargain bin, nor are they really bulk-items yet... they're still mid-end in a way.
Thing is just that AMD's quadcores CPUs were too expensive to produce given the performance they delivered, so they couldn't get up to the desired markup. AMD had to position them under then Q6600 pricetag. Intel nicely used that opportunity to crush AMD's profits.
So AMD needs CPUs that are cheaper to build, and escape the pressure of the Q6600 by being better performers. And then the markup can go up to comfortable levels.
In this sense, the X2 series is actually a healthier product for AMD, which is why it wasn't replaced with a dualcore variation of K10.

Aside from that, I don't think the absolute bargain bin stuff is where the money is. It's probably somewhere halfway, just before the prices start to go up with a big exponential curve (which also means that sales figures go down exponentially).
 
Without the L3 and on 45nm the Quadcore "budget" CPU would be quite viable- if not performance then clocks.

It would be smaller than the sum of a Core 2 Quad which requires further packaging (MCM) costs.
 
Well, quadcores aren't absolute bargain bin, nor are they really bulk-items yet... they're still mid-end in a way.
Thing is just that AMD's quadcores CPUs were too expensive to produce given the performance they delivered, so they couldn't get up to the desired markup. AMD had to position them under then Q6600 pricetag. Intel nicely used that opportunity to crush AMD's profits.
So AMD needs CPUs that are cheaper to build, and escape the pressure of the Q6600 by being better performers. And then the markup can go up to comfortable levels.
In this sense, the X2 series is actually a healthier product for AMD, which is why it wasn't replaced with a dualcore variation of K10.

Aside from that, I don't think the absolute bargain bin stuff is where the money is. It's probably somewhere halfway, just before the prices start to go up with a big exponential curve (which also means that sales figures go down exponentially).

Interesting.. any follow-up links for this? I always assumed that the lower-end chips made up in volume what AMD couldn't make in terms of unit sales price?
 
Interesting.. any follow-up links for this? I always assumed that the lower-end chips made up in volume what AMD couldn't make in terms of unit sales price?

Just my educated guess.
Aside from that, you can't make the assumption that cheaper chips also sell in higher volumes by default.
In the case of AMD, a lot of their chips are so extremely slow and cheap that they're not really interesting anymore to a large part of the market. Especially here in the Western world, where we don't really have to buy the cheapest stuff, because we're plenty rich anyway. When I look around and see what kind of new PCs people buy, they're mostly mid to high-end PCs, not super-cheap crapboxes.
 
I suppose it all comes down to how high they can scale the frequency up now. Can they become competitive in a larger range of quadcores now, rather than just touching the Q6600? Because the real money is made with the faster models, not the bargain basement quadcores.

The "Bargain Basement" processors are their bread and butter... They
will "Underclock" higher quality/performance dies to fill lower performance
die orders if need be... That's why you will get CPU's that are considered
to have good "Steppings"... They are high end chips pulled from inventory
and "Toned Down" to meet the order volume for the lower end chips...

How CPU's are made
http://www.embedded.com/columns/significantbits/17501489?_requestid=7412
After each chip is tested to see if it works, it's usually tested again to see how fast it runs. Surprisingly, a 500MHz processor and a 700MHz processor aren't really different chips. They're probably neighboring chips from the same wafer that happen to run at different speeds. Slight variations in chemistry, contamination, or the phase of the moon seemingly can affect a chip's speed. It's common for microprocessor companies to sort their chips into at least two or three speed grades. The fastest 10% get sold at a premium price, while the slowest ones go to the bargain basement—or get called something else.

Chip makers commonly lie about a chip's features. Well, maybe not lie exactly, but omit certain facts. You see, embedded processors with different features or peripherals often aren't different chips at all. Vendors will produce a single silicon design but then package and market it as different chips. For example, one version might have two UARTs and Ethernet while another version has five UARTs and no Ethernet. Chances are, they're really the same chip. Sometimes the "missing" features are disabled with a laser or by blowing a fuse. Sometimes they're disabled with firmware. As often as not, they aren't disabled at all, but just aren't mentioned on the data sheet. Programmers have occasionally found "secret" peripherals that aren't connected and aren't mentioned in the manuals.

Production quality tends to improve over time, so faster chips will become more plentiful. Sometimes it's not in the vendor's best interest to let customers know that, however. Even if half of the mature parts run at the peak speed, the vendor might arbitrarily limit the number of fast chips to, say, 15% of its volume to maintain an air of exclusivity. Enterprising customers have discovered this and over-clock their parts to gain a speed advantage.

The ethusiast market is probably their smallest market... Processor
manufacturers live on the sales to industry and business... Think of
all the computers in an office building or your local hospital... Most
of them don't need to be that powerful to perform standard business
operations...

For every one home PC there are probably thousands of PC's used
in the commercial sector... ;)
 
The "Bargain Basement" processors are their bread and butter... They
will "Underclock" higher quality/performance dies to fill lower performance
die orders if need be... That's why you will get CPU's that are considered
to have good "Steppings"... They are high end chips pulled from inventory
and "Toned Down" to meet the order volume for the lower end chips...

That doesn't quite go for quadcores, which aren't bargain basement processors in the first place.
Their dualcores are the bread-and-butter.
I think you are mistaking the above phenomenon for an absolute truth, while it is more a temporary response to a certain trend. There are also times when bins are shifted upwards rather than downwards (sometimes you'll suddenly hear reports of decreased overclocking potential for certain CPU models).

The current situation is that AMD has a high-end processor (in terms of design and production overhead) which performs as a low-end (compared to the competition), meaning they don't get the return-on-investment that they were planning. Which is also why AMD has been bleeding cache for the past few years. Since AMD has only 'bargain basement' quadcores, by your logic, they should be making a lot of money... but they don't. What they need is more performance, so they can get more control over pricing.
 
That doesn't quite go for quadcores, which aren't bargain basement processors in the first place.
Their dualcores are the bread-and-butter.
I think you are mistaken the above phenomenon for an absolute truth, while it is more a temporary response to a certain trend. There are also times when bins are shifted upwards rather than downwards (sometimes you'll suddenly hear reports of decreased overclocking potential for certain CPU models).

Quad core processors were not invented for the home PC enthusiast...

They were invented so that companies could double, triple, and quadruple
the number of "Cores" with the same amount of sockets in things like servers
without having to double, triple, or quadruple the cost...

The home PC enthusiast market is a "Niche" market... A residual market
to what the processor manufacturers are really in business for, which is
to supply processors and computer systems to the commercial sector...

The reason for all the "Hype" about high-end CPU's is simply because
it's good advertisement for their product... The people that do the buying
of computer systems are after all human and can be influenced by the
excitement generated by the enthusiast market...

That's not to say they don't make MONEY in the home PC market, but
it doesn't hold a candle to what they make in the commercial market... ;)
 
Except that we're talking about Phenom and Deneb here, which are aimed at the home PC market.
 
Back
Top