Wednesday, July 29th 2009

Clarkdale 3.06 GHz Faces a Preview, Series Pricing Surfaces

"Clarkdale" is the codename for Intel's upcoming dual-core processors derived from the Nehalem/Westmere architecture. The move marks a leap for Intel in two ways: introducting the first commercial-grade 32 nm microprocessor, and implementing a radical new design that involved relocating the platform's northbridge component entirely to the CPU package. Slated for Q1 2010, Clarkdale will go by three brand indentifiers to grade it according to a performance and feature scale. You have the Core i5 class that enables the entire feature-set of processor, there's the Core i3 class that offers some features, excluding Intel Turbo Boost technology for example, finally there's the sub-$100 Pentium part (yes, Pentium lives on), which offers a smaller feature-set. HyperThreading technology is disabled on this one.

Chinese tech-site IT168 published a comprehensive performance (p)review of the 3.06 GHz Clarkdale part. In the article, the 3.06 GHz Clarkdale was pitted against the 3.00 GHz "Wolfdale" Core 2 Duo processor. The memory (Dual-channel DDR3-1333, 4 GB) and graphics hardware (ATI Radeon HD 4870, 1 GB) were kept common between the two test-beds. Tests ranged from memory and CPU internal bandwidth tests, math-intensive tests, synthetic multimedia and 3D tests, and finally, modern 3D games.

With an integrated memory controller, Clarkdale is swimming in higher bandwidth, and lower latency compared to its predecessor, as tested on SANDRA 2009. Having four threads on two cores thanks to HyperThreading, however, doesn't seem to have a significant impact on its performance at the multi-threaded Cinebench R10. Even at the single-core run, the increment is passable at best.



Notable CPU benchmarks from Everest 5, the CPU Z-lib and FPU Julia series (Julia and Sin-Julia) is where the new chip seems to shine with significant increments. A roughly 40 percent increment at Z-lib, and a healthy 90+ percent boost with Sin Junlia. Clarkdale churns out surprisingly high scores with 3DMark Vantage (CPU test), something it doesn't manage with 3DMark06.



If purely gaming is your thing, and you have a current generation machine, these charts will interest you. Popular games such as Enemy Territory: Quake Wars, Far Cry 2, and Crysis Warhead show the two chips to be performing neck and neck.

In related news, HKEPC compiled a neat table listing out Intel's first-wave of Clarkdale-based processors, starting from the sub-$100 Pentium, to the entry-mainstream Core i3, and over to the mainstream-performance Core i5. The models and their tentative prices were sourced from motherboard manufacturers. A bulk of the lineup is positioned in the $126~$196 range. At these prices, they share space with at least one quad-core LGA-1156 processor, the 2.66 GHz Core i5 "Lynnfield" 750, while the 3.46 GHz Clarkdale will be locking horns with the 2.80 GHz Lynnfield at the ~$280 price point. IT168's complete review Google-translated to English can be read here.


Sources: IT168, HKEPC
Add your own comment

31 Comments on Clarkdale 3.06 GHz Faces a Preview, Series Pricing Surfaces

#1
Kenshai
73W TDP from bottom to top of the line there sounds great. Should help keep temperatures down.
Posted on Reply
#2
Steevo
AMD has lower power, and cheaper CPU's.



Don't get me wrong these look great, but untill they face a real world set of tests from a few sites this is all fluff.
Posted on Reply
#3
TheLaughingMan
by: btarunr
implementing a radical new design that involved relocating the platform's northbridge component entirely to the CPU package
Hasn't AMD been doing this for years? I mean what AMD now calls a Northbridge is nothing like a standard Northbridge.
Posted on Reply
#4
btarunr
Editor & Senior Moderator
by: TheLaughingMan
Hasn't AMD been doing this for years? I mean what AMD now calls a Northbridge is nothing like a standard Northbridge.
No. AMD didn't completely move the NB component to the CPU package.
Posted on Reply
#6
btarunr
Editor & Senior Moderator
Dual-core is where the bulk of the money is, and it's simpler/cheaper to make (better yields), so it serves as a good starting point.
Posted on Reply
#7
mdm-adph
Well, not as impressive as whatever the hell that was running 4.0GHz at like .8v from Intel a while back. :D
Posted on Reply
#8
toyo
At this kind of performance (not stellar vs the E8400), I would expected them to be pushed cheaper, under 200$. Intel said that it will make quads mainstream, I hoped that in a year or so there will be full-featured (virtualization, TXT, HT etc.) quads at the 150$ mark, even if at slower speeds.
I can only hope AMD will come up with something and influence the prices or even put out some affordable 32 nm quads... we'll see I guess.
Posted on Reply
#9
toyo
by: mdm-adph
Well, not as impressive as whatever the hell that was running 4.0GHz at like .8v from Intel a while back. :D
That 0.8 was only in idle state it seems. Load voltages were approximately on par with C2D.
Posted on Reply
#10
HalfAHertz
I'm a bit dissapointed from the gaming results to be honest. I thought that by integrating the pci-e controller on the cpu die, you'd get better gaming results due to the reduced latency and signal noise

Maybe we'll see better results with the nextgen of video cards and mature non-ES cpus ?
Posted on Reply
#11
EastCoasthandle
by: odameyer
Who cares, it has a x23 multiplier!!
:laugh: I saw that too! Reminds me of the P4 days in which that multi had no real meaning in regards to performance.
Posted on Reply
#12
phanbuey
by: odameyer
Just think how far you could overclock with a x23 multiplier, that means it wouldn't raise the FSB would in-turn make the NB run cooler?, am I right?
well yeah, given the same platform... but it all depends on how cool/high the NB on these can go... if 180Mhz is the limit on that NB then even with 23x you will only get to 4.140Ghz.

But I will still buy the cheapest one and see if I could get 5.0Ghz out of it.
Posted on Reply
#13
TheLaughingMan
by: btarunr
No. AMD didn't completely move the NB component to the CPU package.
My old 939 board had no Northbridge. It had a motherboard chipset, but there was no notherbridge.

"The memory controller, which handles communication between the CPU and RAM, has been moved onto the processor die in AMD64 processors. Intel has integrated the memory controller onto the processor die with their Nehalem microarchitecture-based processors.

Another example of this kind of change is NVIDIA's nForce3 chipset for AMD64 systems that is a single chip. It combines all of the features of a normal southbridge with an AGP port and connects directly to the CPU. On nForce4 boards they consider this to be an MCP (Media Communications Processor)."
Posted on Reply
#14
btarunr
Editor & Senior Moderator
by: TheLaughingMan
My old 939 board had no Northbridge. It had a motherboard chipset, but there was no notherbridge.

"The memory controller, which handles communication between the CPU and RAM, has been moved onto the processor die in AMD64 processors. Intel has integrated the memory controller onto the processor die with their Nehalem microarchitecture-based processors.

Another example of this kind of change is NVIDIA's nForce3 chipset for AMD64 systems that is a single chip. It combines all of the features of a normal southbridge with an AGP port and connects directly to the CPU. On nForce4 boards they consider this to be an MCP (Media Communications Processor)."
And that "single chip" AMD chipset was the northbridge+southbridge, just like how nForce 740i for Intel LGA-775 processors is a single chip chipset too. All AMD moved was the memory controller (the portion of which it called northbridge), while the PCI-Express complex + IOH functions remained with the chipset. Such is the level of integration with Intel that it needs just a 2 GB/s connection to the chipset, while the rest of the system talks directly to the CPU package.
Posted on Reply
#15
TheLaughingMan
by: btarunr
Such is the level of integration with Intel that it needs just a 2 GB/s connection to the chipset, while the rest of the system talks directly to the CPU package.
Now I am lost. Is that not what the i3 is doing in this case? By "normal" definition of Northbridge it is a bridge from the CPU to Memory and GPU. The two systems you mentioned, the GPU controls were moved to the Southbridge.

I am not trying to argue, just learn and understand. In that case, the GPU and Memory controllers would both be on the CPU for the i3. Is that right?
Posted on Reply
#16
btarunr
Editor & Senior Moderator
by: TheLaughingMan
Now I am lost. Is that not what the i3 is doing in this case? By "normal" definition of Northbridge it is a bridge from the CPU to Memory and GPU. The two systems you mentioned, the GPU controls were moved to the Southbridge.

I am not trying to argue, just learn and understand. In that case, the GPU and Memory controllers would both be on the CPU for the i3. Is that right?
The northbridge is what connects the CPU to important system bus(es), while the southbridge handles peripheral connectivity. Post PCI-Express, a "northbridge" is supposed to house a memory controller, a major PCI-Express complex, the front-side bus (or other higher-bandwidth buses) that connect the CPU to the rest of the system.

All AMD did was split the northbridge to two, take the memory controller component onto the CPU, while the rest of it stayed on as the motherboard chipset, with HyperTransport connection between the two "halves". Now that half can remain as a lesser motherboard northbridge (as with present AMD 7-series chipsets) or unify with the southbridge in a single chip design (as with present nForce 700a series chipsets).

Intel completely relocated the northbridge to the CPU package, so now the package needs only a 2 GB/s DMI connection with all that remains on the board as "chipset", which is the Intel 5-series chipset, Intel P/H/G-55/57 is what Intel calls a PCH "peripheral hub?", which is just a glorified southbridge.

Hope this helps you understand: http://img.techpowerup.org/090729/btacpuarch1.png
Posted on Reply
#19
hat
Maximum Overclocker
I want to see a 4GHz stock cpu already... even if it's an extreme edition
Posted on Reply
#20
3870x2
by: HalfAHertz
I'm a bit dissapointed from the gaming results to be honest. I thought that by integrating the pci-e controller on the cpu die, you'd get better gaming results due to the reduced latency and signal noise

Maybe we'll see better results with the nextgen of video cards and mature non-ES cpus ?
remember that reduced latency wouldnt affect the actual FPS of the game. The graphics response would be faster, smoother, and more thorough. Probably 28-32 FPS as opposed to 22-38 FPS. Same average, different/better deviation.
Posted on Reply
#21
Mussels
Moderprator
AMD moved just the memory controller. Intel moved the whole northbridge (memory controller included)


The original definition of northbridge, was the chip with the memory controller - hence the confusion when discussing it.
Posted on Reply
#22
FordGT90Concept
"I go fast!1!11!1!"
by: HalfAHertz
I'm a bit dissapointed from the gaming results to be honest. I thought that by integrating the pci-e controller on the cpu die, you'd get better gaming results due to the reduced latency and signal noise
I think the architecture is the problem more so than northbridge/PCI Express. Clarksdale, therefore, is expected to take the same gaming disadvantage as Nehalem. Sandy Bridge is your next shot at a kickass gaming chip from Intel.
Posted on Reply
#23
OnBoard
Just upgraded from E7200 to E8400 and those game charts make me feel good inside :) Granted Clarkdale beats it on synthetic benchmarks and a nice boost in 3DMarks too, but not so much that warrants building a new system.

Good move on Intel to scrap the 45nm Havendale, they would have been even less of an upgrade and more expensive to make.
Posted on Reply
#24
inferKNOX
by: odameyer
Who cares, it has a x23 multiplier!!
Higher FSB gives better performance than higher multiplier.:slap:
Posted on Reply
#25
Mussels
Moderprator
by: inferKNOX
Higher FSB gives better performance than higher multiplier.:slap:
no it doesnt. test it yourself, no change - make sure ram speed doesnt change between the tests.
Posted on Reply
Add your own comment