Penryn Build

Q

quadnad

Supreme [H]ardness
Joined
Oct 24, 2005
Messages
7,656
Hi Guys-
Many of you know my straight up AMD bias, but I'm not one to overlook facts. So that means come Feb/Mar I'll be picking up a Q9450 and moving the build in my sig (sans Ultra) into an HTPC case for home theater duties.

However, I have been out of the Intel loop for a long time, so I'm in need of suggestions.

My budget is fairly flexible (as I think most enthusiast budgets are), but coming from AMD's full featured $70 motherboards I find $300+ incredibly hard to swallow. In the $100-$200 range would be best.

I'm looking for:
  • Great overclocking capabilities -- Meaning it's overclocking friendly and stable when cranked to eleven. CPU will be watercooled by parts in sig, and paired with some 1066 Tracers to assist.
  • Reasonable feature set -- I'm a simple guy, so I don't really care about extras like wi-fi, firewire, etc. SLI/CF support isn't necessary, though I may go with it if the price difference isn't too steep.
  • Speed -- I'd like a board somewhere in the top end of the bench charts for subsystems.
  • Native Penryn Support -- This may sound obvious, but considering a lot of people don't read the whole post I thought I'd add it here too :D

How does that sound as a set of requirements? It's funny how if someone asked me this for the K8/K10 side of things I'd be able to reply in a second, but there you go.

A quick aside: am I correct in understanding that the FSB of Penryn is 333mhz, which would call for DDR2-667 RAM to run 1:1?

Any and all help is much appreciated.
 
The Penryn Family has some 400Mhz FSB parts as well. Though, yes, the Q9450 has a 1333Mhz FSB QDR:

Q9450: 8 × 333 = 2.6Ghz, DDR2-667 << STOCK speeds

nForce 780i is your best solution for SLI, though the NVIDIA chipset based boards aren't as stable as the P35/X38 boards. X38 is your best solution if you want Crossfire. However, if you don't need SLI/Crossfire, don't bother with it unless you plan on upgrading that 20" LCD of yours to a 24" 1920x1080 or larger.

Most P35 boards can OC great. If you don't want to deal with aftermarket cooling for the chipset, get something that uses heatpipes for the NB and mosfets at least. If you'll be WC'ing the chipset, then my point is useless.

There are many boards that fit your needs. The two basic boards I generally recommend are the Abit IP35-E and the Gigabyte GA-P35-DS3L (good if you need legacy ports); both have 4 SATA ports.

The next step up is the Abit IP35 (heatpipe cooling for nb/mosfets; firewire; 6 SATA ports w/ RAID) or the Gigabyte GA-P35-DS3R (if you need 8 SATA ports and RAID). The DS3R is typically $130 and the IP35 vanilla is typically $120 - $25MIR at mwave.com. I like the DS3R for OC'ing since it saves BIOS profiles, but I hate its PCB length.

Then you get into the fancier boards, like the GA-P35-DS4, Asus P5K-E, and the crowd favorite, IP35 Pro. These are the higher end P35 boards with full features.
 
If you're main goal is to OC the living daylights out of a quad, you will be much better served by a current gen Kentsfield Q6600 G0 chip than the Penryn.

The issue at hand is that while the Q6600 ships with a generous x9 multiplier, allowing for an achievable 3.6ghz OC w/ a 400mhz FSB with a good air cooling solution (and a 4ghz+ OC at 425-450mhz provided you get a good chip and have a great air or good water solution,) the Q9450 has an x8 multiplier, which makes 3.6ghz the best you'll likely be able to get from the chip--and that's still with a good, well behaved motherboard. The Q9450 will need 450mhz to run 3.6ghz, compared to the G0 Q6600's 400mhz, and if you are after 4ghz (which should be obtainable with a water cooling setup such as the one you are planning,) your only two viable options are a very good (good batch / low VID) Q6600 G0 or a QX9650 for $1000. I would go the G0 route, as even if you have to spend some time exchanging chips until you get a good one, it'll still be much cheaper in the end. Due to the Q9450's multiplier, you will need a 500mhz FSB, which is currently unobtainable on 99.999% of all motherboards out today (and it looks like the X48 series isn't going to reach it either.) You could get lucky with a golden (and I mean golden) north bridge, or you could spend a great deal of time and effort trying to tweak your MB to 500mhz, but in the end the best route remains either $1000 for the QX or $250 for the Q6600 G0, and some effort spent ensuring you get a good one.

The Penryn / Yorkfield chips do run exceptionally cool, but the X8 multiplier kills any dreams about wild OCs. It's bad news indeed, but considering that Nehelem is due out in a year (which means if all goes to plan, Penryn chips will have half the shelf life of Conroe,) sticking with a Kentsfield may not be a bad move in the end.

Mark.
 
Markyip1 said:
If you're main goal is to OC the living daylights out of a quad, you will be much better served by a current gen Kentsfield Q6600 G0 chip than the Penryn.
Click to expand...

That's a good point about the multiplier; I'm not necessarily going for the WR Penryn overclock, but I forgot about how low the multis are on the Core processors.

I'll have to consider that, although the increased IPC of Penryn may make this point moot.
 
enginurd said:
The next step up is the Abit IP35 (heatpipe cooling for nb/mosfets; firewire; 6 SATA ports w/ RAID) or the Gigabyte GA-P35-DS3R (if you need 8 SATA ports and RAID). The DS3R is typically $130 and the IP35 vanilla is typically $120 - $25MIR at mwave.com. I like the DS3R for OC'ing since it saves BIOS profiles, but I hate its PCB length.

Then you get into the fancier boards, like the GA-P35-DS4, Asus P5K-E, and the crowd favorite, IP35 Pro. These are the higher end P35 boards with full features.
Click to expand...

The IP35 Pro looks to be a great board, but I'm surprised since I've heard so much about Abit having lost their way. I'm willing to pay the premium for a board with great features, and if that one overclocks well it may be the one I'm looking for.

So all those boards support Penryn natively?

About Nehalem: After having looked at the latest info, it seems that Nehalem isn't due out in the non-extreme form until early 2009. Provided there are no delays, that is. I think a Penryn-based quad will serve me just fine until then, and hopefully in 2009 we'll have a little bit of competition restored to the market.
 
All the boards I listed will OC well. The problem Mark mentioned is specifically with quads. Quads don't like high FSB speeds. Though, P35s with dual cores hit 500 more consistently than any other chipset that I know of.

Bearlake Family, which includes P35 Express, was designed for Penryn.
 
enginurd said:
All the boards I listed will OC well. The problem Mark mentioned is specifically with quads. Quads don't like high FSB speeds. Though, P35s with dual cores hit 500 more consistently than any other chipset that I know of.

Bearlake Family, which includes P35 Express, was designed for Penryn.
Click to expand...

I see, so the max FSB of those boards is in the region of 475-500 but it's the quads that hold them back.

So is there any reason I should be going with the Abit IP35 Pro vs the GA-P35-DS4 or Asus P5K-E? I've found most of the Asus AMD motherboards to be horrifyingly low on O/C options, but I don't know if that has changed.

Also: is there any reason to consider the X38 based motherboards?
 
500MHz is possible on around 1/2 of all P35 boards. To keep this going stable, and keep everything cool, you will want some form of active cooling. Otherwise, just get a Q6600. As you say that you have a generous budget, maybe the QX9650 is a possibility?:confused:

Asus boards are generous with overclocking options. If you want the best, consider a DFI board. Get some more standard memory at DDR800 rather than 1066 stuff.
 
lamboman said:
As you say that you have a generous budget, maybe the QX9650 is a possibility?:confused:
Click to expand...
Affording it isn't the problem; I would never let myself live down the fact I just paid $1k+ for a processor whose overclocking capabilities will yield only marginal real-world benefits over one a third of the cost.

To be honest, I just wish they had some more reasonable multis on the penryns. If I'm correct, isn't the step up from the Q9450 running an annoying 8.5 multi? What is the cost of the proc running a 9X multi (I found a great table with all the info, but naturally forgot to save it)?
 
There is no non-Extreme Penryn Quad with a 9x multiplier. There could easily be one, but then there would be no reason to buy the Extreme chip, would there?

500MHz is possible on around 1/2 of all P35 boards. To keep this going stable, and keep everything cool, you will want some form of active cooling. Otherwise, just get a Q6600. As you say that you have a generous budget, maybe the QX9650 is a possibility?
Click to expand...

Many P35 boards can hit 500mhz with a dual core, but very few (if any) can do so with a Quad. Again, it's not impossible, but very difficult to achieve. You would have to luck out with both a good chip and motherboard, and be willing to spend a great deal of time tweaking and cooling.

Mark.
 
The only reason to go with X38 is for dual 16x Crossfire, IMO. PCI-E2.0 isn't a good enough reason for the price premium because current cards don't even saturate the PCI-E1.1 bandwidth. So by the time PCI-E2.0 cards are severely limited by the PCI-E1.1 slot, I think it will be time for a new mobo anyhow.
 
Markyip1 said:
There is no non-Extreme Penryn Quad with a 9x multiplier. There could easily be one, but then there would be no reason to buy the Extreme chip, would there?



Many P35 boards can hit 500mhz with a dual core, but very few (if any) can do so with a Quad. Again, it's not impossible, but very difficult to achieve. You would have to luck out with both a good chip and motherboard, and be willing to spend a great deal of time tweaking and cooling.

Mark.
Click to expand...

True. It is possible to get pretty close to that for a broad amount of users though (475 or something like that, would be less of a problem).

quadnad, same here, I would never pay that much for a chip. But, he might feel like it. Some people do. A bit like my obsession with Lamborghini. I don't need one, I could have a faster car, like a turbo or supercharged Corvette. But, if I had the money, I would have a Lambo.
 
lamboman said:
quadnad, same here, I would never pay that much for a chip. But, he might feel like it. Some people do. A bit like my obsession with Lamborghini. I don't need one, I could have a faster car, like a turbo or supercharged Corvette. But, if I had the money, I would have a Lambo.
Click to expand...

hah, I'll take a Reventon please.


More on topic though: if the X38s don't yield any other benefits over the P35 than the PCI-E 2.0 compatibility and dual x16 Crossfire (I won't be going SLI/Crossfire in any event), I think the P35 will be the way to go. Now it's time to start shuffling through reviews of the Abit, Asus and Gigabyte P35s. Thanks all for your responses so far, and if you have anything more to add please do.
 
Markyip1 said:
If you're main goal is to OC the living daylights out of a quad, you will be much better served by a current gen Kentsfield Q6600 G0 chip than the Penryn.

The issue at hand is that while the Q6600 ships with a generous x9 multiplier, allowing for an achievable 3.6ghz OC w/ a 400mhz FSB with a good air cooling solution (and a 4ghz+ OC at 425-450mhz provided you get a good chip and have a great air or good water solution,) the Q9450 has an x8 multiplier, which makes 3.6ghz the best you'll likely be able to get from the chip--and that's still with a good, well behaved motherboard. The Q9450 will need 450mhz to run 3.6ghz, compared to the G0 Q6600's 400mhz, and if you are after 4ghz (which should be obtainable with a water cooling setup such as the one you are planning,) your only two viable options are a very good (good batch / low VID) Q6600 G0 or a QX9650 for $1000. I would go the G0 route, as even if you have to spend some time exchanging chips until you get a good one, it'll still be much cheaper in the end. Due to the Q9450's multiplier, you will need a 500mhz FSB, which is currently unobtainable on 99.999% of all motherboards out today (and it looks like the X48 series isn't going to reach it either.)
The Penryn / Yorkfield chips do run exceptionally cool, but the X8 multiplier kills any dreams about wild OCs.

Mark.
Click to expand...

Hi. i was thinking what if the 3.6Ghz Penryn out performs the 4.0Ghz G0 Q6600. I mean the Penryn has twice the L2 cache. Runs at 1333MhZ as compared to 1066MhZ (i know its not a big difference but it will add to the % diff. ). Penryn is an entirely new architecture (I'm no computer hardware specialist but the Penryn does bring state-of-the-art technology to the drawing board with these processors). If the Penryn > Core Q6600 and runs cooler and uses less power and is faster then I would definitely wait till Feb. or march to grab a 9450. Oh and the quad 6600 and the 9450 will be similarly priced with a $10-$20 diff. The 9450 will be $316 and the q6600 is close to the $300 mark.

Also, there are not a lot of q6600's that touch 4Ghz (in fact I've heard of very few that can do that.) There's a good chance that the G0 the OP will get will be able to hit ~3.6Ghz. I know cos I've browsed forums and i can tell you that the number of 4Ghz q6600 are few and far between. I have my hope still for the 9450 as it will bring many improvements over the Conroe based quads. Just my 2cents :)
 
I've now taken a hard look at the Asus Formula (mainly because of what everyone says about its overclocking ability), but I've been hearing rumblings about Ballistix compatibility.

Does anyone know anything more about this supposed Ballistix problem?
 
Given the same clock speeds, a Penryn/Wolfdale/Yorkfield has been benchmarked to be about 10-15% faster than a Merom/Conroe/Kentsfield. Of course, YMMV with real world usage. I expect that a 4ghz G0 Q6600 will outperform a 3.6ghz Penryn / Yorkfield Q9450.

Hi. i was thinking what if the 3.6Ghz Penryn out performs the 4.0Ghz G0 Q6600. I mean the Penryn has twice the L2 cache. Runs at 1333MhZ as compared to 1066MhZ (i know its not a big difference but it will add to the % diff. ). Penryn is an entirely new architecture (I'm no computer hardware specialist but the Penryn does bring state-of-the-art technology to the drawing board with these processors). If the Penryn > Core Q6600 and runs cooler and uses less power and is faster then I would definitely wait till Feb. or march to grab a 9450. Oh and the quad 6600 and the 9450 will be similarly priced with a $10-$20 diff. The 9450 will be $316 and the q6600 is close to the $300 mark.

Also, there are not a lot of q6600's that touch 4Ghz (in fact I've heard of very few that can do that.) There's a good chance that the G0 the OP will get will be able to hit ~3.6Ghz. I know cos I've browsed forums and i can tell you that the number of 4Ghz q6600 are few and far between. I have my hope still for the 9450 as it will bring many improvements over the Conroe based quads. Just my 2cents
Click to expand...

I believe you have a couple of misconceptions here. First, Penryn is not an entirely new architecture. Nehelem will be, but Penryn is no more than a shrunken Core 2. The one other difference is that Penryn adds the SSE4 instruction set, which will only be used with software written specifically for these instructions. I haven't yet seen a benchmark relating specifically to SEE4, but I do believe one or two exist if you search around. Second, remember that Intel prices for chips are wholesale and in batches of 1000. The price at retail will likely be higher. Based on pre-order prices, you can expect a $100 difference between the two chips at launch. Third, the stock bus speed each chip runs at is moot if you are overclocking. A 3.6ghz Q6600 G0 will run on a 1600mhz FSB (400mhz base) or an 1800mhz (450mhz base) FSB if the multiplier is reduced. A 3.6ghz Q9450 will similarly run on an 1800mhz FSB. Lastly, if you look over on XtremeSystems, you'll see many G0 Quads hitting 4ghz on good water systems. Some of the best batches have produced chips that can do 4ghz on nothing more than a $60 Thermalright Ultra 120 Extreme.

4ghz is a distinct and real possibility for a G0. It's not easy, and you may need to exchange around until you get a good, proven overclocker, but it is possible. 4ghz on a Q9450 will in all but the most extreme cases be impossible.

Mark.
 
True, but it's the same architecture for all intensive purposes. The only major differences are the 45-nm die shrink, the addition of SSE 4 instructions, and 1mb more cache per core. The architecture (i.e. the design of the processor) is of course revised a bit, as all are over time -- but again it's not a new architecture like Netburst was or like Core (Conroe) is or how like Nehelem will be.

There are indeed a few other differences between Conroe and Penryn, such as the switch in transistor gate technology, but these are not architectural differences. They are manufacturing differences, and along with the smaller die, allow for cooler operation. Nehelem, with an integrated memory controller and (I believe) four cores on one die, represents a whole new processor design. It's a new architecture. Penryn is technically a die shrink.

I am by no means attempting to debase the hype Penryn has brought. A chip that OCs well past 4ghz while remaining cool on air is an awesome thing. However, I do believe the x8 multiplier on the mainstream chip kills one's hope to reach Penryn's OC heights w/o paying $1000. This is a major bummer, but not too significant in the end since Nehelem is only a year away.

Mark.
 
Markyip1 said:
True, but it's the same architecture for all intensive purposes. The only major differences are the 45-nm die shrink, the addition of SSE 4 instructions, and 1mb more cache per core. The architecture (i.e. the design of the processor) is of course revised a bit, as all are over time -- but again it's not a new architecture like Netburst was or like Core (Conroe) is or how like Nehelem will be.
Click to expand...

I'm not saying it's a new core, but it's more than a straight shrink. Anandtech overviews it quite well.

I do agree with you about the multiplier though. That certainly does limit the overclockability.
 
Hi Mark. thanks for the correction. I agree its not a new architecture (more so revised than new) However I think the OP will find it rather inconvenient to look for a golden o'clocking chip that will hit 4Ghz. I know there are some batches (as with all cpu's) that do clock higher but then there are those that are average and below average. I think expecting a 4Ghz is being optimistic and it's not what one should expect. If I had a G0 I would expect it to do 3.6 Ghz anything above is a bonus. You may get lucky but luck is something unreliable and a rig should just be the opposite.

To the OP:
The way I see it Intel is gonna revise the socket in the next 6-8months. If you can live with a board that will have the soon-to-be-outdated socket aka LGA 775 then by all means build your comp. with the mobo recommendations mentioned in the posts above. I'd like to think that upgrading a mobo is a little more tedious than the cpu IMO. and would like to wait until the next mobo socket is released. Buy a MSI NeoFR2 ($100) and it will keep you happy till the next gen of mobo's comes out. You may go for a Conroe or a Penryn and plant them into the P35 till the new mobo's are released. (obviously the new socket mobo's will be released b4 the cpu's)​
 
You must log in or register to reply here.
Back
Top