Core I9 Gulftown Benchmarks

Ohh Babyyyyy.

http://translate.google.com/transl [...] %26hl%3Den

^ look at the cinebench R10 Render x CPU and the Excel benchmark, the extra threads certainly make a difference.

then add in stuff like FSX, encoding complete HD movies, and apps which can use the extra threads, and you'll see a major difference between the i7 and the i9.

Given a bit of time to mature, and for some re-coding of certain apps like cs4, maya3d and acad, looks like next year may see some upgrades for some of my clients. Guess that means I'll have to build me a new workstation...

I've got my money ready. Gimme Gimme Gimme

Read the link again. The Gulftown's run cooler then the I7's at idle and full load

It's going to go after AMD's niche crowd of megataskers.


But on a serious note, this is pretty frickin' cool for people who are going to do rendering. I like Intel's approach of segmenting their product offerings into the various categories i3/i5/i7/i9.

Would I ever get one? Probably not anytime soon.

I would expect them to run cooler, although the difference between the 2.4 GHz Bloomfield and 2.4 GHz Gulftown was pretty small for both power draw and temperature. The Gulftown is a 32 nm chip but with only about 50% more transistors than the Bloomfield (4 -> 6 cores, 8 MB L3 -> 12 MB L3). Thus it has a smaller die and should run a bit cooler as a result. The only reason why the Gulftown would pull more power at the same clock speed would be if the manufacturing process, CPU, or platform is immature. None of those should be true in this case. Intel's manufacturing arm is absolutely top-notch and they are not under a bunch of pressure to release 32 nm parts early, so the new 32 process should be pretty decent if Intel is starting to release engineering samples. Gulftown doesn't look like much more than a die-shrunk Bloomfield-and-a-half with a few more SIMD instructions, so there shouldn't be too many bugs in the CPU design to work out. The Gulftown drops into current LGA1366/X58 boards, so the platform is plenty mature.

The only bad part about the Gulftown is that it is currently supposed to be a ~$1000 Extreme Edition-only part, if Intel decides to stick to its roadmap. The only 32 nm LGA1366 parts under $1000 would be single-socket Xeons, which carry a bit of a price premium over standard desktop chips. Hopefully Intel will decide to release some less-expensive non-Extreme Edition quad-core and six-core Core-branded chips for LGA1366.

"Bloomfield" was Intel's codename for the desktop Nehalem i7 CPU, 4 cores on 45nm. Gulftown is the codename for the 6-core, 32nm shrink DT part.

However these Westmere's are early engineering samples, and most likely the released parts will be better on power consumption & oc, if history is any guide.

Yeah, but then a dual processor system is going to take it out anyways...

I guess I am more of a dual core high ghz or dual processor system type of person. It just makes more sense this way. Especially when the benchmarks are similar anyway. And who would think of multitasking while playing a game when you need all the performance you can get? More often then not your going to want more ghz and overall features.

Which is why I am currently waiting for the dual core i3/i5. I made no comment on i7, and have said in previous posts it's actually good.

Also, i7 isn't the best thing out there overall. Seriously an e8600 @ 4.4 ghz+ on air cooling isn't impressive?

And to finish this off, I thought we were talking about i9...

^ they're based on the same general architecture so it's fair to compare the i9 and i7.

Only those that were dumb enough to spend $300 on a X58 and triple channel DDR3 would be dumb enough to pay that price for that.
99.9% of users dont even need a quad core. Maybe as a workstation, MAYBE 10% would utilize 6 cores.
Of course THG will be rampant with uber-geeks bragging in their signature that they got one, 2 days after they release. I'll be laughing.
Lube up fellas.

dual vs qaud but ok ok how bout an i7 at 5ghz vs any e8600 at 5ghz lets up the stakes how bout a i7 920 not to mention a 975 at 4ghz vs e8600 at 5ghz

i will upgrade from the 920 i have if and only if reports of a 5 Ghz OC is the norm and a 4.5 OC min with air lol. Regardless of the core count, 4 is more than enough and HT just make it plain who cares.

or IPC rise of doom, then I will consider it (chances are thats a socket change tho)

The i7 is faster than the Core 2 clock-for-clock in just about all applications, so the i7 at 5 GHz will beat an E8600 at 5 GHz. A 4 GHz i7 920 versus a 5 GHz E8600 would be more of a "how multithreaded is this application?" benchmark than anything. The 4 GHz i7 would probably come fairly close to but not beat the 5 GHz E8600 in single-threaded tasks but wipe the floor, walls, and ceiling with the E8600 in multithreaded tasks. But you do have to realize that both a 4 GHz i7 920 and a 5 GHz E8600 are a bit exaggerated as neither chip can OC that high for very long. The E8600 would die from too many volts in not very long and the i7 920 would require The Heatsink of Doom or a very good water-cooling setup to keep it just under the thermal throttling temperature. It's more like a 4.4 GHz E8600 vs. a 3.6 GHz i7 920. The results should still largely remain the same, although the i7 920 will probably be pretty close to the E8600 in single threaded applications in that scenario.

I don't blame him. The E8600 may overclock farther than the Core i7 but it isn't as fast as the Core i7 and lives on a end-of-life platform. Besides, a heavily-overclocked i7 isn't all that far off the pace of a highly-overclocked E8600 in single-threaded applications but murders the E8600 in multithreaded applications.

You'll probably be waiting a while, probably to Ivy Bridge or Sandy Bridge. The 32 nm Westmere-based chips aren't going to be much faster clock-for-clock than the current Nehalems because they're just a die shrink with a few new SIMD instructions. I don't know how well they will overclock, but I'd estimate that the 24/7 OC range would be in the mid-4 GHz range based on the clock speeds of the retail chips (upper 3 GHz range). I wouldn't be surprised if the 22 nm Ivy Bridge (Sandy Bridge) shrink does get to 5 GHz overclocks at the current rate of increase.

Sandy Bridge is a reworked microarchitecture, so you might see some IPC gains there. I wouldn't hold your breath for the "IPC rise of doom" as most of the low-hanging fruit in improving IPC has already been harvested. I'd expect something like the K8 -> K10 or P6+ -> Core -> Nehalem as far as IPC improvement. That is to say respectable (~15% or so) but not outrageous like NetBurst -> Core or K6 -> K7 -> K8. Still, a 15% increase in IPC with a concomitant 1.0-1.5 GHz rise in core speed due to process node shrinks is a fair bump. I have no clue if Intel will replace the sockets for the Sandy Bridge CPUs. I would doubt they'd replace LGA1156 since it would have only been out a year or so, but they've been known to give some sockets a very short lifespan. Socket 423 was the worst as it supported just the P4 Willamettes. Socket 478 was also pretty short-lived as it basically just handled the P4 Northwood. On the Xeon side, Socket 603 was a dud as well as it only handled the Xeon equivalent to the P4 Willamette and 400 MHz FSB Northwood. The mobile CPU sockets churn a bunch as well.

Dude, just read your last paragraph, I am pretty well informed with the new hardware and you almost lost me. That says a lot. And how is a regular peson supposed to deal with all that mess? At least intel has numbering schemes that make sense for the i7/i9 and i3/i5 that are coming out. They should have done something about the 775 cpu's though, became a norm that people asked if the cpu they wanted was going to work with such and such mobo. And the other way around. Not a big deal and most people just tried to deal with cause let's face it, the core duo architecture was almost the best thing that happened for some people.

Yeah, but for the customer point of view, it's really a huge mess.

As for the 775 comment, I agree with you totally, that was what I was trying to get at. Having the 775 from the p4 all the way to a quad core wasn't a good way to implement their progression. But most people never complained cause it performed well enough that they didn't take notice.

Nah.... what power bill???? I'm sure those of us with SLI/CrossFire/QuadFire are already paying this much.

Also E6500K any one?
http://valid.canardpc.com/show_oc.php?id=642872

What you're seeing here is the beginning of a poor 32nm vs a very good 45nm. Forget about huge 5ghz+ overclocks, that's not gonna happen.

Was intel's 45nm really *that* much better than it's 65nm? Not really. For example I'm pretty sure a 65nm Q6600 at 3.6ghz is the match for most Q9xxx.

It's an evolution, not a revolution.

Yep one had HKMG and the other didn't. It didn't do a lot for the quads though for some reason.

Intels 45nm was *very good*, their 32nm isn't such a forward step.

Intels 32nm is just a small evolution on 45nm. There is no quantum leap coming from intel anytime soon, there will be no insane overlocking 5ghz parts or anything like that. What you'll get is small performance increases and slightly better power draw - amazingly enough that's exactly what you're seeing in previews.

If you fold, yes, but unless you game for 12+ hours a day (go and play in the street) you wouldn't be because these renders can take days of constant 100% CPU usage depending on image complexity.

It wasn't really having more cores genius. It was the cache and overall features. Which is why the six core and quad core with the same exact specifications showed no significant gains in a whole series of benchmarks.

+1 PsychoSaysDie

I think enthusiasts will love a 6core CPU. There were no real gains that i see, yes it was faster than the Core i7, but nothing over 10% faster in my opinion. The core i7 still rules the playing field and will for a while.

They should have made their technology evolve a little more: Example; instead of all the cores being active all at once, the CPU should only draw power and activate core according to the type of usage from the operator. I know they remain idle while other cores are functioning with other tasks, but the ones that are idle keep drawing power and waste it. If they were deactivated and reactivated according to usage, the operator could overclock the CPU without having to worry about heat. Also, I think they should have limited the activation of cores to a minimum 2.

Really, i didn't know that.

But, as the old joke about AMDs chips being more prudent at idle awhile back, whats the point of cores that arent running?
My point is, whats this thing going to do IF you can get all 12 running? How well at peak usage is it going to hang in there? Isnt that the whole point of having more cores? And, as enthusiasts, not just having them there when needed, but to kick some arse when used?
We need more data for those scenarios