Intel Core i5 Benchmarked

[dattimr]

http://www.techpowerup.com/78383/Preliminary_Tests_on_Intel_Core_i5_Conducted.html

i5? i5! Core i5 would be the brand name Intel's mainstream desktop derivatives of the Nehalem architecture based on the Lynnfield core would carry. It is similar to its big brother, the Core i7 for the most of the part except for a few differences:

* A current generation Direct Media Interface (DMI) Interconnect as chipset interface
* A 128-bit wide DDR3 memory interface (Dual Channel) instead of triple-channel
* Some more machinery from the northbridge migrated to the CPU, such as the PCI-Express root complex
* The newer LGA 1160 socket

Lynnfield continues to have four x86 processing cores with HyperThreading enabled, with 256 KB of L2 cache per core and a shared 8 MB L3 cache. Chiphell got its hands on not only the processor, but also a compatible motherboard and run a quick preliminary evaluation of the processor. The processor, clocked at 2,127 MHz, was put though SuperPi, wPrime, Cinebench, Fritz Chess, and 3DMark Vantage. The processor is expected to release in the second half of 2009.

Opinions?

 

[wh3resmycar]

theres a reason why its called "i7".. google it up. its not like its an SKU or something.

 

[dattimr]

Well, I doubt Intel would be stupid enough to create the mayhem of having 2 different sockets for desktop users while using the same branding for both lines of CPUS.

 

[michiganteddybear]

Well, I doubt Intel would be stupid enough to create the mayhem of having 2 different sockets for desktop users while using the same branding for both lines of CPUS.

they have done it before... pentium 4 with both 478 and 775 sockets. even overlapped performance on them too.

yes, they had different part numbers, but they were all branded Pentuim 4

 

[WR]

Well, I doubt Intel would be stupid enough to create the mayhem of having 2 different sockets for desktop users while using the same branding for both lines of CPUS.

they have done it before... pentium 4 with both 478 and 775 sockets. even overlapped performance on them too.

I guess more and more people are building their systems, perhaps out of need for customization and to get more bang for the buck. OEMs don't get confused at identical brand names with different part numbers - neither do enthusiasts. Now they're making it a bit easier for the rest of us to avoid mismatched sockets.

I think they ARE stupid enough. I mean every time they come out with a new socket, it starts out at $300. Why? Because people are stupid enough to pay $300 for a mobo and another $3-500 for a new processor and a couple hundred for 6gb of memory. For what? A 15-20% gain in performance

Intel are stupid, or consumers are stupid?

People have made this argument a long time, but it's ignorant. That 15-20% performance (which varies really between 0% and 50%), you can't get without exotic cooling that would cost even more than the price premium on the i7. A BMW costs 2x a Camry and doesn't drive 2x as fast or have 2x the HP, yet BMW is still very much in business.

Some of us do have yearly IT budgets of $3k or more, which is usually 3-6 weeks of pay. What do you propose, that we add mainstream parts to our systems every month? Run 4 average computers and try to bounce our work among them? Can you imagine the configuration nightmare? IT would envelop our spare time. We wouldn't get to know our hardware well at all.

Those with high computing budgets are a small population. But do you think making 100,000 i7's (as they reportedly did) would be near enough to satisfy the mainstream market? Of course not. They release technology a few months ahead of its time and work with us (developers, end-users, admins) to squeeze out the bugs and optimize performance before an even more massive release. It doesn't matter whether we're after bragging rights, getting a critical task done first in line, or just looking for envy when others with mainstream systems discover they don't have the oomph to catch up. The business model has worked for a long time; ask questions about why instead of calling it impossible.

 

[Pico_w]

Since when are the prices of Qxxxx's dropping like a rock? Q6600 just got $10 more expensive lol.

 

[snarfies1]

I think they ARE stupid enough. I mean every time they come out with a new socket, it starts out at $300. Why? Because people are stupid enough to pay $300 for a mobo and another $3-500 for a new processor and a couple hundred for 6gb of memory.

But wouldn't that make Intel smart? If they can successfully milk the same cash cow over and over, that's downright brilliant.

 

[JAYDEEJOHN]

After seeing those "benches" the other day, Im wondering what all the i7 gamers are going to think when i5 comes along and kills their precious i7s? And for a few hundred $ less?

 

[cjl]

I'm not quite sure how you get that the i5 will "kill" the i7 out of those benches, as my i7 destroys the shown score in every case. In Cinebench 64 bit for example, I get over 22k, while the i5 is shown at 12,523. That same margin of victory is true in basically all of the benchmarks they showed.

 

[WR]

1) i7 is out now, can't buy i5 yet
2) i5 probably can't SLI/Xfire nearly as well. At least with a Q9xxx, you can move from a P35 to an X38/X48 board.
3) i5 isn't out yet; how do we know so much about bclk/multiplier headroom?

 

[JAYDEEJOHN]

Gaming will be better on the i5s. People have already done tests and show i5 as faster. Have you checked XS? http://www.xtremesystems.org/forums/showthread.php?t=210386&page=3

 

[cjl]

How could gaming be faster on the i5 when it is quite literally slower than the i7 in every way? It has slower clocks, the exact same physical cores (hyper threading and all), less memory bandwidth, and fewer available PCI-E lanes. It also has exactly the same cache, though the uncore clock (and therefore the L3 cache clock) is likely to be a bit slower.

I have no doubt that the i5 will be fast, but to say that the i5 will beat the i7 is ludicrous.

 

[JAYDEEJOHN]

It wins in super pi, and is a bit faster in 3DMark, tho within Margin of error. Itll ramp to 2.93 or thereabouts also. Fewer PCI lanes means just less for other things, and itll probably be only for single card solutions, but damned fast at that

 

[WR]

I read through the entire thread about "i5 performance" on XS and they did not test any game. If you have a different thread in mind, can you link it?

On page 3, someone adjusted his i7 920 to mimic the i5's bclk and frequency, but all his benchmarks suggested his CPU was 1-3% behind. The whole socket is different, so is PCI-e, the L3 is unknown, RAM is different, and so forth.

Additionally, mainstream boards may not be as flexible with bclk, and multipliers are lower. That means OC's are unlikely to reach as high.

 

[JAYDEEJOHN]

Not sure which one youre reading either
SuperPi 1M
i5: 19.017s
i7: 19.350s
and here
wPrime
i5 32M: 11.42 sec
i7 32M: 12.013 sec

i5 1024M: 352.199 sec
i7 1024M: 363.435 sec
and like I said, 3DMark was a dead heat. So for huindreds less, to me its a no brainer

 

[dattimr]

How could gaming be faster on the i5 when it is quite literally slower than the i7 in every way? It has slower clocks, the exact same physical cores (hyper threading and all), less memory bandwidth, and fewer available PCI-E lanes. It also has exactly the same cache, though the uncore clock (and therefore the L3 cache clock) is likely to be a bit slower.

I have no doubt that the i5 will be fast, but to say that the i5 will beat the i7 is ludicrous.

Not at all. Remember the PCI Express controller will be inside the CPU silicon, and this will probably reduce the latency quite a bit. Dual vs Tri Channel has been proved useless in desktop apps, so it doesn't matter [now]. They still have lots of time to 'tune it' for a desktop environment.

 

[cjl]

Not sure which one youre reading either
SuperPi 1M
i5: 19.017s
i7: 19.350s
and here
wPrime
i5 32M: 11.42 sec
i7 32M: 12.013 sec

i5 1024M: 352.199 sec
i7 1024M: 363.435 sec
and like I said, 3DMark was a dead heat. So for huindreds less, to me its a no brainer

Here are my scores at 2133MHz:

Wprime 32m: 11.091 sec
Wprime 1024m: 355.524 sec
Superpi 1m: 19.221 sec

Honestly, it seems like pretty much a dead heat to me (as it should be - these are the exact same processing cores working on a CPU limited process). I would say the bigger factor is what the peak clocks are, and the i7 is a higher binned part.

As for the PCI-E latency, I don't see it mattering terribly much. Perhaps I'll be proven wrong, but I seriously doubt that it is limiting framerates.

 

[JAYDEEJOHN]

But isnt top Lynnefield = to 920? in clocks? Thats what Id thought, as I remember am old roadmap showing the top Lynnefield as considered highend

 

[cjl]

It's similarly clocked to the 920, but it's still lower binned silicon. At least that's what I can find so far. Honestly, I wouldn't worry about either's performance in games though.

 

[JAYDEEJOHN]

Heres that roadmap. It looks like 2.93-3.0 Ghz for top Lynnefield

http://pc.watch.impress.co.jp/docs/2008/0602/kaigai01l.gif

 

[cjl]

I'd say that's a fairly safe bet.

 

[spathotan]

So if the i7 and i5 chips are "even" with the Q9550 and what not in gaming, then this truly is a perfect time for AMD to strike in the desktop/gaming market.

I know the i7 is superior to Core 2 in every way pretty much, but these new chips are a good example as to what happens when there is no legit competition for an extended time, stagnation. Gaming performance and basic usage "numbers" from Q6600 to Q9550 to i7 940 is flatlined for the most part. If you have one you have them all.

 

[spathotan]

Also, as somebody mentioned in the feedback to that review in the link, I think having all these sockets is a bad idea. Would have been much easier and much cheaper to just make weaker chips for LGA 1366 instead of this absurd LGA 1160 or whatever

 

[cjl]

It has to be a different socket - it has a physically different connection method. The 1366 has a QPI, while the 1160 does not, but it does have an onboard PCI-express. Because of this, the 1160 CPUs need a physically different motherboard than the 1366.

 

[chookman]

So if the i7 and i5 chips are "even" with the Q9550 and what not in gaming, then this truly is a perfect time for AMD to strike in the desktop/gaming market.

I know the i7 is superior to Core 2 in every way pretty much, but these new chips are a good example as to what happens when there is no legit competition for an extended time, stagnation. Gaming performance and basic usage "numbers" from Q6600 to Q9550 to i7 940 is flatlined for the most part. If you have one you have them all.

http://www.tomshardware.com/reviews/core-i7-gaming,2061-11.html

No increase in gaming huh? Seems its the GPU's letting the i7 down to me... Tri SLI has big gains at 2560x1600 over the QX9770. ~40fps increase in farcry 2 at 2560x1600 is nothing to diss