Sandy Bridge i7 & i5 Reviews

[iTravis]

Impressive gaming performance, look like I'm gonna return the i7 950 along with the memory then. Now I just need a high end mobo, any words about how much it's gonna cost?

[Chrono Detector]

At first I was kinda expecting Sandy Bridge to perform the same as the first generation i7's, but it beats the 980x in half of the tests which is quite impressive but not impressive enough for me to replace my X58/1366 setup, waiting for 2011 or 1356 or whatever is next. But still, its quite a decent product from Intel, the i7 2600K quite a beast of a CPU. I wonder if Intel will drop the 980x price though.

[cpuz]

hi guys,

did you notice that :
http://www.hardware.fr/articles/815-...dy-bridge.html

it is in french, but the graph is clear : under heavy CPU load, the IGP performance is falling down (and drastically).

[Andrea deluxe]

hi guys,

did you notice that :
http://www.hardware.fr/articles/815-...dy-bridge.html

it is in french, but the graph is clear : under heavy CPU load, the IGP performance is falling down (and drastically).

any cpu under load decrease rendering performance of the gpu....

[-Sweeper_]

very impressive performance and power efficiency on a ~210mm² die, well done intel

now bring us SB-E and IB

[Mumak]

This is the standard Turbo behavior, since the TDP limit is shared among IA and GT Cores.

hi guys,

did you notice that :
http://www.hardware.fr/articles/815-...dy-bridge.html

it is in french, but the graph is clear : under heavy CPU load, the IGP performance is falling down (and drastically).

[cpuz]

any cpu under load decrease rendering performance of the gpu....

True, but not in such a range.
The test of the 2600K is very clear :
As the load of the CPU inscreases, the FPS in HAWK drops from 80 to 20 FPS (divided by 4).
With a discrete GPU (HD5450), the frame rate is 68 whatever the CPU load is.

I know this can be easily explained by TDP sharing, nethertheless, it is a real issue, IMO.

[xVeinx]

It also could be due to sharing of the L3 cache. With too much going on, the GPU may become cache starved I would imagine...

[zalbard]

K-series do not support neither TXT nor VT-d

damn, that sucks. bad for people who were looking to build some cheap workstations

That sucks. One of the reasons that puts me off...

Anand mentions 4 and 6 cores on LGA2011 in Q4, without IGP. No 8 cores though.

I was surprised by this too. Maybe Anand made a typo there? Of course,there is a possibility that intel is going for maximum of 6 cores with SB(power limited) and 8 cores with IB refresh.

I think we will see 4 and 6 cores for consumer market, and 8 and 12 cores for servers.

True, but not in such a range.
The test of the 2600K is very clear :
As the load of the CPU inscreases, the FPS in HAWK drops from 80 to 20 FPS (divided by 4).
With a discrete GPU (HD5450), the frame rate is 68 whatever the CPU load is.

I know this can be easily explained by TDP sharing, nethertheless, it is a real issue, IMO.

Yes, it is. Not sure anything can be done, though. If it's TDP, then perhaps there will be some BIOS option to turn this off... If it's caches and memory bandwidth being starved then nothing can be done most likely.

[Olivon]

Really powerfull CPU's, cruncher's dream

[Vozer]

It seems my guess about Sandy Bridge prices was right.

The biggest loser in the overclock locks is the Core i3 which now ships completely locked. Thankfully AMD has taken care of the low-end segments very well over the past couple of years. All Intel is doing by enforcing clock locks for these lower end chips is sending potential customers AMD's way.

Now all they need to do is release a Core i3 2120K for ~150 USD and the lower-end chip customers will be happy.

[informal]

At first I was kinda expecting Sandy Bridge to perform the same as the first generation i7's, but it beats the 980x in half of the tests which is quite impressive but not impressive enough for me to replace my X58/1366 setup, waiting for 2011 or 1356 or whatever is next. But still, its quite a decent product from Intel, the i7 2600K quite a beast of a CPU. I wonder if Intel will drop the 980x price though.

2600K beats 980x in some tests because consumer applications are not well threaded(past 4 or max 8 threads). SB does have,apart from ~10% better IPC, higher turbo modes that engage even if all 4 cores are 100% loaded so there is one more performance feature added,which is very good(TDP of underutilized GPU logic is enabling this feature).
As AT mentioned,there will be 4 and 6 core chips on 2011 socket so you will get all SB benefits with more cores and more cache.

hi guys,

did you notice that :
http://www.hardware.fr/articles/815-...dy-bridge.html

it is in french, but the graph is clear : under heavy CPU load, the IGP performance is falling down (and drastically).

Like someone mentioned if L3 is utilized by CPU cores a lot then GPU might be getting less of it so there is a potential GPU bottleneck.

[Tom128]

My AM3 system with my 1090TBE tells me I don't need to upgrade, I can do anything any everything I need to do. But yet, I have this urge that tells me my AM3 system is wrong and I could do anything and everything more betterer. Me want.

[ap4lifetn]

is there any review that does a clock for clock comparison?

i run my 920 @ 3.8ghz daily and want to see if i see any huge gains

its been too long since i upgraded! need to find out if i should bite now or wait for SB-E

[Movieman]

Really powerfull CPU's, cruncher's dream

Yes, and especially when you see the electrical usage essentially halved vs a i7-920 at the same speeds and output of a 2600 non K in WCG at 3492 just outdoes a i7-920 at 4GHz..

[Mechromancer]

The old Intel would have charged $2000 for the Core i7-2600K. I'm glad Intel woke up and realized folks can't spend house payments on CPUs. This pricing can really stomp a hole in AMD if all does not go well for Zambezi. Honestly, I think 8 real cores will end up being very competitive with Intel's 4+HT cores. Lets just hope AMD can sell an 8 core CPU for $350 or less.

SB is truly a great CPU! Congrats to Dr. Who?

[kl0012]

is there any review that does a clock for clock comparison?

i run my 920 @ 3.8ghz daily and want to see if i see any huge gains

its been too long since i upgraded!

http://www.computerbase.de/artikel/p...formancerating
http://www.hardware.fr/articles/815-...dy-bridge.html

[C-N]

Maybe I shouldn't post this as I'm sure I'm going to get a bit of stick but here goes.

I'm have a similar CPU/GPU issue atm on my Giga P67 & 2500k....

Haven't investigated it much so could be all sorts of problems like unstable clock, not enough vcore lack of the magic internal pll fix etc but here it is & I thought it would be worth mentioning.

As always 3DMARK03 & 05 CPU tests require a tad more vcore than any other test I run Prime, LinX etc.... nothing new there. 03 is constant with a bit more vcore but 05 only passes 1/2 the time & out of those times rarely achieves a decent score. I can consistently achieve a similar score at lower CPU speed like 200MHz or so.

It probably doesn't help I am still using an old 3870X2 as the dual core cards are known to give the CPU a good pasting but this is something I didn't expect. It looks to me crossfire or even quadfire will be tricky at max CPU speed with the 2500.... maybe the 2600 with its larger L3 cache will do better but we'll have to see.

Your thoughts please.....

#### Edit #####

Gave up & put a HD6950 in & all is well, looks like my setup dont like the old 3870x2!!!

.

[kl0012]

one more review (SB as HTPC cpu)
http://www.missingremote.com/review/...bl-motherboard

[informal]

http://www.computerbase.de/artikel/p...formancerating
http://www.hardware.fr/articles/815-...dy-bridge.html

The hardware.fr link is pretty interesting since they shown the exact gain that HT bring on both Lynnfield and SB. With and without HT the gap is around ~13.2 and 11.3% at 2.8Ghz.As can be seen, with HT it narrows down quite a bit in few tests while it jumps up in 1 or two totaling in lower advantage for SB with HT ON. I heard that HT should have been improved considerably with SB since it has much more potent memory subsystems(much higher L/S bandwidth) ,but in tests they ran it is not showing at all. I suppose that when AVX optimized software arrives it will show up in tests as a good performance uplift.

[ap4lifetn]

http://www.computerbase.de/artikel/p...formancerating
http://www.hardware.fr/articles/815-...dy-bridge.html

english please?

or maybe a quick summary..

is it <20% clock for clock?

[Clunk]

P8P67 Deluxe Review Here!

[kl0012]

english please?

or maybe a quick summary..

is it <20% clock for clock?

I don't speak french or german as well, but you actually don't need it to understand the charts. The clock for clock perf diff varies from 5% to 20% depends on application.

[Hornet331]

True, but not in such a range.
The test of the 2600K is very clear :
As the load of the CPU inscreases, the FPS in HAWK drops from 80 to 20 FPS (divided by 4).
With a discrete GPU (HD5450), the frame rate is 68 whatever the CPU load is.

I know this can be easily explained by TDP sharing, nethertheless, it is a real issue, IMO.

Na its the same for dedicated gpus, when i run boinc in backgorund while I play TF2, my avg fps is a lot lower then when I stop boinc. And thats with a hexacore and a HD5870.

[informal]

Na its the same for dedicated gpus, when i run boinc in backgorund while I play TF2, my avg fps is a lot lower then when I stop boinc. And thats with a hexacore and a HD5870.

How many cores can TF2 utilize ? If you limit boinc to 4 cores and TF2 uses up to 2 cores then you shouldn't see much of a drop off.