| View previous topic :: View next topic |
| Author |
Message |
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
 Posted: Thu Mar 02, 2023 11:29 am Post subject: Nehalem/Bloomfield i7 vs i7-Extreme |
 |
|
So, inspired by John's recent i7-975x sale, https://www.cpu-world.com/forum/viewtopic.php?t=38036 i got to think of it, that the first generation Extemes were merely 6.4 GT/s vs. 4.8 GT/s speed on the QPI interface (and as far as I can tell, maybe also unlocked multiplier?).
But how much difference does the faster QPI make? I can't find any reviews, could anyone chime in with something? Ideally, a head to head between i7-960 and i7-965x would be cool  |
|
| Back to top |
|
 |
svmlegacy
Joined: 15 Jun 2016
Posts: 551
Location: Canada
|
| Posted: Thu Mar 02, 2023 6:16 pm Post subject: |
|
|
Can confirm the extreme editions had an unlocked multiplier (a bit moot considering the BCLK was freely adjustable, but it was there, and allowed for more refined OC'ing)
Can't speak much to the QPI clock differences at this time. |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Fri Mar 03, 2023 1:31 am Post subject: |
|
|
Yeah, since there's no FSB, I imagine that overclocking BCLK would be pretty "safe" no matter.
As for what I can see, the faster QPI link should only matter in regards to peripheral speeds, i.e. graphics, I/O etc., and where it would really shine is in multi socket platforms ... |
|
| Back to top |
|
|
Calbris
Joined: 06 Feb 2019
Posts: 157
Location: Singapore
|
| Posted: Fri Mar 03, 2023 7:19 am Post subject: Re: Nehalem/Bloomfield i7 vs i7-Extreme |
|
|
| H3nrik V! wrote: | | But how much difference does the faster QPI make? |
Have a look at this old review from 2008: https://bit-tech.net/reviews/tech/cpus/intel-core-i7-920-945-965-review/4/
| H3nrik V! wrote: | | As for what I can see, the faster QPI link should only matter in regards to peripheral speeds, i.e. graphics, I/O etc., and where it would really shine is in multi socket platforms ... |
Assuming this is an X58 platform and not a P55 platform, yes. A faster QPI link would only benefit devices that are directly connected to the 82X58 hub, however. Anything that is connected to the ICH10(R) will not as they are linked to the 82X58 via a separate link, you'd still be bottlenecked by a slow DMI link. The gains would be very minimal at best.
 |
|
| Back to top |
|
|
svmlegacy
Joined: 15 Jun 2016
Posts: 551
Location: Canada
|
| Posted: Fri Mar 03, 2023 8:19 am Post subject: |
|
|
I'd think the QPI speed is much more important on dual socket LGA1366 systems.
Again, X58 is saved by its flexible BCLK, when OC'ing a processor in this way, the QPI speed can be raised.
P55 / LGA1156 doesn't use QPI at all, instead uses DMI. |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Fri Mar 03, 2023 8:47 am Post subject: Re: Nehalem/Bloomfield i7 vs i7-Extreme |
|
|
| Calbris wrote: | | H3nrik V! wrote: | | But how much difference does the faster QPI make? |
Have a look at this old review from 2008: https://bit-tech.net/reviews/tech/cpus/intel-core-i7-920-945-965-review/4/
| H3nrik V! wrote: | | As for what I can see, the faster QPI link should only matter in regards to peripheral speeds, i.e. graphics, I/O etc., and where it would really shine is in multi socket platforms ... |
Assuming this is an X58 platform and not a P55 platform, yes. A faster QPI link would only benefit devices that are directly connected to the 82X58 hub, however. Anything that is connected to the ICH10(R) will not as they are linked to the 82X58 via a separate link, you'd still be bottlenecked by a slow DMI link. The gains would be very minimal at best.
|
Yes, it is X58.
It seems that faster QPI could remove a bit bottlenecks regarding PCI-Express bus. |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Fri Mar 03, 2023 8:47 am Post subject: |
|
|
| svmlegacy wrote: | I'd think the QPI speed is much more important on dual socket LGA1366 systems.
Again, X58 is saved by its flexible BCLK, when OC'ing a processor in this way, the QPI speed can be raised.
P55 / LGA1156 doesn't use QPI at all, instead uses DMI. |
Yes, from what I could read, I also got the understanding that faster QPI would be most valuable for multi socket systems. |
|
| Back to top |
|
|
Calbris
Joined: 06 Feb 2019
Posts: 157
Location: Singapore
|
| Posted: Fri Mar 03, 2023 10:22 am Post subject: Re: Nehalem/Bloomfield i7 vs i7-Extreme |
|
|
| svmlegacy wrote: | | P55 / LGA1156 doesn't use QPI at all, instead uses DMI. |
I know this is a little bit off-topic, but wasn't Clarkdale an LGA1156-compatible architecture that utilized QPI to communicate from its memory controller + graphics core to the processor cores? My memory's rusty on this part. |
|
| Back to top |
|
|
svmlegacy
Joined: 15 Jun 2016
Posts: 551
Location: Canada
|
| Posted: Fri Mar 03, 2023 10:25 am Post subject: Re: Nehalem/Bloomfield i7 vs i7-Extreme |
|
|
| Calbris wrote: | | svmlegacy wrote: | | P55 / LGA1156 doesn't use QPI at all, instead uses DMI. |
I know this is a little bit off-topic, but wasn't Clarkdale an LGA1156-compatible architecture that utilized QPI to communicate from its memory controller + graphics core to the processor cores? My memory's rusty on this part. |
You are correct, forgot about this little technicality.
 |
|
| Back to top |
|
|
crazybubba64
Joined: 03 Jul 2018
Posts: 1371
Location: WI, USA
|
| Posted: Fri Mar 03, 2023 11:20 am Post subject: |
|
|
Might be easier to do some comparisons using the more widely-available Xeon analogs for the i7 parts. (heck of a lot cheaper too)
These Xeon parts even have unlocked multipliers, just like their i7 counterparts. The only difference I can think of between the two is ECC memory support on the Xeons.
Core i7-965 SLBCJ = Xeon W3570 SLBES
Core i7-975 SLBEQ = Xeon W3580 SLBET
Core i7-980X SLBUZ = Xeon W3680 SLBV2
Core i7-990X SLBVZ = Xeon W3690 SLBW2
The theoretical i7-995X does not have any known retail Xeon counterpart.
_________________
My collection |
|
| Back to top |
|
|
xsecret
Joined: 01 Feb 2004
Posts: 1846
Location: France
|
| Posted: Sun Mar 05, 2023 5:34 pm Post subject: |
|
|
| crazybubba64 wrote: |
Core i7-965 SLBCJ = Xeon W3570 SLBES
Core i7-975 SLBEQ = Xeon W3580 SLBET
Core i7-980X SLBUZ = Xeon W3680 SLBV2
Core i7-990X SLBVZ = Xeon W3690 SLBW2
The theoretical i7-995X does not have any known retail Xeon counterpart. |
That's correct. To be sure, you just have to check their ordering code. If the digits are the same, the validation process is the same (then specific features are fused-down in the last step to make them a Core or a Xeon).
Here we have :
Core i7-965 SLBCJ (AT80601000918AA) = Xeon W3570 SLBES (AT80601000918AB)
Core i7-975 SLBEQ (AT80601002274AA) = Xeon W3580 SLBET (AT80601002274AB)
Core i7-980X SLBUZ (AT80613003543AE) = Xeon W3680 SLBV2 (AT80613003543AF)
Core i7-990X SLBVZ (AT80613005931AA) = Xeon W3690 SLBW2 (AT80613005931AB)
_________________
ES-Only Collector : http://www.engineering-sample.com
Universal Chip Analyzer (UCA) : https://x86.fr/uca / http://www.cpu-world.com/forum/viewtopic.php?t=34349 |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
| Posted: Mon Mar 06, 2023 4:45 pm Post subject: |
|
|
Thanks for the input ... Better get me some Xeons, though it's kind of cheating  |
|
| Back to top |
|
|
H3nrik V!
Joined: 15 Apr 2014
Posts: 1246
Location: Denmark
|
|
| Back to top |
|
|
Guest
|
| Posted: Tue Mar 05, 2024 3:34 pm Post subject: |
|
|
| crazybubba64 wrote: | Might be easier to do some comparisons using the more widely-available Xeon analogs for the i7 parts. (heck of a lot cheaper too)
These Xeon parts even have unlocked multipliers, just like their i7 counterparts. The only difference I can think of between the two is ECC memory support on the Xeons.
Core i7-965 SLBCJ = Xeon W3570 SLBES
Core i7-975 SLBEQ = Xeon W3580 SLBET
Core i7-980X SLBUZ = Xeon W3680 SLBV2
Core i7-990X SLBVZ = Xeon W3690 SLBW2
The theoretical i7-995X does not have any known retail Xeon counterpart. |
hi, would the i7-995X have a counterpart with Xeon X5698 SLC32 |
|
| Back to top |
|
|
crazybubba64
Joined: 03 Jul 2018
Posts: 1371
Location: WI, USA
|
| Posted: Tue Mar 05, 2024 10:15 pm Post subject: |
|
|
| Anonymous wrote: | | hi, would the i7-995X have a counterpart with Xeon X5698 SLC32 |
X5698 is a 4.4GHz dual-core, the theoretical (ie, very little concrete evidence they exist) i7-995X is a six-core part with a 3.6GHz base clock IIRC.
_________________
My collection |
|
| Back to top |
|
|
|
