Yorkfield vs. sandy bridge

auntarie

Distinguished
Feb 27, 2012
737
0
19,060
Hello,

I was wondering if someone could shed some light on why Sandy Bridge outperforms Yorkfield by so much. I know there are the beasts like 2500k, 2600k and even 3930k, but I'm not talking about them. I'm talking about a simple dual-core i5 mobile processor @ 2.5 GHz that was listed as 8% better than my Q9300. Or an i7-2630QM @ 2.0 listed as 78% better. I can understand the i7 case because it has hyper-threading and turbo boost, but what about the i5? Do people take HT and turbo boost into account in benchmarks or is it just that the chip is so much better?

Thanks in advance.
 
Solution


Yorkfield was an amazing chip, I still have 3 computers based on them and a 2 based on Kentsfield.

There are far too many differences to list them all...



it depends on benchmarks you are using, some prefer real time benchmarks(moving file speed, encoding time, rendering speed etc) others just use a passmark or passmark like tool(can be heavily inaccurate sometimes) to compare processors.
 
Updating from Yorkfield to Sandy Bridge can be beneficial depending on what you plan on doing.

I encoded a 1 hour and 44 minute video using the XviD codec (2-pass method) on both my desktop rig (Q9450 @ 3.0GHz OC'ed) and my laptop (i5-2410m @ 2.9GHz w/ Turbo Boost). The Q9450 to 55 minutes to complete the encode while the i5-2410 took 39 minutes which represents a 30% improvement in performance.

The XviD codec only uses one core.
 


Yorkfield was an amazing chip, I still have 3 computers based on them and a 2 based on Kentsfield.

There are far too many differences to list them all but I'll name the big ones

1. Kentsfield and Yorkfield are native dual-core dies. Each Kentsfield and Yorkfield processor was actually 2 Conroe or Wolfdale chips glued together. This means that each 2 processor block shared half of the total L2 cache and communication between the two processor blocks was problematic. Nehalem, Westmere, Sandybridge and Ivybridge are native quad and octal core processors (the 3930k and 3960x have two cores and 5-8MiB of L3 cache disabled for yield purposes) with unified cache. This improves cache access and interprocessor communication

2. Beginning with Nehalem the memory controller was moved from the northbridge onto the CPU itself. This reduced the nasty bottleneck that could be caused by the Front Side Bus, lowered latency by by necessitating only a single bus transfer rather than two, and greatly expanded the memory capacity in multi-socket systems.

3. Beginning with the mid range Nehalem processors (800 series and below) PCIe lanes were made available directly from the CPU. The first generation i7s still pulled them out from the northbridge only. Subsequently all Sandybridge and Ivybridge processors made 16 PCIe lanes available directly from the CPU and an additional 8 available with the standard 6 and 7 series chipsets. The on-chip PCIe lanes reduce access latency.

4. Sandybridge also included a ton of front-end improvements to the branch prediction and decoded instruction cache. These improvments also allowed for Hyperthreading to be reintroduced which allowed for unused back-end resources to be shared with different front-end threads. This improves performance by reducing idle time and causes confusion in those who don't know the difference between a thread and a core.
 
Solution
G

Guest

Guest
I agree with amuffin, but if you think is better to move to sandy now go with i7 3820 or i7 3930K to LGA 2011 motherboards this way you get better chance to upgrade in feature incoming processors.
 

auntarie

Distinguished
Feb 27, 2012
737
0
19,060
Thanks for the replies guys. I guess I should have mentioned in my first post- I'm not planning to upgrade. I'm only using my rig for gaming and the Q9300 is tearing through any game I throw at it. I was just wondering why the benchmarks were showing such a superiority for Sandy over Yorkfield.

it depends on benchmarks you are using, some prefer real time benchmarks(moving file speed, encoding time, rendering speed etc) others just use a passmark or passmark like tool(can be heavily inaccurate sometimes) to compare processors.

I was looking at the benchies on cpubenchmark.net and I have a profile over at game-debate where each user has their components listed and it automatically compares them. I think according to passmark, since both sites returned pretty much the same results.

Updating from Yorkfield to Sandy Bridge can be beneficial depending on what you plan on doing.

I encoded a 1 hour and 44 minute video using the XviD codec (2-pass method) on both my desktop rig (Q9450 @ 3.0GHz OC'ed) and my laptop (i5-2410m @ 2.9GHz w/ Turbo Boost). The Q9450 to 55 minutes to complete the encode while the i5-2410 took 39 minutes which represents a 30% improvement in performance.

The XviD codec only uses one core.

That might be because of Quick Sync. It really helps with conversion of videos.

@Pinhedd: Thanks mate, that was just what I wanted to know :)
 

Pailin

Distinguished
Dec 1, 2007
851
1
19,015
I just happen to have been comparing these CPU's as I am just now in the middle of Finally upgrading my Q6600
- and then only as I can resell old parts for decent price right now - and will appreciate the extra folding@home PPD on my new Xeon X5650
(oh, and possible money saved on electricity bills ;) <--- I really like this upgrade justification lol )


Intel Core i7-970 (Gulftown 6c/12t, DDR3-1066, 1 MB L2, 8 MB L3) @ 3.2 GHz (Turbo 3.46 GHz)
Intel Core i7-940 (Bloomfield 4c/8t) @ 2.93 GHz (Turbo 3.2 GHz), DDR3-1066, 1 MB L2, 8 MB L3
Intel Core i5-2300 (Sandy Bridge 4c, DDR3-1333, 1 MB L2, 8 MB L3) @ 2.8 GHz (Turbo 3.1 GHz)
Intel Core 2 Extreme QX9770 (Yorkfield 4c, DDR3-1333, 12 MB L2) @ 3.2 GHz
Intel Core 2 Extreme QX6850 (Kentsfield 4c, DDR3-1333, 8 MB L2) @ 3 GHz

Toms CPU Comparison
(had to use a short url converter as the real link breaks the forums url hotlink due to having [ ] brackets in it -.- )

more or less tried to keep CPU speeds the same - and threw in the 6core/12thread CPU closest to mine too
There are some really nice speed increases going from generation to generation even keeping Clocks pretty similar.

For most games there is not so much call to upgrade the CPU I think as long as your CPU can push your GFx to a min framerate of 60+ fps (maybe 120 for some 3D setups) . I am sure some games out there might like that bit more, but when you play such a game and feel the limitations - Then upgrade ^^
 
I keep wondering what it will take before my i7-930 OCed to 3.9Ghz will make me want to trade her in. Maybe Haswell will do it - hard to say. To the OP - most games rely on gpu performance as long as you have a decent multi core cpu.
-Bruce
 

auntarie

Distinguished
Feb 27, 2012
737
0
19,060


Yeah, I know. A lot of people say that. That's why I'm not planning on upgrading. My GPU is more than up to the task.
 


I am not using Quick Sync. I simply use the XviD codec and VirtualDub 1.8.8. The increase in performance is solely due to IPC (Instructions Per Cycle) improvements. VirtualDub 1.8.8 is not designed to make use of Quick Sync since that program was released back in 2009.
 

TRENDING THREADS