Sandy Bridge Based Pentiums

[Beraid]

Recently Intel launched their Pentium line of Sandy Bridge cpus. Has anyone here seen any reviews on them? I would especially love to see how efficient the G620T is.

http://ark.intel.com/Compare.aspx?ids=53480,53481,53490,53491,

 

[ShadowVVL]

I looked around but haven't found anything.I would like to see some reviews and benches as well. Maybe we will see some pop up this week or the next.

 

[LOL_Wut_Axel]

The Pentium G850 should be around the same speed as the Core 2 Duo E8600. Meh.

The only case in which I can see these being successful is in... well, nowhere. If you want this for an HTPC or low-power rig, you're better off looking into the Dual-Core Celeron Sandy Bridge. The only real difference between both is lower GPU turbo for Celeron, a bit lower cache (mostly irrelevant), and somewhat lower clock speeds. The Celeron G530 is clocked at 2.4GHz and will be priced at $50, while the Pentium G620 is clocked at 2.6GHz and is priced at $75. Those extra $25 will buy you a 5-10% performance increase, give or take. At that price point, the Athlon II X3 445 is a better choice as it's faster in multi-threaded apps.

 

[pantsaregood]

Intel is making a point to hit every market segment.

Intel's suggested pricing for the G620 is $64. G840/850 are suggested at $75/$85, respectively. I doubt an Athlon II can compete with them in performance - Sandy Bridge is just too far superior to K10. Even an extra core on an Athlon II X3 is likely unable to hold up against any SB-based CPU.

Intel's low-end offerings are a bit unusual, though. Sandy Bridge based Celerons are coming, and they aren't really being distinguished from the Pentiums in any way, save for price. It would've made sense for SB Celerons to be Hyper Threaded single core CPUs, while the Pentium would be a true dual core.

 

[LOL_Wut_Axel]

Intel is making a point to hit every market segment.

Intel's suggested pricing for the G620 is $64. G840/850 are suggested at $75/$85, respectively. I doubt an Athlon II can compete with them in performance - Sandy Bridge is just too far superior to K10. Even an extra core on an Athlon II X3 is likely unable to hold up against any SB-based CPU.

Intel's low-end offerings are a bit unusual, though. Sandy Bridge based Celerons are coming, and they aren't really being distinguished from the Pentiums in any way, save for price. It would've made sense for SB Celerons to be Hyper Threaded single core CPUs, while the Pentium would be a true dual core.

Suggested retail pricing and what we actually see when it filters through the different channels is another thing. The G620 is currently selling for $10 more than MSRP at $75. The Celeron G530 has an MSRP of $45, so I'm expecting actual pricing to be $50-55.

Intel has gimped the instruction sets on the Pentiums and Celerons, not to mention they have no HyperThreading. The performance gains in comparison to Wolfdale and Clarkdale are very minimal, so much that Pentium Clarkdale was actually the exact same performance clock-for-clock as Wolfdale. Based on what I saw on a Chinese review, though, IPC is up around 10% in comparison to Pentium Clarkdale. It lost in multi-threaded apps to the Athlon II X3, but won in single-threaded apps. Multi-threaded apps are the ones that are more time sensitive, though, so I consider them to be more important.

Considering the fact that Celeron and Pentium Sandy Bridge are almost equally gimped, I think the Celeron is a better choice. $25 less and only 5-10% slower overall.

 

[Phynaz]

Intel has gimped the instruction sets on the Pentiums and Celerons

Lol Wut?

 

[Beraid]

When should we expect to see SB Celerons hit store shelves?

 

[LOL_Wut_Axel]

Lol Wut?

They have. Look it up. It's the reason why the Clarkdale Pentiums were the same speed as the Wolfdale Pentiums.

 

[pantsaregood]

They have. Look it up. It's the reason why the Clarkdale Pentiums were the same speed as the Wolfdale Pentiums.

They haven't gimped the instruction sets on SB Pentiums at all. They support SSE 1/2/3/3S/4.1/4.2, XD Bit, EM64T, and even VT-X. They aren't "gimped."

EDIT: Proof.
http://ark.intel.com/Product.aspx?id=53491 (Pentium G850)
http://ark.intel.com/Product.aspx?id=53422 (i3 2100)

 

[Phynaz]

They have. Look it up. It's the reason why the Clarkdale Pentiums were the same speed as the Wolfdale Pentiums.

Outside of virtualization, what exactly are you claiming they did? Made instructions perform slower?

 

[Vesku]

Is there even limited overclocking of these models?

 

[Zap]

The only thing "gimped" would be no hyperthreading and lower amount of cache.

Is there even limited overclocking of these models?

No overclocking.

What I like about the Sandy Bridge Celerons and Pentiums are that the dual core Celeron seems (on paper) pretty nice at hopefully under $50, and the Pentium "T" version should make for really nice super low power draw systems with great performance (worlds better than Atom/Zacate).

 

[ShadowVVL]

I dont think you can oc them same with the i3 2100 or so Ive heard.

 

[LOL_Wut_Axel]

The only thing "gimped" would be no hyperthreading and lower amount of cache.



No overclocking.

What I like about the Sandy Bridge Celerons and Pentiums are that the dual core Celeron seems (on paper) pretty nice at hopefully under $50, and the Pentium "T" version should make for really nice super low power draw systems with great performance (worlds better than Atom/Zacate).

Right... you guys in the thread should read up. The Core i3 2100 is around 50% faster in average in multi-threaded apps. Accounting for the loss of HyperThreading, that's 20% lost. Accounting for the higher clock speeds, that's 15% lost. We still have some 15% unaccounted for.

It is gimped by a huge amount, unfortunately. We're looking at Core 2 Duo E8400 performance, and lower than Athlon II X3 450 performance. The IGP is decent, though, so that's good. If you want something in this price range that's good for the money, get the Celeron G530. It should only be around 10% slower and is just as gimped, while only costing $50-55. You'll have to wait around a month to get it, though.

http://translate.google.com/transla...=http://news.mydrivers.com/1/193/193434_7.htm

 

[ShadowVVL]

what about power consumption chart?

 

[LOL_Wut_Axel]

what about power consumption chart?

System power consumption:

Idle:

Load:

Definitely better than the Athlon II X3 in that regard. Still, since we're talking desktops, it's not as meaningful. For an HTPC or file server, the Pentium makes more sense. For a do-it-all system, the Athlon II X3 makes more sense.

 

[Ken g6]

Right... you guys in the thread should read up. The Core i3 2100 is around 50% faster in average in multi-threaded apps. Accounting for the loss of HyperThreading, that's 20% lost. Accounting for the higher clock speeds, that's 15% lost. We still have some 15% unaccounted for.

Your math is fuzzy.

120%*115%*X = 150%
X = 108.7%

So, 8.7% difference. I wonder what could do that. Maybe the cache difference? :hmm:

 

[pantsaregood]

Intel's site clearly states that the Pentium supports SSE 1/2/3/3S/4.1/4.2, just as the i3 does. The also has an 11% higher clock speed and, from what I've read, only natively supports DDR3 1066, rather than the i3's DDR3 1333.

Drop an i3 2100's multiplier down to 28x and run its RAM at 1066 MHz. Until then, I trust Intel's word.

 

[LOL_Wut_Axel]

Your math is fuzzy.

120%*115%*X = 150%
X = 108.7%

So, 8.7% difference. I wonder what could do that. Maybe the cache difference? :hmm:

What? You should make your results clearer...

The Core i3 2100 is 50% faster. HyperThreading should give it an advantage of 20%. The clock speed is 19% higher, but that's not always completely linear. So 15-19% advantage there. Even taking the best case scenario, that's an 11% difference that's gone unaccounted for. The cache will only make a 1% difference at most if we can take the Core i5 2500K and i7 2600K as examples. Intel gimped it.

EDIT: I'm taking the 2100 as the baseline, not the G620.

 

[LOL_Wut_Axel]

Intel's site clearly states that the Pentium supports SSE 1/2/3/3S/4.1/4.2, just as the i3 does. The also has an 11% higher clock speed and, from what I've read, only natively supports DDR3 1066, rather than the i3's DDR3 1333.

Drop an i3 2100's multiplier down to 28x and run its RAM at 1066 MHz. Until then, I trust Intel's word.

1066MHz 7-7-7-20 to 1333MHz 9-9-9-24 is an under 1% difference. But keep defending Intel all you want...

I already broke it down, and presented factual info.

 

[pantsaregood]

This has nothing to do with defending Intel.

I'm just pointing out that you're making comparisons that aren't 1:1, not to mention that you completely ignore the fact that Intel's site clearly states that it supports the same instruction sets as every other SB.

Also, your assertion that a loss of cache should result in minimal performance loss isn't very scientific. You can throw 12MB of L3 cache onto an SB that is otherwise identical to a 2600K, and you'd probably see under 5% performance gain. Removing cache, however, may result in a huge loss of performance. Performance scales much less linearly with cache than it does with clock speed.

 

[LOL_Wut_Axel]

This has nothing to do with defending Intel.

I'm just pointing out that you're making comparisons that aren't 1:1, not to mention that you completely ignore the fact that Intel's site clearly states that it supports the same instruction sets as every other SB.

The comparisons are completely fair. You're not gonna find a review comparing them both at the same clock speed, but given that the clock speed is 19% higher on the 2100 taking it as the baseline, we can add up to 19% performance for it. There's absolutely nothing wrong with the comparison. A review with them at the same clock speeds will put the 2100 as being 31-35% faster.

Also, your cache argument clearly doesn't hold water in the real world if you look at a Core i7 2600(K) with HyperThreading disabled vs a Core i5 2500(K).

You lose 33% of the cache and less than 1% performance difference. Why would the Pentium with the same amount of lost cache in comparison to the 2100 lose a whole lot more?

 

[ShadowVVL]

The i3 looks better in performance per watt. The Pentium dont looks so good now.

 

[pantsaregood]

The clock speed of the 2100 is not 19% higher. It is 11% higher.

My source is just as valid as yours; Intel clearly states that it supports the same instruction sets.

11% loss of performance to clock speed. 20-30% loss of performance to lack of HT.

Those factors alone place the Pentium at anywhere from 62.3% to 71.2% the performance of the i3. That would average out to 66.75% the performance of the i3, making the i3 almost EXACTLY 50% faster on average.

Now, that's all in an ideal situation. Clock speeds don't scale linearly and HT doesn't offer the 20-30% performance boost that Intel advertises.

Also, the Pentium may lose more performance from cache loss for obvious reasons.

I have 8GB of RAM in my PC. If I pulled out 4GB, I wouldn't see any real difference in performance. If I pulled 2GB from that, I'd take a decent hit to performance. If I took 1GB from that, I'd take a very significant hit to performance.

 

[LOL_Wut_Axel]

The clock speed of the 2100 is not 19% higher. It is 11% higher.

My source is just as valid as yours; Intel clearly states that it supports the same instruction sets.

11% loss of performance to clock speed. 20-30% loss of performance to lack of HT.

Those factors alone place the Pentium at anywhere from 62.3% to 71.2% the performance of the i3. That would average out to 66.75% the performance of the i3, making the i3 almost EXACTLY 50% faster on average.

Now, that's all in an ideal situation. Clock speeds don't scale linearly and HT doesn't offer the 20-30% performance boost that Intel advertises.

HT does not give a 30% performance advantage. Many tests will tell you the same: it's 20%. Now, what you posted simply proves my point. There's a disparity that shouldn't be there. The fact that you put "almost EXACTLY" made me LOL. Come on, you know a difference of 10% or higher is not "almost EXACTLY". Your source says nothing regarding performance numbers themselves.

Also, like I said earlier, if you take the 2100 as the baseline it's clocked 19% higher. If you take the G620, then 11%. Like I also said earlier, clock speed does not scale linearly either, which is why I said 15-19%.

Anyway, I think the point has been made. There's something that's clearly limiting the Pentium's performance, so it's gimped.