weak quad vs strong dual

[621ogob]

Hi Guys,

In your opinion, what's better in terms of gaming (new games): 2 ghz quad sandy bridge vs 3 ghz dual core with hyperthreading sandy bridge ? No turbo boost or oc allowed.

Thanks in advance,
bogo126

 

[T_Yamamoto]

Quad.

Some games are able to utilize 4 cores (IE: Battlefield games)

 

[LurchFrinky]

I would expect the specific games you had in mind would make a pretty big difference.
But seeing as your dual core has hyperthreading and a 50% speed advantage, I would guess this would be the better choice.

 

[Exophase]

Assuming the former has no HT, I'd go with the latter. You'd already get most of the performance in a lot of very well threaded loads due to the combination of clock speed and HT (1.5 * 1.3). And a lot of software isn't very well threaded.

 

[2is]

Back when this discussion was between the E8400 vs Q6600 the decision was pretty clear (for me) that the Q6600 was the better choice. E8400 only had a 25% clock speed advantage but NO hyper threading. In this case, we are talking about a 50% speed advantage for for dual core, and it also has the benefit of HT so the performance hit from >2 threaded games isn't as big as a dual core without HT.

 

[Jacky60]

This would be a silly question two years ago but today's games need a quad and the faster the better. Quad and QFT:whiste:

 

[621ogob]

To be specific, I want you to help me choose between i7-2620m (2c 4t 3,2 ghz) and i7-2630qm (4c 8t 2 ghz (turbo disabled)). Which is better for crysis 2/3, most wanted 2012, skyrim, grid 2 etc. ?

 

[Ventanni]

Why would the turbo be disabled for the 2630qm? It turbos up to 2.8ghz in single thread applications.

 

[JimmiG]

Why would you disable Turbo? It's extremely useful on mobile quad-core CPUs which are limited to low clock speeds when all four cores are active. With the i7-2630qm you get the best of both worlds - Four full cores when you need them and up to 2.9 GHz in lightly threaded applications.

 

[crashtech]

Yeah, the quad with turbo is a no-brainer. Without, it's a lot harder to say.

 

[Yuriman]

As far as raw computing power goes, here's some simple math:

2Ghz x 4 cores = 8Ghz-core effective total
3Ghz x 2 cores x 1.3 = 7.8Ghz-core effective total
(1.3 being arbitrarily chosen best-case performance improvement from hyperthreading)

In apps that utilize all 4 cores and extracts a 30% performance boost from the two logical cores in the dual, a 2Ghz quad is still marginally faster.

As others have already said though, having a turbo on the quad allowing it to clock close to as high as the dual makes it a no-brainer in favor of the quad.

 

[2is]

As far as raw computing power goes, here's some simple math:

2Ghz x 4 cores = 8Ghz-core effective total
3Ghz x 2 cores x 1.3 = 7.8Ghz-core effective total
(1.3 being arbitrarily chosen best-case performance improvement from hyperthreading)

In apps that utilize all 4 cores and extracts a 30% performance boost from the two logical cores in the dual, a 2Ghz quad is still marginally faster.

As others have already said though, having a turbo on the quad allowing it to clock close to as high as the dual makes it a no-brainer in favor of the quad.

No dude, I can't believe people still add up the GHz. It does not at all work that way

 

[Blitzvogel]

No dude, I can't believe people still add up the GHz. It does not at all work that way

Across similar architectures, I don't see too much of a problem. It's not like we're going to be able to instantly know what kind of difference a larger L2/L3 cache and the effects of main memory bandwidth per core are going to be like.

 

[2is]

Across similar architectures, I don't see too much of a problem. It's not like we're going to be able to instantly know what kind of difference a larger L2/L3 cache and the effects of main memory bandwidth per core are going to be like.

It's not about architecture. Even multi-threaded software does not scale linearly with cores. Some scale well, some scale poorly, some scale only up to x number of threads and some don't scale at all. The effective clock speed of a 2GHz quad core is 2GHz. Period.

 

[Yuriman]

No dude, I can't believe people still add up the GHz. It does not at all work that way

I'm aware things don't scale perfectly linearly with clock, and most programs don't load cores even close to evenly, but it's more useful to guesstimate this way than to say "the quad will be faster because my brother has a quad and it's pretty fast."

 

[2is]

Sorry, but I don't see how it's useful when it doesn't at all work that way. In fact, it almost never does.

 

[crashtech]

This is a hard question to answer because there isn't a Sandy quad that is locked to such a low frequency. There's nothing to point to to say, "it will perform something like x."

 

[Yuriman]

No dude, I can't believe people still add up the GHz. It does not at all work that way

I'm aware things don't scale perfectly linearly with clock, and most programs don't load cores even close to evenly, but it's more useful to guesstimate this way than to say "the quad will be faster because my brother has a quad and it's pretty fast."

 

[2is]

I understand what you're saying, but I don't agree. Its not better when it's a completely inaccurate method of gauging performance. I'd take anecdotal evidence over adding up ghz any day.

 

[Concillian]

This is a hard question to answer because there isn't a Sandy quad that is locked to such a low frequency. There's nothing to point to to say, "it will perform something like x."

It's not a hard question to answer, as the example he gave is very specific. In HIS EXAMPLE, the dual would be clearly superior in games that were not well threaded and roughly equivalent in games that were well threaded. So dual with HT is the answer.

In real life, sure, there's more doubt, as the difference is not as stark. Absolutely. But there's very little doubt for the specific scenario asked about in the OP.

 

[crashtech]

...roughly equivalent in games that were well threaded...

This is the part that seems like it must be pure conjecture. If there is concrete evidence on which to base your assertion, please correct me, though. That's why I think it's not so easy to say, not without actual testing.

btw, I tend to agree with you, it's just that I like to have some real-world example to point to, and say, "this is why."

 

[Concillian]

This is the part that seems like it must be pure conjecture. If there is concrete evidence on which to base your assertion, please correct me, though. That's why I think it's not so easy to say, not without actual testing.

Well threaded applications, by definition, can "add cores" together to get their relative computing power, so in these cases you will see 90+% of the quad core capability from a 2C/4T CPU that has a 50% speed advantage. Even games that take advantage of many threads favor 2 more heavily, so there is a bit of an x-factor. The worse a game handles multiple cores the more advantage a 4T CPU that more heavily weights 2 cores will fare.

Of course there will always be specific cases of software that is better on one or the other, but if having to make a decision based on the average performance, I'd pick the 2C/4T. The advantage of 3 GHz vs. 2 GHz in the games that still aren't well threaded (+50%) far outweighs the ~10% advantage the quad will have in games that are well threaded. Since games MUST design around a lowest common denominator CPU, you also see well threaded games being a little less demanding on the CPU in general to get decent frame rates (minimums in the 40+ range).

It's a matter of what's the better balanced, and 2C/4T is more balanced than 4C / 4T.... when 2C/4T is given a 50% speed advantage. The worse an application handles threading, the more it will benefit a 2C/4T CPU.

Now in the real world, where it's not a 50% speed advantage for the 2C/4T, sure... the water is a LOT more muddied, but I think 50% speed advantage is enough to bring it into a clear lead.

If you ask me, if i3s were still OCable and in the $130 range, we'd see MANY gamers using 2C4T Ivy and Haswells at 4.5+GHz and handling modern gaming just fine. There is a bit of a threshold effect to games... there comes a point where a CPU does get "good enough" to be virtually perfect and I think an Ivy or Haswell 2C/4T at slightly higher than typical quad OCs (due to additional thermal margin) would reach that plateau for all but the least budget conscious gamers.

 

[BrightCandle]

I have captured all the CPU information from gamegpu.ru. Comparing a 2500k to an i3 2100 on average the quad core achieves just 29% better framerates. Sometimes the 2500k is slower by abouut 15% but 25% of games run between 43% and 135% faster.

So we are clearly in a world where quad cores do bring some benefit, but its not "massive" in the sense that a dual core is still achieving playable frame rates in many games. But not all. Now this is with comparable clock speeds so if you take the quad core and rip away 50% of its clock speed its unlikely to compete in todays games well compared to the dual core. But there will be circumstances where it does better.

 

[bbhaag]

Fanboyism and Brand Loyalty

Interesting read and a real insight into the human psyche. It makes you wonder if Brand X really is better than Brand Y or is it the emotional attachment to the object that really forms our opinions.

Sorry for the OT post but your sig link piqued my interest.

 

[Enigmoid]

Get the dual core + HT. Guys lets not forget that he want's a laptop for gaming and that said laptop in question is unlikely to be held back by effectively a desktop sandy i3. On the other hand 2 ghz may not be quite enough for some games (SC2, Skyrim for example).

i7-2620m (2c 4t 3,2 ghz) and i7-2630qm (4c 8t 2 ghz (turbo disabled)). Which is better for crysis 2/3, most wanted 2012, skyrim, grid 2 etc. ?

And if you actually look at his posts (which no one seems to be doing) you clearly see that its a dual @ 3ghz + HT and a quad @ 2 ghz + HT. Both have HT.

For your listed games, Crysis 2 and Skyrim will do well on the dual while crysis 3 may do slightly better on the quad. No clue about most wanted or grid 2 but I doubt with a mid range gpu of the sandy bridge time period (up to 560m) you will be held back by your cpu on either of those games.