[AT] Haswell Refresh comes with improved TIM, unlocked Pentium due mid-2014

[cbn]

AnandTech Review (finally): www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k

That was a great review, but for the gaming comparison I was very disappointed they did not test the OC Pentium 3258 against one of the 3MB cache Haswell Core i3's (either the i3-4130 or i3-4150).

Instead they used the more expensive 4MB Core i3-4330.

 

[Techhog]

AnandTech Review (finally): www.anandtech.com/show/8227/devils-canyon-review-intel-core-i7-4790k-and-i5-4690k

So this "new TIM" is just the one from Ivy Bridge? -_- I'm speechless.

 

[StinkyPinky]

The new pentium is a pointless CPU. Why bother going to the expense of buying it, purchasing a cooler, and a decent Mobo to overclock it when you can just get an i3, a stock cooler, and a budget mobo and get better overall performance at the same price (or even cheaper since coolers aren't cheap)

 

[Techhog]

The new pentium is a pointless CPU. Why bother going to the expense of buying it, purchasing a cooler, and a decent Mobo to overclock it when you can just get an i3, a stock cooler, and a budget mobo and get better overall performance at the same price (or even cheaper since coolers aren't cheap)

Because it can overclock well on the stock cooler and some budget boards are getting BIOS updates for overclocking that Pentium.

 

[cbn]

The new pentium is a pointless CPU. Why bother going to the expense of buying it, purchasing a cooler, and a decent Mobo to overclock it when you can just get an i3, a stock cooler, and a budget mobo and get better overall performance at the same price (or even cheaper since coolers aren't cheap)

If you read the following thread, you'll see a person does not need anything better than the stock cooler for a good overclock:

http://forums.anandtech.com/showthread.php?t=2391096

 

[StrangerGuy]

The new pentium is a pointless CPU. Why bother going to the expense of buying it, purchasing a cooler, and a decent Mobo to overclock it when you can just get an i3, a stock cooler, and a budget mobo and get better overall performance at the same price (or even cheaper since coolers aren't cheap)

Yeah, put me in the "what is the big deal camp." when I can get a H81M mobo + 4590 for like mere 15% more over a cheapo Z87/97 mobo + Pentium AE + Hyper 212+. Then there is also the real world performance, what the AE wins in IMO irrelevant benchmarks it get crushed by Haswell quads in relevant stuff like gaming.

Simply isn't as value worthy as ~4GHz Pentium E6800 on a ~$50 G41 board or a 3GHz $180 E6300 Conroes that actually comes close to a OC $1000 Extreme Edition.

 

[SPBHM]

perfect CPU to run emulators...

PCSX2
110.73 - SLUS 20672 - Pentium G3258 @ 4.7GHz
109.22 FPS - SLUS 20672 - Intel Core i7 4770K - 4.8 GHz OC
102.24 FPS - SLUS 20672 - Intel Core i5 2500K - 5.2 GHz OC
http://forums.pcsx2.net/Thread-CPU-Benchmark-designed-for-PCSX2-based-on-FFX-2





I don't know if it's a repost but Digital Foundry made some interesting testing with the g3258 and a gtx 760/750 Ti...
Crysis 3 at high details was pretty poor compared to i7s, medium was fine, but with a lock at 30FPS the experience was good on high.
http://www.eurogamer.net/articles/digitalfoundry-2014-pentium-g3258-review

a big shame they ignored Core i3s which is realistically the only comparison to be made... i3 + potentially cheaper MB vs G3258 + OC friendly MB.



"Please Intel, create an i3 overclocking processor."

oh yes.

That was a great review, but for the gaming comparison I was very disappointed they did not test the OC Pentium 3258 against one of the 3MB cache Haswell Core i3's (either the i3-4130 or i3-4150).

Instead they used the more expensive 4MB Core i3-4330.

the difference is really, really small... 4150 (3.5Ghz) will perform almost the same as the 4330

pclab.pl compared the g3258 to the 4150 and some of their other tests have both 3 and 4MB i3s, like:
http://pclab.pl/art57916-12.html

the difference could be totally explained by clock, I'm sure in some very specific case it could be something, but don't expect it to change any conclusions you could make on this comparison (i3 vs pentium)


here some testing comparing 3 to 6 and 8MB of l3
http://www.madshrimps.be/articles/a...sary-Edition-G3258-CPU-Review/5#axzz37XbO2000

 

[TheRealSintel]

Can the memory be overclocked on the little pentium?
If so, is it dependent on the board (manufacturer tweaks/allowances) or chipset or unlocked always?
Anyone found a review that actually measures memory speed impact on the overclocked performance (if possible)?

 

[DrMrLordX]

It is board-dependent. I'm fairly certain that you'll need Z87 or Z97 for full overclocking of the memory subsystem. I'm sure memory performance will have some effect, though by how much is uncertain. The L3 cache tests might indicate that it won't be as big a deal as you would think.

 

[BadThad]

The new pentium is a pointless CPU. Why bother going to the expense of buying it, purchasing a cooler, and a decent Mobo to overclock it when you can just get an i3, a stock cooler, and a budget mobo and get better overall performance at the same price (or even cheaper since coolers aren't cheap)

$99 + tax mobo and CPU, OC's with the STOCK cooler:

http://www.microcenter.com/site/brands/G3258Bundle.aspx

Let's see your i3 system that competes price wise.

 

[Insert_Nickname]

Can the memory be overclocked on the little pentium?
If so, is it dependent on the board (manufacturer tweaks/allowances) or chipset or unlocked always?
Anyone found a review that actually measures memory speed impact on the overclocked performance (if possible)?

It is board-dependent. I'm fairly certain that you'll need Z87 or Z97 for full overclocking of the memory subsystem. I'm sure memory performance will have some effect, though by how much is uncertain. The L3 cache tests might indicate that it won't be as big a deal as you would think.

Yup, Z87 or Z97 is required. That's one of the Z-series least publicized features, it allows memory overclocking on ALL Intel CPUs.

Its not a big deal, Asrock has a couple of low-cost Z97 anniversary boards available.

 

[jana519]

I wonder if we will start seeing more and more of these "special edition" desktop processors as the advancements in fabbing become more and more costly.

 

[cbn]

a big shame they ignored Core i3s which is realistically the only comparison to be made... i3 + potentially cheaper MB vs G3258 + OC friendly MB.

AFAIK, there are not a lot of review sites that made the G3258 to Core i3 comparison. (Anandtech and Tom's used the i3-4330 and PCLab used the i3-4150).

I would be great to see testing of stock cooler Non-Z overclocked 4.5 Ghz G3258 vs. i3-4130. This particularly at lower resolution and detail settings with a relatively affordable Video Card such as R7 260X (as low as $85 AR at Newegg) or 750 GTX.

I'd imagine a 4.5 Ghz 3MB 2C/2T might even do surprisingly well against a 3.4 GHz 3MB 2C/2T in a highly multi-threaded game like Crysis 3 once we drop resolution and detail settings down.

Notice even, in the following test with a large 780GTX Ti and high detail settings, the 4.7 Ghz G3258 does quite well against the 3.5 Ghz i3-4150.

https://translate.google.com/transl...sl=auto&tl=en&u=http://pclab.pl/art57691.html

 

[cbn]

here some testing comparing 3 to 6 and 8MB of l3
http://www.madshrimps.be/articles/a...sary-Edition-G3258-CPU-Review/5#axzz37XbO2000

If only there were more than three games tested.

The three game test we conducted at the high detail preset show a slight FPS improvement from 3 to 6MB cache. Only the Sleeping Dogs benchmark seems to scale big time with an added 6fps on the High detail preset. Once Ultra mode is selected the GPU bottleneck jumps in.

However, the fact one of the titles did scale quite well (in such a small sample) makes me wonder how many other games might be affected by cache?

 

[SPBHM]

AFAIK, there are not a lot of review sites that made the G3258 to Core i3 comparison. (Anandtech and Tom's used the i3-4330 and PCLab used the i3-4150).

I would be great to see testing of stock cooler Non-Z overclocked 4.5 Ghz G3258 vs. i3-4130. This particularly at lower resolution and detail settings with a relatively affordable Video Card such as R7 260X (as low as $85 AR at Newegg) or 750 GTX.

I'd imagine a 4.5 Ghz 3MB 2C/2T might even do surprisingly well against a 3.4 GHz 3MB 2C/2T in a highly multi-threaded game like Crysis 3 once we drop resolution and detail settings down.

Notice even, in the following test with a large 780GTX Ti and high detail settings, the 4.7 Ghz G3258 does quite well against the 3.5 Ghz i3-4150.

some of the settings on crysis 3 greatly reduces the CPU load, so like on DF test with medium settings, I would expect it to be GPU limited with the 260x/750 for both the i3 and g3258 OC...

trouble is that as watch dogs shows some MT games can act very badly with 2 threads, a lot worse than crysis 3.

If only there were more than three games tested.



However, the fact one of the titles did scale quite well (in such a small sample) makes me wonder how many other games might be affected by cache?

but 3 to 6 is not 3 to 4

the other link I posted have a 3MB 3.5GHz i3 and 4MB 3.6Ghz with almost no difference for that game at least, I'm sure the Anandtech results for the 4330 would be almost the same with the 4150 (biggest difference would be the IGP)

I don't think that it's a big deal... the same happens with the i5 vs i7 on games loading less than 4t,

 

[cbn]

I brought up using a smaller video card at reduced resolution and detail settings for at least two reasons:

1. It seems to me like it would be the most likely use case scenario for the typical end user who buys G3258, i3-4130 or Athlon x4 750K/760K.

2. Something doesn't seem right with the testing results when large GPU(s) are used. While increasing GPU size or number of GPUs should reduce GPU bottleneck, I see another problem being a factor ( the ability of a small CPU to keep large GPU(s) fed.)

Therefore I propose some type of testing with reduced resolution and detail settings to 1.) reduce GPU bottlenecking 2.) reduce stress on the CPU to feed the GPU.

 

[cbn]

so like on DF test with medium settings

What does "DF test" mean?