Is it a good time to buy?

[pendragon1984]

There are a lot of screen issues with the current machines. Wait. That's my advice. I replaced my original one because of an uneven yellow tint, and the replacement although a little better has it as well.

 

[simon48]

If you don't know what "or" means, I don't know that I have that much more to say.

That's just... I don't even know what.

Battlefield 3 HQ: The 650M performs at 36.6FPS. Take the Iris Pro 5200 and scale it up 40%, and that's 32.1FPS. That's within spitting distance as far as I'm concerned—a heck of a lot closer, in fact, than your claim that 83.8% is about equal to 100%.

So 36.6 fps is within spitting distance of 32.1 fps (14% change), but me being off by 16.2% is a heck of a lot more?

And at this point, you're just making up numbers. That's cute and all, but is hardly grounds for claiming, "NO U WRONG BRO." Also, I love how you cherry-picked the game with the biggest gap. Again, still cute.

And you're omitting numbers. Adding VRAM doesn't do anything for games? I guess it's just for show.

Edit: Here, I'll do the work for you. Complete with all the calculations. And then average it. Shocking, what do you know—the average without cherry-picking comes out to 36%, right in the middle of that "30-40%" number I've been talking about. It sure is great when one doesn't bias data selection in favor of the argument they're trying to make.

Game GT650M Iris Pro 47W Percentage Difference
Metro 20.0 15.3 30.7%
Bioshock 35.5 24.2 46.7%
Sleeping Dogs 28.4 19.3 47.2%
Tomb Raider 44.0 31.4 40.1%
Battlefield 36.6 22.9 59.8%
Crysis 3 32.0 22.3 43.5%
Crysis: WH 34.5 29.0 19.0%
Grid 2 30.2 30.8 -1.9%
Total 35.6%


See above.

You have a really good point, but you're omitting the difference between the 650m and the 750m which includes a big VRAM boost.

I picked Battlefield 3 because I knew it was a big gap, I have no qualms about that, we were using it and I stuck with it.

 

[Jeanloup]

Btw in how many months do you think the next revision/refresh will be ? (i know it´s going to be a minor one)

 

[simon48]

Btw in how many months do you think the next revision/refresh will be ? (i know it´s going to be a minor one)

We don't know how big it will be yet. If you don't count the chance for a tiny CPU clock speed boost in ~4 months, ~10 months.

 

[john123]

Btw in how many months do you think the next revision/refresh will be ? (i know it´s going to be a minor one)

It most likely depends on Intel and Broadwell. Right now, that's slated for "second half of 2014." I'd guess July, but if that slips, look for a speed bump in the interim and a release later in the fall.

 

[john123]

So 36.6 fps is within spitting distance of 32.1 fps (14% change), but me being off by 16.2% is a heck of a lot more?

This is why percentages are often poor metrics. I'd rather not derail this thread with a discussion of statistics, which is my primary background, although I suppose I can if there's interest. Suffice it to say that I don't think 4 FPS makes that much difference.

And you're omitting numbers. Adding VRAM doesn't do anything for games? I guess it's just for show.

It's not a linear performance boost. As you alluded to previously, more VRAM helps with some games and especially at higher resolutions. Are the types of resolutions and games Anand et al. are testing on high enough for it to matter? That's TBD.

Of note for other readers who might be following this discussion and are wondering about the 128MB of VRAM on the Iris Pro: that's potentially misleading the Iris Pro 5200 can access the main system memory too. That said, it certainly does make a difference.

You have a really good point, but you're omitting the difference between the 650m and the 750m which includes a big VRAM boost.

Some of the benchmarks people have been posting for the 750M around here have actually been unimpressive, including one person suggesting it's running at 925Mhz (it's supposed to be 967Mhz), which would be underclocked, versus the 900Mhz on the overclocked 650M. Some suggest that performance is roughly equivalent, but are they testing at higher resolutions where it would matter? It's unclear. Again, I'm very eager for Anand's review to wrap up to figure these issues out.

 

[simon48]

Suffice it to say that I don't think 4 FPS makes that much difference.

I agree, but why be so ridiculously specific then?

It's not a linear performance boost. As you alluded to previously, more VRAM helps with some games and especially at higher resolutions. Are the types of resolutions and games Anand et al. are testing on high enough for it to matter? That's TBD.

Of note for other readers who might be following this discussion and are wondering about the 128MB of VRAM on the Iris Pro: that's potentially misleading the Iris Pro 5200 can access the main system memory too. That said, it certainly does make a difference.

Yep, the Iris Pro in the rMBP can use up to 1GB of RAM.

Some of the benchmarks people have been posting for the 750M around here have actually been unimpressive, including one person suggesting it's running at 925Mhz (it's supposed to be 967Mhz), which would be underclocked, versus the 900Mhz on the overclocked 650M. Some suggest that performance is roughly equivalent, but are they testing at higher resolutions where it would matter? It's unclear. Again, I'm very eager for Anand's review to wrap up to figure these issues out.

I'm waiting too, any day now, I hope.

 

[john123]

I agree, but why be so ridiculously specific then?

LOL, this all just started with the whole "will Broadwell graphics be comparable" thing. But I think after going through it, we're roughly on the same page.

I'm waiting too, any day now, I hope.

Amen.

 

[simon48]

LOL, this all just started with the whole "will Broadwell graphics be comparable" thing. But I think after going through it, we're roughly on the same page.

Amen.

Well, now that we have just about the same outlook, I feel sorry for anyone who gets in our way.

Come on AnandTech!