GT3e/GT3 v Desktop Radeon 4850

champion-7891

Member
Jun 7, 2011
99
0
0
How will the GT3e/GT3 in Haswell compare to a 512 MB Radeon 4850? I think it would be interesting if after a gap of 5 years we finally get integrated graphics that can perform upto that level. I've actually owned a 4850 for about 4 years and apart from the Witcher 2 I've been able to play most games at medium settings at 1280*960.

GT3 is supposed to be at par with GT640M. GT3e with 650M. How do they compare to the 4850?
 

Stuka87

Diamond Member
Dec 10, 2010
6,240
2,559
136
I am not holding my breath that they will be as fast as those cards. Intel was really picking and choosing benchmarks when they said that. We will find out soon, but I am willing to bet they wont be as fast in actual games as those cards.

I cannot recall off hand how well a 4850 does compared to current cards.
 

lagokc

Senior member
Mar 27, 2013
808
1
41
Considering the 7750 isn't even twice as fast as a 4830 I'm going to predict the GT3e is going to be slower than the 4850 for DX9 games. On the other hand the 4850 can't run DX11 which can be a bit more efficient.
 

Arkadrel

Diamond Member
Oct 19, 2010
3,681
2
0
well.....

i7-3770k has a HD4000 igp.
That scores about ~4300 in 3Dmark Vantage (3371 gpu, 26,6k cpu)

Supposedly the GT3e will be around x2,5 of that (in 3Dmark vantage).
So 3371 GPU score x 2.5 = ~8427.
(source: http://images.anandtech.com/doci/6926/Screen%20Shot%202013-05-01%20at%205.36.57%20PM_575px.png)


Meanwhile a 4850 (stock) gets around ~6400 in 3Dmark Vantage GPU score alone.

So the GT3e should be like ~30% faster than a 4850.


Considering the 7750 isn't even twice as fast as a 4830......
A 7750 gets a GPU score of around 11,000 in 3Dmark Vantage.

o_O <-- my face right now.

GT3e is "supposedly" closer to a 7750 than I would have thought.
(note this is for the Desktop versions of CPUs, not the ones going into notebooks ect)
 
Last edited:

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136
GT3e is "supposedly" closer to a 7750 than I would have thought.
(note this is for the Desktop versions of CPUs, not the ones going into notebooks ect)

From what I've heard, the GT3e will only be available on the absolute highest-end mobile i7's and desktop R-series BGA (e.g. non-upgradeable SFF) desktop i7's. So for most desktop users, the GT3e is a moot point. For mobile I'll wait and see what they bring to the table...
 

Face2Face

Diamond Member
Jun 6, 2001
4,100
215
106
well.....

i7-3770k has a HD4000 igp.
That scores about ~4300 in 3Dmark Vantage (3371 gpu, 26,6k cpu)

Supposedly the GT3e will be around x2,5 of that (in 3Dmark vantage).
So 3371 GPU score x 2.5 = ~8427.
(source: http://images.anandtech.com/doci/6926/Screen%20Shot%202013-05-01%20at%205.36.57%20PM_575px.png)


Meanwhile a 4850 (stock) gets around ~6400 in 3Dmark Vantage GPU score alone.

So the GT3e should be like ~30% faster than a 4850.


A 7750 gets a GPU score of around 11,000 in 3Dmark Vantage.

o_O <-- my face right now.

GT3e is "supposedly" closer to a 7750 than I would have thought.
(note this is for the Desktop versions of CPUs, not the ones going into notebooks ect)

That's pretty impressive. The 4850 is = to a 9800 GTX and they can still game somewhat at low-med settings @ 1080 or below. To have this power in a laptop without a dedicated power-sucking leach attached will be very nice :)
 

Arkadrel

Diamond Member
Oct 19, 2010
3,681
2
0
Three words: memory bandwidth bottleneck

The cake is a lie!

^-^

>> Memory bandwidth isnt always going to be a issue. <<
Currently it is, but in the near future, eDram & GDDR5 system memory solve that.
(and allow IGPs performance to go up a good bit)

Thats short term though, long term....
stacked memory cubes
, will skyrocket system memory bandwidth.


Whats really holding IGPs back is TPD.
You need some big cooling solutions to keep those IGPs cooled,
if they plan to grow nearly as big as the discrete card's.

Watercooling is the future.
I know there are already closed system 1 peice watercooling solutions,
but I expect that ll (watercooling) become the norm at some point.


The future:

either this (air) (Silver Arrow):
391_thermalright_silver_arrow_onmobo_big.jpg

*** Not sure how much "bigger" air cooling for a PC can grow, without it breaking the motherboard.
Air coolers keep growing in size and its becomeing a issue lol.




or this (water) (Corsair H100):
h100_rad_v2.png
 
Last edited:

NTMBK

Lifer
Nov 14, 2011
10,232
5,012
136
The cake is a lie!

^-^

>> Memory bandwidth isnt always going to be a issue. <<
Currently it is, but in the near future, eDram & GDDR5 system memory solve that.
(and allow IGPs performance to go up a good bit)

Thats short term though, long term....
stacked memory cubes
, will skyrocket system memory bandwidth.


Whats really holding IGPs back is TPD.
You need some big cooling solutions to keep those IGPs cooled,
if they plan to grow nearly as big as the discrete card's.

Watercooling is the future.
I know there are already closed system 1 peice watercooling solutions,
but I expect that ll (watercooling) become the norm at some point.


The future:

either this (air) (Silver Arrow):
24-big-thermalright-silver-arrow.jpg



or this (water) (Corsair H100):
h100_rad_v2.png

Oh god, eww, no, you can keep your watercooling. :p Today's "desktop" range is just repurposed laptop parts, they're not going to go blowing out the TDP.
 

Arkadrel

Diamond Member
Oct 19, 2010
3,681
2
0
I think IGPs will keep raiseing in performance, and that ll mean TPDs will go up.

I think the future is in watercooling.

Eventually air coolers will become to big to just hang there on the motherboard.

Watercoolers will just move the radiator outsides the PC case (ei. future pc cases, are buildt to accomidate this).
 

blastingcap

Diamond Member
Sep 16, 2010
6,654
5
76
The cake is a lie![/IMG]

No offense but you wrote "TPD" instead of "TDP" two posts in a row which makes me wonder.

Anyway, mem bandwidth will be an issue for some applications, not an issue in others, but I just don't see Haswell's GPU knocking off a 4850 for most games.
 

boxleitnerb

Platinum Member
Nov 1, 2011
2,601
2
81
IGPs will be TDP limited, too. With discrete GPUs, you can always allocate more die size and TDP since it's optional and therefore cheap. I don't see Intel or AMD approaching 100W and 200mm^2 for the GPU alone, ever. But AMD and Nvidia can and do, therefore discrete GPUs will always have a large performance lead if we look at the whole picture. And TMSC and Glofo improve their processes too. It's not like only iGPUs benefit from node shrinks.
 

NTMBK

Lifer
Nov 14, 2011
10,232
5,012
136
53963.png


Just a taster of how the HD4000 (running at tablet TDPs in the Surface Pro) compares against yesteryear's graphics cards...
 

lagokc

Senior member
Mar 27, 2013
808
1
41
Watercoolers will just move the radiator outsides the PC case (ei. future pc cases, are buildt to accomidate this).

One day laptops will have nozzles on the side so that when they're docked they will connect to a radiator that allows them to run at full speed, then when you undock them the nozzles seal and the chips reduce clockspeed :biggrin:
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,691
136
IGPs will be TDP limited, too. With discrete GPUs, you can always allocate more die size and TDP since it's optional and therefore cheap. I don't see Intel or AMD approaching 100W and 200mm^2 for the GPU alone, ever. But AMD and Nvidia can and do, therefore discrete GPUs will always have a large performance lead if we look at the whole picture. And TMSC and Glofo improve their processes too. It's not like only iGPUs benefit from node shrinks.

Agree. You simply can't shove a 250W IGP into a 100W CPU without issues. Would be interesting to see though... :p

>> Memory bandwidth isnt always going to be a issue. <<
Currently it is, but in the near future, eDram & GDDR5 system memory solve that.
(and allow IGPs performance to go up a good bit)

Thats short term though, long term....
stacked memory cubes
, will skyrocket system memory bandwidth.

No, but you do realize that the same improvements can be applied to discrete GPU's too? (Maxwell springs to mind). IGP's are always going to be behind in performance. They're far from useless of course.
 

skipsneeky2

Diamond Member
May 21, 2011
5,035
1
71
I was silly enough a while back to even consider the possibility that a 5 year old 9500gt 1gb was slower then the hd2000 on my i5 2500 non k,boy was i wrong about that.

Always kinda think about my 2006 build with the 6200le and the p4 that struggled to play BF2 even on 1024x768 low...hd2000 is capable today of medium 1366x768 with 60+fps with no aa on a core i5 notebook,terrible today but would have been awesome some years ago for a desktop.
 

Arkadrel

Diamond Member
Oct 19, 2010
3,681
2
0
One day laptops will have nozzles on the side so that when they're docked they will connect to a radiator that allows them to run at full speed, then when you undock them the nozzles seal and the chips reduce clockspeed :biggrin:

hahaha ^^ I doubt it.


No, but you do realize that the same improvements can be applied to discrete GPU's too? (Maxwell springs to mind). IGP's are always going to be behind in performance. They're far from useless of course.

The point was, that memory bandwidth wont always be holding back performance.

Yes the same improvements will go to discrete GPUs, but at some point giveing them more memory bandwidth wont do much of anything. Thats how big the jump in memory bandwidth will be once they start doing memory cubes.

Whats gonna hold back IGPs is the power consumption & cooling, compaired to seperate CPU + discrete GPU.
 

tvdang7

Platinum Member
Jun 4, 2005
2,242
5
81
I have an nvidia 540m in my laptop and i just want to know if the new intel will be somewhat close in performance to this.