X48 vs P45 Crossfire Comparison... OUCH 8x/8x = fail

MaStaFLaSH said:
This is false. He was using a 790i, the 4870 is just that good. It's not a bandwidth limitation of PCIE 1.1.
Click to expand...
I went back to the review to prove you wrong, but it said 790i in the page about CF, where I could have sworn it just said 680i. I wouldn't be surprised if they're still correcting typos.

I really wish a reputable forum member who owns both chipsets would just put this to rest...
 
bump for that, I would like to know just what the penalty is for 8 x 8. the P45 would be much cheaper. then again if AMD would pull there heads out and give us a good CPU would have these nice options.
 
RamGuy said:
You have to remember. P45 = 65nm, X48 = 90nm, it's much more expensive to produce X48 boards, that's why they cost more I guess..
Click to expand...
Not even close. It costs more because it is a better chipset in terms of performance. It is designed with a significant increase in bandwidth. The catch there is very few users actually need the increase in bandwidth. Those whom really need the increase are overclocking the $1400 extreme processors or running high end crossfire setups, as you are starting to see roll out now.

Think about it from a vehicle point of view. It costs alot more to make that Corvette than it does a Honda Civic. Between the two choices, you would buy the Corvette for raw performance but that doesn't mean it has a soft ride or fuel efficiency...so most people buy the Civic and are more than happy with it waiting on the vtec to kick in yo.
 
Yeorwned said:
Not even close. It costs more because it is a better chipset in terms of performance. It is designed with a significant increase in bandwidth. The catch there is very few users actually need the increase in bandwidth. Those whom really need the increase are overclocking the $1400 extreme processors or running high end crossfire setups, as you are starting to see roll out now.

Think about it from a vehicle point of view. It costs alot more to make that Corvette than it does a Honda Civic. Between the two choices, you would buy the Corvette for raw performance but that doesn't mean it has a soft ride or fuel efficiency...so most people buy the Civic and are more than happy with it waiting on the vtec to kick in yo.
Click to expand...

Actually the only advantage X48/X38 has the dual PCIE2.0 x16. It doesn't have more bandwidth other than that, and it's not really that great of an OC'r compared to P35/P45, heck even P965 can OC better. It does OC fine, just not as good as P35/P45/P965. Also the fact that it's a 90nm fab instead of a 65nm fab does raise it's cost.
 
I'd venture a pretty strong guess that Crossfire'd 4870s are able to better meet the bandwidth for 2560x1600 w/ 4x AA. Consider how at lower end cards can deliver playable results at lower resolutions, but quickly drop off as resolution and AA get higher. Crossfire'd 4870s may be getting by some x8 limit, but at part of it has to do with greater combined processing power giving less of a drop off at higher settings. 8800GT vs. 8800GT SLI shows the same greater than 100% boost, while 9800GTX+ only gets a 74.4% boost in SLI at 2560x1600. Of course, different games benefit differently with multiple GPUs, such as Crysis' pathetic gains compared to COD4.
 
Valset said:
bump for that, I would like to know just what the penalty is for 8 x 8. the P45 would be much cheaper. then again if AMD would pull there heads out and give us a good CPU would have these nice options.
Click to expand...
Yeorwned said:
Not even close.
Click to expand...
LOL.

While the extra features would inflate the cost somewhat if it cost the same to manufacture the two chipsets, Intel's 65nm process simply does provide a cost advantage over their 90nm process. Die size matters.

Back in the days of 865/875 (anyone remember the P4P800? :D)you might have had a point. But even then the higher prices of 875 boards were often offset by better bundles and higher quality motherboards.
 
Seen those, they are using a 3870 so doesn't help :(

Edit: Ok so the guy on XS says that so far he hasn't seen his asus P5Q take a hit like the gigabyte P45, he should have more results tomorrow *crosses fingers*

The only game he's tested was TF2 at 19x12 8xAA 16xAF and he was getting 150-200 frames which sounds exactly right compared to numbers I've heard of on X38/X48 systems.
 
Like *his* Gigabyte P45 board, or the one in that review?

I'd be surprised if the Asus board had that much of an advantage over the Gigabyte one.
 
i have x38 so if you want me to run anything just let me know
 
MaStaFLaSH said:
Actually the only advantage X48/X38 has the dual PCIE2.0 x16. It doesn't have more bandwidth other than that, and it's not really that great of an OC'r compared to P35/P45, heck even P965 can OC better. It does OC fine, just not as good as P35/P45/P965. Also the fact that it's a 90nm fab instead of a 65nm fab does raise it's cost.
Click to expand...


Wrong. The x38 OC's my Q6600 and my dads E8450 way better than my P965 board did.
 
RADEoN said:
Wrong. The x38 OC's my Q6600 and my dads E8450 way better than my P965 board did.
Click to expand...

Well for quads P965 is definitely a bad choice, but for duals it was one of the best till P45 came out (Rev C2 P965 chips). E8450???? what chip is that?
 
MaStaFLaSH said:
Edit: Ok so the guy on XS says that so far he hasn't seen his asus P5Q take a hit like the gigabyte P45, he should have more results tomorrow *crosses fingers*

The only game he's tested was TF2 at 19x12 8xAA 16xAF and he was getting 150-200 frames which sounds exactly right compared to numbers I've heard of on X38/X48 systems.
Click to expand...

Is there a link to this? I haven't been able to find it yet. Thanks.
 
MaStaFLaSH said:
Actually the only advantage X48/X38 has the dual PCIE2.0 x16. It doesn't have more bandwidth other than that, and it's not really that great of an OC'r compared to P35/P45, heck even P965 can OC better. It does OC fine, just not as good as P35/P45/P965. Also the fact that it's a 90nm fab instead of a 65nm fab does raise it's cost.
Click to expand...
You might want to consider brushing up on your research because this is definately speculation. Put an extreme quad in each board and try to say that a 12% difference isn't significant benefit from bandwidth overhead...
 
I guess one question is, will there be another "X" generation chipset coming out soon. We got the X48 with the full bandwidth PCIe 2.0 x16s, but the old southbridge. The P45 looks new and shiny, but has a possible bottleneck with PCIe 2.0 x8s when using crossfire. I'd like to see more testing done in the 4870 (and x2) to see if a PCIe 2.0 x8 will be sufficient bandwidth or not.
 
Indeed there will be, depending on what you consider soon. It is the X58, it will (likely) be out by the end of the year. It will have 36 lanes of PCI-E 2.0, so most likely 32 (16x2) for graphics-card use and 4 for other peripherals (e.g.: gigabit Ethernet). And it will indeed use ICH10 for the Southbridge. Of course you'll have to get DDR3 and a Bloomfield (Nehalem) in order to actually use any motherboard utilizing the chipset. ;)

In the meantime, X48 is plenty good. The only really difference between ICH9 and ICH10 is that plain ICH10 has AHCI support (you had to get ICH9R to get it before) and an obscure Wake-on-LAN feature. There's really no reason to get excited about ICH10 over ICH9R for most people.
 
farfromhome said:
In the meantime, X48 is plenty good. The only really difference between ICH9 and ICH10 is that plain ICH10 has AHCI support (you had to get ICH9R to get it before) and an obscure Wake-on-LAN feature. There's really no reason to get excited about ICH10 over ICH9R for most people.
Click to expand...

Ditto, WOL while nice to have is pointless for me since my rig is connected wirelessly.
 
Wait, so ICH9 doesn't support WOL at all? Or only in specific situations?
 
Just a fun update on this. I got my 2 4850s when the Best Buy sale was on. I've been using a Gateway P6831FX for gaming so the only board I had to throw them in was an Asus P5W DH 975X. I knew that the 8x/8x v1.1 slots were going to be a problem until I got an X48 board in but here is something funny to look at:

3DMark06 (With Crossfire Enabled): ~9800
3DMark06 (With Crossfire Disabled: ~9200

Yes, that is how much these cards are bandwidth crazy. Here are the current system specs:

Asus P5W DH
Core 2 Duo E6400
2GB DDR2-5300
2 x Visiontek 4850s
Vista 32bit
ATI "Hotfix" drivers

I have a Q6600 and Crucial BallistiX for when my DFI X48 board comes in. Really interested to see the comparison numbers between that setup and the current one. I may even toss the Q6600 and BallistiX in now so that I can get a better idea of how just going to 16x/16x v2.0 affects things.
 
Maximus825 said:
Just a fun update on this. I got my 2 4850s when the Best Buy sale was on. I've been using a Gateway P6831FX for gaming so the only board I had to throw them in was an Asus P5W DH 975X. I knew that the 8x/8x v1.1 slots were going to be a problem until I got an X48 board in but here is something funny to look at:

3DMark06 (With Crossfire Enabled): ~9800
3DMark06 (With Crossfire Disabled: ~9200

Yes, that is how much these cards are bandwidth crazy. Here are the current system specs:

Asus P5W DH
Core 2 Duo E6400
2GB DDR2-5300
2 x Visiontek 4850s
Vista 32bit
ATI "Hotfix" drivers

I have a Q6600 and Crucial BallistiX for when my DFI X48 board comes in. Really interested to see the comparison numbers between that setup and the current one. I may even toss the Q6600 and BallistiX in now so that I can get a better idea of how just going to 16x/16x v2.0 affects things.
Click to expand...

3D Mark doesn't mean jack, P45 and X48 were getting similar numbers on it. Can you run some game benchies matching the settings used here I want to see what numbers you get. even tho you are only x8/x8 PCIE1.1 I would still like to see the nubers ;)

If you could that would be great :D
 
They don't give enough information about the settings they used for me to do a comparison.

And no, 3DMark isn't useful for comparing different systems. But it is useful for comparing how different tweaks affect a single system.
 
I have an interesting update. I changed out my CPU and memory for Q6600 and BallistiX. Running at 2.7GHz I got 14821 in 06 with Crossfire enabled. So a jump of over 5000 points. I'm going to do some more testing when I have the time, but it seems that it wasn't so much the limited bandwidth as it was my CPU just not keeping up.
 
You must log in or register to reply here.
Back
Top