[CoffeeFiend]

And it seems to be not all that great. I was hoping for AMD to come up with a match for the Core 2 Duo, but seemingly it's not faster, it's not cheap (100$ less for a pair of FX-74 than a QX6700, but the pair is slower) , and uses twice as much power (595W for the system tested). I doubt the motherboards will be cheap either, like all dual socket motherboards, so that "100$ less" might not help much. Doesn't use AM2 either (short lived socket seemingly)

http://www.hothardwa...x?articleid=911
http://www.extremete...,2065493,00.asp
http://enthusiast.hardocp.com/article.html...oZW50aHVzaWFzdA

Maybe it's just me, but I had expected more of it.

This post has been edited by crahak: 30 November 2006 - 05:22 PM

[ripken204]

wow, what a pice of crap that is, i was hoping for that to beat the c2d... oh well

[jcarle]

AMD better come out with something amazing in 2007 because right now Intel's making them look like fools, same as AMD did back last year.

[cluberti]

Note that neither Intel's nor AMD's current quad-core offerings are actually quad-core at all, but dual dual-core processors. Hopefully both will show better products when we see REAL quad-core processors (and not the current trash of using the system bus to communicate - what a mess).

[Zxian]

I haven't even read the whole article yet, but there's one part that worries me from the get go... max temperature of the FX-74 is 56C, and yet it's got a TDP of 125W? You'd need a pretty good heatsink to pull that off.

Their testing pushed it to 60-something C anyways...

[jcarle]

cluberti, on Nov 30 2006, 10:20 PM, said:

Note that neither Intel's nor AMD's current quad-core offerings are actually quad-core at all, but dual dual-core processors. Hopefully both will show better products when we see REAL quad-core processors (and not the current trash of using the system bus to communicate - what a mess).

Whether it's a true quad core or a pair of dual cores isn't the system bus bandwidth usage the same? I mean aside from the fact that the two pairs of cores don't share their L2 Cache, where's the difference? They'd be part of the same physical package either way that interfaces with the same socket interface either way, so I'm a bit confused?

[Aegis]

It's a marketing strategy . They're releasing this to put Intel off-guard by having them believe AMD is not a threat. Then when they release the Intel-killer (Barcelona), Intel won't have anything ready for months on end.

[CoffeeFiend]

Zxian, on Nov 30 2006, 11:38 PM, said:

I haven't even read the whole article yet, but there's one part that worries me from the get go... max temperature of the FX-74 is 56C, and yet it's got a TDP of 125W? You'd need a pretty good heatsink to pull that off.

Talking about heat sinks, there's probably enough of them for the system to hover: 2 for the 2 CPUs, and 3 on the motherboard (5 in total). The thing's just covered by heat sinks. Along with power usage, it's a sure thing you don't have to heat during winter, and get crazy AC bills during summer... Especially if combined with nowadays's power-hungry video cards!

There's only one nice thing about the whole setup: 12 SATA ports onboard (I've dreamed of that for a while - the more, the better). But then the RAM is a big downside. It needs all 4 slots filled, and if you want to upgrade, you can't add any, throw away the old ones and buy bigger sticks. Maybe it's just me, but in a high-end 2x dual core rig with 1000$ CPUs, I'd like 8 memory slots (hard to fit on a normal sized board for sure, if even possible). Especially when the board has a MSRP of 480$! (enough to buy a nice E6400 and nice motherboard by itself), so almost 1500$ with the FX-74 (now add 4 sticks of RAM and everything else).

As for "real" quad-cores, I doubt it will really deliver much more speed - especially on the AMD side, as HT is already pretty fast, and CPUs already have their own RAM banks. It's already widely used like that by Operon systems and seems good enough for the job (and scales well). Faster interconnects are only going to help so much.

Either ways, K8L better be a LOT nicer than this.

[jcarle]

OMG... you know, at first I had only glanced over this post. I actually took the time to look at the reviews and the pictures this time. WOW. What a mess... you know, I'm sorry, but that's a really ridiculous attempt by AMD. It even seems a mess just by looking at it.

[cluberti]

jcarle, on Nov 30 2006, 10:48 PM, said:

Whether it's a true quad core or a pair of dual cores isn't the system bus bandwidth usage the same? I mean aside from the fact that the two pairs of cores don't share their L2 Cache, where's the difference? They'd be part of the same physical package either way that interfaces with the same socket interface either way, so I'm a bit confused?

The processor cores need to be able to communicate with all other cores - usually, this is done in the processor core itself (or between cores on the die in multiple-core scenarios), but with these dual dual-core setups, the processors have to actually go OUTSIDE of the processor itself, to the system bus, to communicate with the other set of cores on the same processor. Inter-processor communication within cores on one die is very, very fast. The system bus is MUCH slower than the processor core or die when it comes to communication, so this is a really, really bad idea - we've introduced latency (well, above and beyond the normal latencies processors experience on-die) where there was none before. Think of it as going from computer to computer communication over a 100Mb link to a 14.4k modem - not a good idea if speed is your desire (and it should be with processors!).

[jcarle]

cluberti, on Dec 1 2006, 11:08 AM, said:

The processor cores need to be able to communicate with all other cores - usually, this is done in the processor core itself (or between cores on the die in multiple-core scenarios), but with these dual dual-core setups, the processors have to actually go OUTSIDE of the processor itself, to the system bus, to communicate with the other set of cores on the same processor. Inter-processor communication within cores on one die is very, very fast. The system bus is MUCH slower than the processor core or die when it comes to communication, so this is a really, really bad idea - we've introduced latency (well, above and beyond the normal latencies processors experience on-die) where there was none before. Think of it as going from computer to computer communication over a 100Mb link to a 14.4k modem - not a good idea if speed is your desire (and it should be with processors!).

Okay, I can see how the latency can be introduced and I understand how the system bus is slower, but the part I don't get is why the two sets of cores would need to talk to each other? Doesn't windows just talk to each set of cores independently?

[Zxian]

jcarle, on Dec 1 2006, 12:58 PM, said:

Okay, I can see how the latency can be introduced and I understand how the system bus is slower, but the part I don't get is why the two sets of cores would need to talk to each other? Doesn't windows just talk to each set of cores independently?

If a Windows thread is running on one core, and it needs to access the thread running on the second core (simplest example I can think of is to close a program for example), then there would be communication between the two.

[cluberti]

Windows (or any other OS) doesn't control a lot of what a processor does. A lot of what the OS does is feed it code about the OS, not how to run the code it's sent to the processor, which core to run it on (unless affinity has been set), how to queue if the proc is busy, whether or not to access the data from L1 or L2 cache, etc. You'd be surprised how little your OS has to do with the processor's internal functions.

[Lost Soul]

geesh, i cant believe amd did this, now its a hard debate for me wether to go with a core 2 or hold off to see amd's next move , right now ill probably hold off abit to see how things go

[jcarle]

Lost Soul, on Dec 1 2006, 05:57 PM, said:

geesh, i cant believe amd did this, now its a hard debate for me wether to go with a core 2 or hold off to see amd's next move , right now ill probably hold off abit to see how things go

If anything I'd wait and see what the next generation of Core 2 (1333MHz FSB) is going to be like...

[cluberti]

If you're buying new machines, sticking with the older dual core processors is a wise choice (and making sure they've got at least 1MB of L2 cache is also a good idea). In late 2007, we should be able to get our hands on *real* quad-core processors, and that gives hope that by 2008, they will not suck as bad as these half-baked attempts at getting a product out the door by 2006 (and who cares what quality). Hopefully by 2008 quad-cores will even be affordable (although the cooling may cost more than the processor!!! ).

[jcarle]

I think the bigger problem at the moment, as has been demonstrated by many benchmarks, is that there is a serious lack of true multi-threaded software. The worse part is that the software that could benefit the most from the extra crunching power is solely single threaded.

Best examples I can think of are LAME MP3 encoding and XviD Video encoding.

[Camarade_Tux]

I have a Turion64 X2 here.
lame.exe uses from 95 to 100% of my two cores and encodes mp3 at a speed of 24/25x.

That's what I call a truely multi-threaded app.

[jcarle]

Camarade_Tux, on Dec 2 2006, 12:25 PM, said:

I have a Turion64 X2 here.
lame.exe uses from 95 to 100% of my two cores and encodes mp3 at a speed of 24/25x.

That's what I call a truely multi-threaded app.

Then why is it that every single processor completes in the exact same time as their single cored equivalent? Not to mention that the quad core processor completes in the exact same time as the dual core version? That's not what I call multi-threaded.

LAME Benchmarks

[Camarade_Tux]

Maybe because they used a non-threaded core!
http://rarewares.org/mp3.html