AM2 Testing at THG

[mike64]

There shouldn't be much of a difference between a 4800+ s939 and 4800+ AM2 part, the speed rating is the same! AM2 isnt a waste of time, it is just AMD moving to a different type of memory(a sideways step), so at the moment there is no difference but the fact that 65nm, low power and probably quad core CPU's will all be on this platform it'll hardly be a waste then! If you have a top of the line s939 CPU now then there is no need to change now just cos its a new socket, but in time chips that are a lot faster, cooler with more parallelism around and they'll fit into a AM2 socket!

 

[FirePhoenixCtk]

There shouldn't be much of a difference between a 4800+ s939 and 4800+ AM2 part, the speed rating is the same! AM2 isnt a waste of time, it is just AMD moving to a different type of memory(a sideways step), so at the moment there is no difference but the fact that 65nm, low power and probably quad core CPU's will all be on this platform it'll hardly be a waste then! If you have a top of the line s939 CPU now then there is no need to change now just cos its a new socket, but in time chips that are a lot faster, cooler with more parallelism around and they'll fit into a AM2 socket!

agreed, there are already some scheduled to run at 65w the same as the current semprons.

 

[jmhduck]

HI,

That doesn't sound good really. In essence what kind of speed increase can we expect when the right memory and latencies are available?

Thanks
Raja

We won't really know until final versions of cpu's, decent mobo's and ram are out.

 

[Minstadave]

Yeah I'd take this all with a pinch of salt, until I see some a chip with the fixed memory controller and some DDR-II 800 I'm going just watch and wait.

 

[Dunky]

Why were they running DDRII-667 at DDR-II-400 speeds?

You can blatently see from the bandwidth results that its DDR2-400 that they are using, because the bandwidth is similar to DDR-400. DDR2-667 should have a bandwidth of near 10.6GB/s NOT 5.8GB/s. So they are testing an revision F chip with ram at DDR2-400 (4-4-4-12) latency against a s939 chip with DDR-400 (2-2-2-5) latency. That's right, double the latency!

DDR400 = PC3200 => 3.2GB bandwidth (dual channel makes 6.4GB/sec)
DDRII-667 = PC5400 => 5.4GB bandwidth (dual channel makes 10.8GB/sec)

It's quite possible that they weren't running dual channel on AM2 so the figures are probably correct, but quite an unfair comparison. Review is to be taken with a few bags of salt and a dash of pepper.

Although there may not be much difference, I think I will make the jump quite soon after it comes out. I'm holding out on dual core til then. You may think I'm crazy but I like to play about with the latest tech and have a spare machine for the important stuff. Although I'd prefer some linux support before I make the move but not too sure if that will happen. But defo the real benefits of AM2 will be reduced TDP and 65nm further down the line.

 

[div0]

Seems like some people were expecting a lot more from this review than I was

I'm don't really understand why people think the results are much of a surprise.

Wait until something is actually released before you start trying to look compare it to existing products.

Personally I just loked at the review as an insight to "roughly" where in the development the product is. And from what I read, it looks like it is going pretty well.

I don't think of AM2 as the sort of "upgrade" that will have current 939 users itching for an early change. It's just AMD "sidestepping" onto a slightly different path. DDR seems to be on the way out - and so prices will probably rise - moving to DDR2 will bring another platform that is more up to date.

Reading too much into this sort of pre-release "benchmark" is not very sensible imo. I just looked at it as a bit of interesting reading

 

[Soul Rider]

I Think we will see a much improved performance when the final chips are relased and everything is working as AMD desire. Plus new motherboard chips will be developed by manufacturers to exploit some of the new features available in the chips.

We'll just have to wait and see what happens closer to 6/6/6...

 

[Sumanji]

It's quite possible that they weren't running dual channel on AM2 so the figures are probably correct, but quite an unfair comparison.

Sounds like the most logical explanation... aren't THG supposed to be really Intel biased anyway?

Suman

 

[deadeyedic30]

I only recently went pci-e so my next upgrade will be 939 dual core in the summer when prices are falling due to the arrival of AM2.

 

[tomos]

check the comments to the article @ THG, all these points about how THG messed up the review is in there.

hopefully anandtech etc will do this properly.

in its own way, it shows some of the potential. even with a pre-release board, on a chip that doesnt work fully (wont take ddr2-800 RAM), on really slack timings with no bandwidth increase from DDR, it still matches its 939 in performance when that is running @ 2-2-2-5 timings!

as dunky said above, they goosed up. cpu-z and the sandra tests show the rams speed - or lack of.

cant wait to see how AM2 does when setup properly with chips and mobos closer to their final revision

 

[Sumanji]

Well if it's keeping pace with current chips crippled with what we reckon is single channel DDR2-400 at horrible latencies, I'm sure retail products using dual channel DDR2-800 at something like 5-5-5-12 will be pretty nippy

Suman

 

[Dunky]

Yep, one thing to remember is that AMD will already have thought about DDR2's current latencies and no doubt will have tweaked the memory controller based on this. When DDR2 latencies start dropping eventually, no doubt AMD will update the memory controllers then to reflect these changes.

 

[Robert]

I'll have a gander at it when I come to buy around summer/early winter time.

 

[tomos]

any guesses when real tests will be done?

i mean, ddr2-800 ram with not too slack timings on a non alpha/beta mobo/cpu?

it'll be early next year for 65nm cpus and quad core tho wont it?