Err I thought Kuma, Agena, Agena FX, Budapest and Barcelona are K10 with HT 3.0?

Anyway, wiki updated. Kuma's TDP as low 45W for 1.9GHz~2.3GHz models, 65W for 2.4GHz~2.6GHz models and 89W for 2.8GHz model. Somehow I don't understand the Bus Speed to Clock Frequency

http://en.wikipedia.org/wiki/AMD_Phenom

I'm just going to say it once, HT's bandwidth doesn't matter much to overall performance, because all it is at the end of the day is pretty much a glorified interconnect, and i don't see how an interconnect that primarily serves I/O devices that themselves are pretty slow is going to make the final Phenom platform kick any ass.

It's like giving an Autobahn to nation of backward peoples that use bullock carts. It simple isn't something normal users should give a shit about. What would be worth discussing (i think alot of us already discussed it) are it's execution and scheduling pipelines, it's branch prediction, instruction window and scheduler, the bitwidth of ops of it's SSE and Integer units and other stuff like that.

For the server market HT3 will probably make a difference, and that should i think be enough to give them the lead in 4 sockets and higher configs. Also if you're using a system with integrated gfx, ht3 should give some nice improvements there. But other than that.... not so much.

This post has been edited by ikanayam: Jun 30 2007, 07:18 AM

I think the dudes i quoted aren't that high up the scale and are probably talking about HT within a HT to NB context only. What are the chances of anyone here using the Direct Connect as a multiway system

Someone with the bragging rights for 4X4 (X2).

*crickets*

But as a marketing tagline to prolong DDR2's life, it could seem well enough for DAAMIT to not take a substantial loss for moving to AM3/DDR3 much later.

What do you think of SSE128 anyways? Substantial increase like Core was? Because basing on the POV benches that came earlier it doesn't really seem to help this chip float.

In term of latency and bandwidth sure Phenom will improve a lot.

whose kicking whose ass??? phenom launch in novemberr 07??? got link for it's review??? oc performance??? i tried google but found not much bout this phenom except of forums...

I don't think improving latency and bandwidth is any indication of a better performing phenom.
we have to remember it's a native quad core...so roughly, under full load, we can expect the HT needing to be theoretically at most 2x the bandwidth..but i'm sure they will optimise the memory controller.

i'm not being a fanboy, but amd should take cue from intel...their memory controller is still on the northbridge of a motherboard and yet latencies are WAY low. it's either because of their bigger CPU caches or just excellent engineering.

yeah it's quite true what charge-n-go said. we can't modify the hardware, but the nearest point of which we can modify it is through software, and the very lowest of all software would be the bios itself..controls the memory timings, voltage feed to the cpu, things like that. very delicate thing.

it would be great to see the first native quad-core around, but i think this is the main trouble, providing native solutions on how each cores communicate with each other, caching... i dont know whether ppl will agree but intel avoided much of this by just gluing two core duos together.

id like to see the performance numbers on barcelonas when it comes out!

hm..what d difference between bacerlona and phenom??i thought phenom will come out first???hm..

http://www.tomshardware.com/2007/07/16/cpu...2007/page4.html

biostar
am2+560

robyncom rm 250

So cheap?? It's around RM280 per LY's price list.

barcelona = code name
phenom = brand name

ya . correct . hope amd phenom release soon .

oic..thanx for info bro..get confused before this..hehe

That probably won't last when samsung buys amd

Samsung gonna buy AMD??

Barcelona = Opteron
Agena = Phenom

TheINQ says Samsuuuung + DAAAAAAAMIIITTT is UUUUUUUNSTOOOOPPPPABLE for Intel to handle.
If it really is done however, at minimal:
They'll be in the black with Samsung doing so well with electronics, and possibly Fusion has a better chance of growing into a robust GPGPU platform.