Nehalem lonely competitor?

[vizering]

Nehalem is the king of the hill right now. What does AMD have up it's sleeves to compete?

I'm going to get an AM3 board and CPU, I'd just like to know what I can look forward to in the near future.

 

[JAYDEEJOHN]

If you wait til september, maybe a 3.6 cpu, stock.

 

[vizering]

From AMD?

 

[JAYDEEJOHN]

Yes

 

[vizering]

Might as well wait then. Can you give me a link with info on this new AMD CPU?

 

[JAYDEEJOHN]

It was leaked by a tester from another site. Theres no link. He is reliable. Everyones been waiting for the new stepping, this is probably it, as everyone assumed itd be befoe the end of the year

 

[JAYDEEJOHN]

The 965, at 3.4 is coming, but this is to be 3.6 or higher, but most likely, 3.6

 

[vizering]

Found [this]. August 13th for the 3.4 965.

 

[JAYDEEJOHN]

Yep. Thats the latest. But 1 to follow quickly

 

[xsamitt]

Sounds good.Thanks for the info.

 

[megamanx00]

AMD has been promising to raise the IPC of the Phenom II with new stepping. I don't know on how many of those promises they will manage to deliver. For gaming perhaps the most important tweak will be the promised increase in floating point performance. Honestly I'm surprised they haven't been making a high-k version of their CPUs as I'm sure that Global Foundries will be offering it.

 

[jimmysmitty]

AMD has been promising to raise the IPC of the Phenom II with new stepping. I don't know on how many of those promises they will manage to deliver. For gaming perhaps the most important tweak will be the promised increase in floating point performance. Honestly I'm surprised they haven't been making a high-k version of their CPUs as I'm sure that Global Foundries will be offering it.

The GF is just now getting up and running. It needs time to get everything in order plus they have yet to be able to get HK/MG working. Last report I saw, IBM was having touble with it and I doubt they will just get it down pat. Intel was working on it for quite along time (since before Core 2 came out) so they have the lead there.

We will have to see if anything they "promise" will come. I don't take it as a promise though because that just leads to being let down. I prefer to think of it as things they are going to try to do. But if they can improve the IPC they just might be able to beat out C2Q and give Core i5 a run for its money.

 

[JAYDEEJOHN]

The 45nm version of HKMG probably wont happen thos its superior even to Intels, its too late to be working on it for this (45nm) node, well see it on 32nm

 

[apache_lives]

3.6 amd meanwhile Intel isnt sitting around either - no doubt something coming from them soon

 

[apache_lives]

actually just found this http://www.hexus.net/content/item.php?item=19571

 

[JAYDEEJOHN]

h yes, hex cores are coming from both sides, but that depends on usage. Im thinking hex cores with smt are similar to when quads first came out. Almost nothing used them, and most things dont use quads even yet.
It depends on upgrade time and usage. If you plan on keeping your rig for 2+ years, itd be good to at least look into hex cores, but currently not practical for most usage

 

[jj463rd]

More cores sound great for Flight Simulator X though.

 

[JAYDEEJOHN]

Not sure about FSX. If its optimized beyond 4 cores or not, since SMT is apparently useless in gaming

 

[softwarefiend]

Yeah it's pretty S-H-I-T if you ask me and im a parfessional.

 

[jj463rd]

So far FSX can use up to 32 cores but it's capable of using 256 if there was a patch for it.

 

[fazers_on_stun]

The GF is just now getting up and running. It needs time to get everything in order plus they have yet to be able to get HK/MG working. Last report I saw, IBM was having touble with it and I doubt they will just get it down pat. Intel was working on it for quite along time (since before Core 2 came out) so they have the lead there.

We will have to see if anything they "promise" will come. I don't take it as a promise though because that just leads to being let down. I prefer to think of it as things they are going to try to do. But if they can improve the IPC they just might be able to beat out C2Q and give Core i5 a run for its money.

IBM's approach is gate-first HKMG, Intel's is gate-last, so whatever Intel learned probably won't be useable by IBM/AMD. In part, the 1000 degree C annealing process means IBM had to go with a different mix - Intel's wouldn't stand the heat so to speak.

IIRC games are primarily integer code, so FP improvement (or SSE, etc) wouldn't help much except maybe on some game threads such as physics...

Finally, I doubt AMD could do much for IPC without a substantially different core architecture. You can do some things like throw more cache or high-speed, low latency memory access to reduce bottlenecks, but internal to the core would require far more than just a new stepping.

Finally finally , one reason why the Core2 march has higher IPC is that it has 4-issue decoders, instead of the 3 that AMD's K8, K10, K10.5, etc use. There are other advantages such as better OoO execution, too, although K10 has made significantly progress there I believe, compared to K8.

 

[JAYDEEJOHN]

Theres a few things AMD is planning on doing to raise IPC on current arch, but being the fiend I am, Im going to make you , fazers, wade thru all the AMD material to find it heheh

 

[vizering]

For those like me who wish to build a new rig coming this Fall 2009, which company would be best to stick with for a gamer with a $700-$800 budget? Also which has greater long term possibilities?

 

[Helloworld_98]

^ neither on that budget, AM3 is going soon and 1156 won't see any hexa-cores or octo-cores.

the only socket that you can buy this year which has a long term upgrade plan is 1366, as you get a Quad core now, then a hexa-core in H1 2010, then an octo core from Q4 2010 and after.

 

[vizering]

Aren't the i7's being "phased out"? I've no idea what that means but it don't sound to good.

Someone also told me I'd be smart to stick with AMD AM3 board since AMD still uses that and I could upgrade it well into the future.