What do you think of AMD's Thuban

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

DXcellence718

Distinguished
Mar 22, 2010
35
0
18,530
You guys probably already heard the news, but now that amd's six core processor is coming out what are your thoughts. Personally im not really impressed, from what i've heard its only equivalent to an i7 quad but cost under $300. It is still the cheapest hex core a consumer can get, so i guess there still good for price/performance ratio(for a six core processor anyways)
 
Solution
If we can take into account MC vs Xeon (24 real cores vs 12 real/12 fake) than I am not too impressed. That means that Core i7 will probably be on par if not better than Thuban, hence why the highest end one is priced near a Core i7 920.

While 6 real cores is a nice thing to hear, if its performance is meh its not impressive. If it can't beat a quad that has 4 real/4 fake cores than its just not what I would call impressive.

My bet is that the only thing AMD will have that will truly beat Core i7 on a core per core and clock per clock basis is Bulldozer.


Agreed, price really tells the story here. Personally due to the performance advantage I7/I5 has over Phenom II, I doubt the X6 will even consistently out perform an I7 920 even in well threaded applications. It's true adding 2 more real cores is better than hyperthreading, but I7 still has a much better architecture.
 


Quite true, which is why I said "up to". IIRC this figure was from Intel's marketing material, although don't ask me for a linky 😛.
 

I can tell you that apps like Handbrake and RipBot264 LOVE HT and 30% is correct...at least with my 920. If I shut it off, it takes about 30-35% longer to encode a BR to MKV then it does with it on. I haven't benched anything else so I can't speak for games, but encoding...I'm still heavily debating a 980x. Nice little tax break for my home business and I seriously think I could finally encode a BR at high quality in under an hour. Oh that would make my day. Mmmm....12 threads. LOL
 
Awesome... so THAT'S how they did it... interesting. I guess AMD thought that would be a better move than having everyone work on 32nm so they could release the 6 core on time with Intels... Cool stuff.
 


Look it like this, everyone wants to go to 32nm so they have less heat, less power consumption, etc. If they can make this happen by adding special material to there 45nm and the resault is the same (or almost the same*), why not?
 
FYI, both the THG & AT articles on Thuban (kinda ridiculous, and just free AMD advertising since no CPUs have actually been released yet), call AMD's Turbo Core "less elegant" than Turbo Boost :kaola: . So there you have it folks - AMD is for the unsophisticated, bottom-feeding, "El Cheapo" people in sweaty T-shirts and copious amounts of B.O. 😀

In fact, "AMD" is short for "Abstains froM Deodorant" :bounce:
 


Wow, you just can't say anything good about AMD. I mean, I'm not saying this is the Holy Grail or anything, but they DID manage to put 6 cores into the same TDP as their quads, with only a ~200Mhz decrease in core speed.

And everyone knew their Turbo would be "less elegant." But a 500Mhz increase is nice, any way you spin it. :bounce:
 
AMD and Intel took different approaches with the Turbo feature so it isn't comparing apples to apples. I would imagine the Intle approach would be more beneficial for mainstream desktop computing and the AMD approach would be more beneficial for apps that make use of more than 2 cores. It isnt that complicated.
 


Will you just stop this nonsense. You're just as bad as the AMD loons. Its not free advertising. Its a tech site reporting on the upcoming tech.
 


LOL - relax, I was just poking fun at the AMD fanbois, much like Jenny does with Intel fanbois 😀..

 


True AMD put 2 more cores with their quads on the same process node but Intel put s more cores at 65nm within the same TPD envelope with its C2Qs. So its not that amazing. And if I remember correctly, I don't think Intel used a new low-k dielectric to do that.

Either way, what they did is not what needs to impress. Its the performance that does.
 


Yes, of course my post was nonsense, just like the "Thuban - world's fastest desktop CPU", "Turbo is just a benchmarketing gimmick", etc remarks here & other related threads. This thread disintegrated long before my post, in case anybody here was hoping to actually obtain any useful information. C'mon, seriously - after all the ad hominem posts above, you expect to learn anything new here?

And yes, these threads could be much more useful, if the intent of the TOS was observed by everybody participating. However if people routinely post flamebait stuff they know is untrue or is mere wishful thinking, then what's the point of playing by the rules?

What I'd like to know is why Thuban is still under NDA and thus no benchies available?
 


If AMD had spent the design effort using an extra million transistors like what Intel's power-gating required, they likely could have gotten Thuban up past 3.4GHz turbo capability since that's what the 965 does with 4 cores. Instead, they just down-clock the unused cores & jack the voltage up. Really it's just a quick & dirty "brute-force" solution.

IMO, AMD is just trotting out mostly placeholder CPU stuff while spending most of their efforts on Bulldozer, where they hope to recapture the perf crown. MC is sandwiched by Westmere & Beckton, haven't heard much on the mobile front from them, and DT is ever so predictable - just a small GHz boost or # cores boost, using the same old K8 core tweaked a bit here & there..

And it's notable that AMD now has to add other tricks such as low-k dielectric to their SOI. IIRC one benefit of SOI was supposed to be lower gate-to-substrate capacitance due to the isolation, as well as lowered leakage current. So when scaling down it appears that SOI's benefits become less important compared to inter-electrode & other parasitic capacitances.

 
I'm not saying it's impressive, but progress is good. You have to give them a LITTLE credit, they've now made the only native 45nm hex-core!!! Wooooo!

Though, I don't think the performance will be too impressive to anyone. I'd say the 3Gz will be roughly = to the 930 in apps that use 6+ threads. For games it'll perform like 955, and the 965, and the 945, and the X3 720, and all the other good gaming processors.
 

Intel does have the only native 45nm Octal core in Beckton. 😀
 



Many Intel fans like to point out that Turbo generally always works and bumps the speed up one step even with all cores working.

So we have to ask: If there is rarely a time when the cores are running at base clock speed then what motivation would anybody ever have to claim base clock speed for any purpose other than trolling on forums or skewing benchmark results for marketing purposes?

Pointing out that simple fact is NOT nonsense as you are claiming. What is nonsense is people who know that base clock means nothing continually quoting it as a value that is somehow relevant in any conversation or benchmark. (Or even worse nonsense: Knowingly quoting base clock when you know the chip is running a single threaded application and would be running at a speed step even higher.)
 


Yea and AMD had the first naitive 65nm quad core (Phenom) but look how that turned out.... low clock speed, high power, high heat and low performance. Oh and no real OCing until the B3 stepping and even then it barley pushed as far as Phenom II does stock.

As I said, its not what they do that needs to impress. Its the performance. And the price says it all. It will stay near the Core i7 920. AMD knows this hence why their Black Edition wont be $1K.


Sahme shizz, different story. Its still the same MCM setup that AMD laughed, prodded and said Intel was stupid naitive will be better. AMD is now eating their own words. They are using it because if they cannot get more than a 6 core CPU out to the server market, they will lose horribly.

So now if Intel did a MCM Westmere setup runnin on its 2x QPI, would that make it even more elegant than MC since QPI is much faster than HTT? Nah. It would be the same thing. A way to put more cores on one chip.

AMD laughed at Intel. They claimed naitive at 65nm would be better. They were wrong. Now they are doing what they claimed was crap.

its funny to watch other companies eat their words. Just like we all knew Intel would one day go with a IMC, AMD would need a MCM.
 
Status
Not open for further replies.