Sign in with
Sign up | Sign in
Your question

AMD Turion 64 X2 Performance Preview

Last response: in CPUs
Share
May 18, 2006 8:51:04 PM

HardwareZone has some interesting results for the upcoming Turion X2 processor:

Quote:
From the rather small pool of results gathered, the AMD Turion 64 X2 processor seems to garner very positive results. When you take into consideration that what we had time to hands-on was a mere TL-52 processor model and not the top of the line version, I'm sure you can take a guess of what the top performing model has in store and we are very eager to find that out when the possibility arises.

As long as the Turion 64 X2 lives up to it TDP spec and manages power even better than its predecessor, we are pretty sure that the processor combined with the forthcoming mobile platform update from NVIDIA and ATI can really propel AMD's Turion 64 X2 Mobile Technology into strong consumer favor. Hopefully, the Turion 64 X2 Mobile Technology would be a bigger sleeper hit than the first generation Turion 64. If it does, AMD has created yet another winner, but a new first for the mobile segment. While Intel's Core 2 Duo processors are slated to be out in the third quarter of this year, AMD's next revision of mobile processor would be out to play catch then and probably give Intel a good run-in because AMD's next revision of the Turion 64 X2 would be based on the 65nm silicon process technology. Rumored to have a similar TDP as the current processors, you can probably imagine that it's going to be a lot smaller and faster even when nothing else is propped up. While this is up for argument couple of months away, stay tuned for the near timeline when we assess a full retail Turion 64 X2 Mobile Technology notebook for the complete lowdown.


The MSI notebook was configured with just one memory module which forced the machine to run in single channel mode. Still, the results were impressive.

http://www.hardwarezone.com/articles/view.php?cid=1&id=1904
May 18, 2006 9:14:48 PM

yeah, very interesting.. here is another benchies.

m@tbe.com


very interesting Core Duo Yonah.

lolz.. =)
May 19, 2006 12:49:10 AM

Let's wait till more reviews come out. I'd like Anand or techreport to becnh the chip with DDR2-667 running in dual channel. :wink:
Related resources
May 19, 2006 1:11:27 AM

I think the consensus is going to be that it's not as fast as core duo, or as battery efficient (yet), but the price difference will still make it a good option.
May 19, 2006 1:17:06 AM

Ahhh ! Caps gone, good riddance !

'Twas about time...
May 19, 2006 1:40:04 AM

LOL
Just be happy the WHOLE thing wasn't in caps.
May 19, 2006 1:44:12 AM

If you put it that way, the Turion X2 should lose to the Core Duo on all counts. From the m@tbe review, the 2Ghz Core Duo seems faster on average than the Turion X2 at 2GHz while giving 18% more battery life. On the price front, Intel is set to drop prices of the Core Duo processors on May 28th. By that time the 2Ghz Core Duo will only cost $294 compared to the Turion X2s $354.

http://www.hkepc.com/bbs/news.php?tid=600608

This would mean that Core Duo is both faster, uses less power, and is cheaper than the Turion X2 at the same clock speed. The price cuts were done in order to make room for the 2.33GHz Core Duo, which along with the 2.16GHz Core Duo should both outperform AMD's highest mobile offering.

Personally, I don't see the Turion X2 regaining it's footing until a 65nm version is released. AMD's 90nm SOI process may be very good but it's now exhausted itself. All you have to do is look at the Turion X2's TDP ratings. They actually tell you that the 1.6GHz models use 31W, the 1.8Ghz model 33W and the 2GHz model 35W. Generally, companies tell you the maximum TDP for the series with the models falling below that so that you don't know how much scaling room is left. In this case, AMD is making it blantly obvious that they couldn't fit the 1.8GHz model in a 31W TDP, and that they couldn't fit the 2GHz model in the 33W TDP. This means that their scaling room on the 90nm SOI process is virtually exhausted. M@tbe actually points out that power issues were one of the reasons that the Turion X2 platform was delayed. I believe that original plans had a 2.2GHz Turion X2 at the 35W TDP for launch, but The Inquirer reported that postponed a while ago. Power problems are probably also the reason why the L2 cache size has dropped to 256k and 512k per core instead of the 1MB and 512k sizes of the single cores although to be fair Core Duo also halved its equivalent per core cache amount compared to Dothan although Merom will add it back.

The thing is that it's unlikely we will see 65nm Turion X2s this year. AMD has already said that the first 65nm chips won't arrive until December and even then it's only low end desktop X2 parts until yields pick up. This would give Intel a lot of time and freedom in the mobile market, which obviously isn't the ideal position.
May 19, 2006 2:22:39 AM

Yeah, I noticed that too, but since the someone else already linked the m@tbe review I didn't bother commenting. The benchmarks that 9-inch linked were also completely synthetic so they aren't that representative.

Quote:
Note that the CPU's L2 cache is solely for each processor's use and it isn't shared like in the Intel Core Duo design. While the latter's advantage is a larger shared cache storage accessible by either core, the nature of the shared cache design does not allow both processors to access the core simultaneously and they have to arbitrate (take turns) to do so via a shared bus interface. The result? Higher cache latencies. This is not a problem of the Turion 64 X2 processor with each core's dedicated cache. Simpler in design with reduced latency penalties, the only limitation is reduced total processor cache storage since it is not a unified shared cache. Depending on the workload types, both AMD's and Intel's processors have their advantages in this design aspect.

I couldn't help but notice the irony of this statement. They think that the Turion X2 will have lower latency L2 cache because it's not shared, but in reality that isn't true. The Core Duo's L2 latency is only 14 cyles and if memory serves, the Turion X2 like other K8 processors should have an average L2 latency of 17 cycles. This means that despite it's shared architecture, the default sleep mode, and the larger 2MB size compared to 512k on the Turion X2, the Core Duo still has lower L2 latency. What is even more impressive are reports that part of the reason why the L2 latency went up from 10 cycles in Dothan to 14 cycles in Yonah is to already plan for Core since Core supposedly still maintains that low 14 cycle latency despite the cache size being increased to 4MB.
May 19, 2006 5:57:10 AM

nice processor.
May 19, 2006 4:16:21 PM

The Core Duo and Core Single are the first Hybrid Mobile/Desktop chips. The Turion X2 is purely mobile we can't base the lower cost X2 against the Duo. The cost points alone are so great. Even AMD said the same thing in the news confrence. Let's just wait for the data. Intel will have to cut cost to meet the cost divide AMD is using to fuel their sales. Intel may have to cut cost to regain market shares. Which is good for us.
May 19, 2006 9:50:36 PM

Quote:
Yeah, I noticed that too, but since the someone else already linked the m@tbe review I didn't bother commenting. The benchmarks that 9-inch linked were also completely synthetic so they aren't that representative.

Note that the CPU's L2 cache is solely for each processor's use and it isn't shared like in the Intel Core Duo design. While the latter's advantage is a larger shared cache storage accessible by either core, the nature of the shared cache design does not allow both processors to access the core simultaneously and they have to arbitrate (take turns) to do so via a shared bus interface. The result? Higher cache latencies. This is not a problem of the Turion 64 X2 processor with each core's dedicated cache. Simpler in design with reduced latency penalties, the only limitation is reduced total processor cache storage since it is not a unified shared cache. Depending on the workload types, both AMD's and Intel's processors have their advantages in this design aspect.

I couldn't help but notice the irony of this statement. They think that the Turion X2 will have lower latency L2 cache because it's not shared, but in reality that isn't true. The Core Duo's L2 latency is only 14 cyles and if memory serves, the Turion X2 like other K8 processors should have an average L2 latency of 17 cycles. This means that despite it's shared architecture, the default sleep mode, and the larger 2MB size compared to 512k on the Turion X2, the Core Duo still has lower L2 latency. What is even more impressive are reports that part of the reason why the L2 latency went up from 10 cycles in Dothan to 14 cycles in Yonah is to already plan for Core since Core supposedly still maintains that low 14 cycle latency despite the cache size being increased to 4MB.

If you've read the whole thing, the author of the article stated that Turion X2 cache latency has been decreased. He didn't give any numbers but we hope that some other source to fully benchmark the chip.
May 20, 2006 2:19:16 AM

Quote:
Power problems are probably also the reason why the L2 cache size has dropped to 256k and 512k per core instead of the 1MB and 512k sizes of the single cores although to be fair Core Duo also halved its equivalent per core cache amount compared to Dothan although Merom will add it back.


Though since Core Duo uses shared L2 cache, it didn't lose cache space per core except in dual[/quote] core. And the small improvements they put into the cache, like increased L2 cache bandwidth when using dual cores indicates that scaling isn't sacrificed much either.
May 20, 2006 6:28:00 PM

Among the many things I disagreed with former-member MadModMike I have to agree with him on one thing. Benchmarks don't mean squat. I have a midrange system in the other room that is supposedly an economical el cheapo PC yet it can play Doom 3, Quake 4, Half-Life 2, Call of Duty 2 on the highest ( yes that means ultra mode on Quake and Doom with max Antiso and Antial ) settings on the highest resolution the monitor can support (I think it is like 1280X1024 or something). That computer completely blows in benchmarks, yet it does excellent in real world? Tell me then, how much can we depend on benchmarks if they hardly relate to real world performance?
May 20, 2006 6:46:05 PM

Quote:
It's really context. MMM believes benches don't mean squat only when the data does not support his view of the world, in other words, if a bench were positive Intel, it don't mean squat. If it were positive AMD, it is the best way to assess the CPU :)  .

Jack
I must have misinterperated his posts then 8O . I still don't like benchmarks though. I prefer real world performance testing because that is what I do, I don't run benches all day, I play games, do work, encode videos, just to name a few. MMM claimed the same but he was obviously too biased to have any of it be true.
May 20, 2006 7:45:43 PM

Quote:
This is where we agree
Well we agree on most things so that isn't surprising. :) 
a c 100 à CPUs
May 21, 2006 2:42:32 AM

Yeah, what DID happen to MadModMike? He used to practically live in these forums, now I haven't seen him at all in weeks. Did he get banned permanently?
May 21, 2006 4:54:09 AM

Quote:
Yeah, what DID happen to MadModMike? He used to practically live in these forums, now I haven't seen him at all in weeks. Did he get banned permanently?
Yeah he likely got his I.P. permanently banned.
May 21, 2006 8:28:18 AM

I really dont see the point. It's bad enough that there was no comparison to core. Fact is though, that nobody will buy a core, turion X2 or merom chip.
They come in a package. AMD's biggest downfall has been that the packages just aren't there.
Still, if I had to shell out for a laptop, out of my own hard earned, I would probably buy the turion. That is only because the "centrino technology" carries a heavy price premium locally. For the price of a medium to high quality centrino setup, I can build a desktop that would be ultra high end.
Couple that with the fact that I would leave my head behind if it were'nt attached, and you can see why I dont have a centrino laptop.
May 21, 2006 8:42:09 AM

Quote:
I really dont see the point. It's bad enough that there was no comparison to core. Fact is though, that nobody will buy a core, turion X2 or merom chip.
They come in a package. AMD's biggest downfall has been that the packages just aren't there.
Still, if I had to shell out for a laptop, out of my own hard earned, I would probably buy the turion. That is only because the "centrino technology" carries a heavy price premium locally. For the price of a medium to high quality centrino setup, I can build a desktop that would be ultra high end.
Couple that with the fact that I would leave my head behind if it were'nt attached, and you can see why I dont have a centrino laptop.


huh!! helzz yeah you will be able to buy a core. WWID10638725
May 21, 2006 8:43:49 AM

Quote:
I really dont see the point. It's bad enough that there was no comparison to core. Fact is though, that nobody will buy a core, turion X2 or merom chip.
They come in a package. AMD's biggest downfall has been that the packages just aren't there.
Still, if I had to shell out for a laptop, out of my own hard earned, I would probably buy the turion. That is only because the "centrino technology" carries a heavy price premium locally. For the price of a medium to high quality centrino setup, I can build a desktop that would be ultra high end.
Couple that with the fact that I would leave my head behind if it were'nt attached, and you can see why I dont have a centrino laptop.


newegg
May 21, 2006 9:07:00 AM

Dont I wish!
Not available outside the U.S.A.
May 21, 2006 12:47:49 PM

Err... I can only see that youre a Fanboy willing to die for AMD :?
May 21, 2006 5:03:19 PM

Quote:
...seriously, if Core 2 comes out significantly out performing K8, then I am afraid he might do something radical.. I fear for his personal safety.

Ohhh, nooooo! Poor MMM :cry:  ....
May 21, 2006 5:28:54 PM

Quote:
Err... I can only see that youre a Fanboy willing to die for AMD :?
Another stupid Ycon post. :roll:
July 8, 2006 10:26:10 PM

Owned.

Just for 9nm:

Cons:

* Laptop runs hot
* Weak CPU with only 2x256KB L2 cache
July 8, 2006 10:51:39 PM

Quote:
Owned.

Just for 9nm:

Cons:

* Laptop runs hot
* Weak CPU with only 2x256KB L2 cache

AM why did'nt you use the owned cannon?
July 8, 2006 10:54:52 PM

Because its lame.
July 8, 2006 11:03:43 PM

Quote:
Because its lame.
Ok it's your choice.
July 8, 2006 11:13:21 PM

Ya, somewhat odd, though they gave mobility 5 and system performance 4. The lowest things were 3D performance (it has onboard and scored 2) and features which scored 3.

I guess it does what its supposed to do so yeah.
!