AMD Phenom II X6 1055T & 1090T Reviews

[grimREEFER]

It does remind me of fermi reviews, because the way you qualify the benches can make all the difference.
imo phenom II 1060 is better than the entry level core i7s.
one is clearly better in a multithreaded environment, and thats where things are, or are heading.
i also believe that a gtx 470 is better than a 5870.
one is better with dx11 and tessellation, and thats where graphics is headed.

in both cases, if you weigh more heavily the benches that are more old-school, you can get vastly different conclusions.

[zerazax]

Where things are headed and where things are now are going to have different requirements. But, from a market standpoint, CPUs last a lot longer than a GPU, so I'm not sure the same analogy applies the same way

[kl0012]

A good review from ixbtlabs (as usual). Large number of tests:
http://ixbtlabs.com/articles3/cpu/am...-1090t-p1.html

[matose]

Can you check the memory and cache latency/bandwidth figures of Thuban and Deneb?

Yes, you can see here. Memory performance on Thuban is better, this is why in multi-threaded applications the cores work better together.

[onethreehill]

Sooo.... the review on BSN* doesn't count because...?

I am not sure i understand what you are saying

[ajaidev]

-- Intel launched i7-980X a month or so ago, but they slapped on way more the 50%

I am wondering if Intel may not accelerate their plans to produce the i7-970 at a ~300 buck price point sooner rather than later, rumor mill has it pegged at Q3 -- that is a long time to let AMD woooooo people over to their platform.

They cant accelerate it its going on as planned, it will not be around $300 price point at all.

Also intel will respond to the Thuban threat with lynfields not bloomflieds.

[massman]

Yes, you can see here. Memory performance on Thuban is better, this is why in multi-threaded applications the cores work better together.

Yes, but that's just the memory. What about L1/L2/L3 cache latency?

[BoredByLife]

Yes, but that's just the memory. What about L1/L2/L3 cache latency?

Should be the same or better normally.

[Ghostbuster]

I really doubt that.... it does not work the way you are thinking.

Games still can hardly use all 6 cores. I think he meant going the way which Intel did with ClarksField, using SMT. Using 3 cores with SMT would have reduced the die size, lower the TDP and can go for higher clock speed as well as higher Turbo.

Not worth what hype? The extra cores do what they are supposed to in multi-threaded scenarios and the turbo works great. This is a nice jump over Phenom II and a great upgrade path for AM2+ and AM3 owners. Can you point me toward something Thuban didn't deliver on?

This caught my eye -> http://www.tomshardware.com/reviews/...fx,2613-9.html

[Manicdan]

Games still can hardly use all 6 cores. I think he meant going the way which Intel did with ClarksField, using SMT. Using 3 cores with SMT would have reduced the die size, lower the TDP and can go for higher clock speed as well as higher Turbo.

This caught my eye -> http://www.tomshardware.com/reviews/...fx,2613-9.html

i dont think if amd went the SMT route, would automatically get higher clocks. and lets face the fact its a little too late to rethink phenom, lol, its only 2 something years old, might as well be a senior citizen in the tech industry

and the turbo working on only 3 cores is very limiting in certain games. i think we need to see a good list of duel core and quad optimized games and compare them to a 955 (because the 955 and 1090T both run at 3.2 stock) so we should see a very obvious increase, or near identical.

the good news is AMD Overdrive is able to let the user set how many cores can turbo (im 90% sure of this) so people can simply set it to 4 cores, deal with 15 more watts of heat, and see 400-500 more mhz in any game.

i would have really preferred in AMD did let 4 cores use turbo in x6, and leave it at 2 cores for x4 chips. 3 overclocked cores might as well have been left at 2, and get an extra 100mhz out of it, cause honestly how many applications take advantage of 3, nothing i can think of realistically. even if amd let 4 cores turbo, the TDP would be higher, but i dont think it would have been high enough to break the 125W TDP they set, unless if 6 cores at stock was atleast 115W.

[matose]

Yes, but that's just the memory. What about L1/L2/L3 cache latency?

I guess L1 or L2 is slower but L3 is faster hence the better multi-core performance. How do I check the actual latency?

[Lightman]

i dont think if amd went the SMT route, would automatically get higher clocks. and lets face the fact its a little too late to rethink phenom, lol, its only 2 something years old, might as well be a senior citizen in the tech industry

and the turbo working on only 3 cores is very limiting in certain games. i think we need to see a good list of duel core and quad optimized games and compare them to a 955 (because the 955 and 1090T both run at 3.2 stock) so we should see a very obvious increase, or near identical.

Yes, 3 Turbo cores, but other 3 doesn't need to go to low clock!!
I'm just running with 2 cores crunching AOD stability test + normal system usage and typing this text:
2 cores crunching at 100% utilization are Turboed to 17x multi (3.4GHz)
4 cores are running mild tasks (utilization on average 30% per core with spikes to 100% and lows of 3%) and the multi is 14x (stock 3.2GHz)!

the good news is AMD Overdrive is able to let the user set how many cores can turbo (im 90% sure of this) so people can simply set it to 4 cores, deal with 15 more watts of heat, and see 400-500 more mhz in any game.

This is true

i would have really preferred in AMD did let 4 cores use turbo in x6, and leave it at 2 cores for x4 chips. 3 overclocked cores might as well have been left at 2, and get an extra 100mhz out of it, cause honestly how many applications take advantage of 3, nothing i can think of realistically. even if amd let 4 cores turbo, the TDP would be higher, but i dont think it would have been high enough to break the 125W TDP they set, unless if 6 cores at stock was atleast 115W.

With BE CPU there is no problem because you can bypass these limits.
Besides I'm sure once AMD launches 32nm cores we will have power gating and what you're describing will be implemented

[Leopold Butters]

Turbo can be turned off right?

[richierich]

Yes, turbo can be disabled in BIOS.

[Manicdan]

Yes, 3 Turbo cores, but other 3 doesn't need to go to low clock!!
I'm just running with 2 cores crunching AOD stability test + normal system usage and typing this text:
2 cores crunching at 100% utilization are Turboed to 17x multi (3.4GHz)
4 cores are running mild tasks (utilization on average 30% per core with spikes to 100% and lows of 3%) and the multi is 14x (stock 3.2GHz)!

ok well that goes against every bit of info ive seen, everything else has shown that turbo only works if idle cores are detected,

you should check your total pc power consumption with turbo off all 6 cores maxed out at stock, and then test with turbo on, see what the cores try to do, chances are your still under the expected power envelope with 2 on turbo, and 4 doing their own thing.

[Leopold Butters]

Yes, turbo can be disabled in BIOS.

How do you like your 1055? I ordered a 1090, can't wait to try it out.

[richierich]

lovin it.

[Chumbucket843]

Yes, but that's just the memory. What about L1/L2/L3 cache latency?

the more entries you add to cache the more searching you have to do. a bigger L3 is slower but gets more cache hits. my guess would be that L1 and L2 are the same unless they did some tweaks.