AMD Phenom II x6 1090T/1055T reviews..

[sirmonkey1985]

anandtech posted a review of the 1090T and the 1055T with the 1090T beating the i7 920 on a few benchmarks and beating the i7 860 on a few as well.. so now my question is when will F@H add support for the phenom II x6?

Let's clarify one issue: We are talking about SMP1 -bigadv, which is on it way out the door.
SMP2 -bigadv may have different core & memory requirments.

so now we can end the SMP1 allowing the phenom II x6's to run the bigadv WU's and talk about how much PPD we think these puppies could make..

Anandtech review:
http://www.anandtech.com/show/3674/amds-sixcore-phenom-ii-x6-1090t-1055t-reviewed <-- 3.8ghz @ 1.36v stock cooling..
http://www.overclockersclub.com/reviews/amd_phenom2_x6_1090t/4.htm <-- 4.11ghz @ 1.478v..


also theres a been a few screen shots in the AMD forum section showing the 1055T(locked multiplier) running at 4.2ghz @ 1.5v and another at 4.5ghz @ 1.65v(.10 above recommended max voltage).. so i see no reason why the 1090T could not run the bigadv WU's..


ill add more review links as i find them..

 

[sdy284]

and F@H doesn't need to add support, just running the SMP client should do the trick i believe

 

[sirmonkey1985]

and F@H doesn't need to add support, just running the SMP client should do the trick i believe

if im not mistaken its coded to only support processors with 8 or more threads..

 

[SmokeRngs]

if im not mistaken its coded to only support processors with 8 or more threads..

If my memory serves me correctly, -bigadv does need 8 threads spawned, but it doesn't matter how many cores there are. It matters if you can finish the work units in time. A Q6600 overclocked high enough can run and finish -bigadv work units on time. The most you would probably need to do is force a certain number of cores in the command line of the client.

Take this with a grain of salt as I haven't looked up -bigadv stuff much since I have nothing which can run them.

 

[450]

Take this with a grain of salt as I haven't looked up -bigadv stuff much since I have nothing which can run them.

You said that a Q6600 clocked high enough can do bigadv. Your Q6600 is already at 3.6, how much higher can the Q6600 go? And how much higher does it need to go for bigadv?

 

[sirmonkey1985]

If my memory serves me correctly, -bigadv does need 8 threads spawned, but it doesn't matter how many cores there are. It matters if you can finish the work units in time. A Q6600 overclocked high enough can run and finish -bigadv work units on time. The most you would probably need to do is force a certain number of cores in the command line of the client.

Take this with a grain of salt as I haven't looked up -bigadv stuff much since I have nothing which can run them.

ahh ok.. makes sense.. so pretty much the same when we had the old 2 core vm's and being able to run 4 threads per 2 cores in each VM, if i understand that correctly..

 

[APOLLO]

You said that a Q6600 clocked high enough can do bigadv. Your Q6600 is already at 3.6, how much higher can the Q6600 go? And how much higher does it need to go for bigadv?

Super high, well over 4GHz. So high that some people have stated here that not many people managed to succeed in OC a quad that high. Moreover, stability starts becoming a major concern, and -bigadv is the most sensitive WUs released so far.

 

[SmokeRngs]

Super high, well over 4GHz. So high that some people have stated here that not many people managed to succeed in OC a quad that high. Moreover, stability starts becoming a major concern, and -bigadv is the most sensitive WUs released so far.

Yep, this is about what I've heard. I believe 4Ghz on a Q6600 is the bare minimum to finish the work unit on time and that's with absolutely no use and no downtime for the system. Q6600s were and still are great processors, but getting them to 4+Ghz generally needed extreme cooling and even then you still needed a hell of a good clocking CPU to do it. A [email protected] is already a 50% overclock and getting any extra out of that is not exactly easy.

 

[EvilAlchemist]

Let's clarify one issue: We are talking about SMP1 -bigadv, which is on it way out the door.
SMP2 -bigadv may have different core & memory requirments.

 

[10e]

SMP2 with the 6.29 binaries/A3 work units and drop-in client will support the Phenom II X6 out of the box, and the A3 work units should show upwards of 8,000 PPD. My 3.5Ghz 955 shows 6000 PPD and that's with four GPU clients running concurrently.

I have an X6 1090T on the way and will post some SMP2 A3 (60xx) work unit benchies when it finishes some WUs at stock, and then I'll try to OC it.

BigAdv checks the presence/availability of 8 cores on startup. This is an issue getting the X6 running with it. If I can figure out how to get VMware player to "trick" the LinuxRouter image to doing BigAdv by playing with the VMX file I'll also post it also. I am not sure if VMWare player will allow this, or just puke out error messages. This is only theoretical, I haven't tried it in the past.

The Core2Quad BigAdv tests were done via "sneakernet" to by-pass this internal 8-core/thread check on BigAdv clients. When the client is connected to the 'net, it will check the presence of 8-cores and when that is not available, it will switch to an A1, A2, or A3 unit automagically.

A3 work units have no problem finishing on stock Core2Quads and still get bonus.

 

[lehmann]

Anything that will run a smp client gets some bonus. My athlon2 64 @ 2.2 used to get 350ppd running two standard clients. gets close to 800ppd running smp. I would also like to know how to get bigadv running on my i5-750

 

[sirmonkey1985]

Let's clarify one issue: We are talking about SMP1 -bigadv, which is on it way out the door.
SMP2 -bigadv may have different core & memory requirments.

ahh ok didnt know they were completely getting rid of it so quick.. well then my discussions over.. so enjoy the reviews..

 

[tjmagneto]

I'm not holding my breath for the A3 bigadvs, let alone GPU3.

 

[tjmagneto]

TigerDirect has the 1055T for $149.00 after a $50.00 rebate. I'm buying, anybody else?

 

[amdgamer]

What do we mean that SMP1 -bigadv is on its way out the door? The stuff that those of us are doing in Linux(natively or in a VM), are these going to be no longer available then? If we go to new -bigadv stuff, I wonder whether those of us with Core I7 rigs will be able to continue doing them?

 

[FrEaKy]

TigerDirect has the 1055T for $149.00 after a $50.00 rebate. I'm buying, anybody else?

Already put in the order

<-- REINFORCEMENTS ON THERE WAY!

 

[sdy284]

here's my BCB

so now if the rebate goes through, i'll get my 1090 for $220

 

[jebo_4jc]

here's my BCB

[snip]

so now if the rebate goes through, i'll get my 1090 for $220

yup. I have a 1055 on its way for ~$136

 

[sirmonkey1985]

TigerDirect has the 1055T for $149.00 after a $50.00 rebate. I'm buying, anybody else?

im getting the 1090T with a 890FX motherboard.. just waiting for some better looking boards to come out.. the asus crosshair IV looks retarded and ugly as shit..

 

[Mr. Pedantic]

I still don't think I would buy one for a folding rig. The power consumption lead that Intel has over AMD is quite impressive, and I think that any cost savings the AMD chips have will be more than clawed back over the lifetime of the chip by Intel's power savings.

At least it seems the 4GHz 64-bit bug from AMD has gone, though.

 

[sirmonkey1985]

I still don't think I would buy one for a folding rig. The power consumption lead that Intel has over AMD is quite impressive, and I think that any cost savings the AMD chips have will be more than clawed back over the lifetime of the chip by Intel's power savings.

At least it seems the 4GHz 64-bit bug from AMD has gone, though.

for me its cost.. and well the fact that i just cant stand any board with a nvidia chipset on it and most of the high end intel x58 boards all have at least 1 NF200 chipset on the board.. 230-250 for a processor plus 150-160 for a motherboard and another 100 bucks for ram works out perfectly for me..

 

[jebo_4jc]

Hell, I'll change the cost analysis even more

I bought the Microcenter $99 combo with MSI 785G mobo and CPU along with 4GB of RAM for $80. I just sold the CPU for $90, bought the 1055 for ~$125 after MIR.

So I'll have a total of $210 into a six core, 4Gb system. I bet it'll be hard to beat that system in terms of PPD per dollar.

 

[Kendrak]

I find it hard to believe that the current -bigadv based on the old smp1 core will be around all that much longer.
As evil said new clients (and WU) typicaly have new system requirements.

 

[Mr. Pedantic]

for me its cost.. and well the fact that i just cant stand any board with a nvidia chipset on it and most of the high end intel x58 boards all have at least 1 NF200 chipset on the board.. 230-250 for a processor plus 150-160 for a motherboard and another 100 bucks for ram works out perfectly for me..

Most of the "lower-end" (read: cheaper) boards from most major manufacturers don't have the NF-200 or SLI support, but which have the same overclocking performance and same memory bandwidth, stability, power, etc. as much more expensive mobos. The only real difference are number of PCIe slots and SLI support.

Hell, I'll change the cost analysis even more

I bought the Microcenter $99 combo with MSI 785G mobo and CPU along with 4GB of RAM for $80. I just sold the CPU for $90, bought the 1055 for ~$125 after MIR.

So I'll have a total of $210 into a six core, 4Gb system. I bet it'll be hard to beat that system in terms of PPD per dollar.

It depends on how you think about it. Here the difference in electricity cost between the 1090T and the 860 is about $40 a year, extrapolated out 24/7.

 

[APOLLO]

What do we mean that SMP1 -bigadv is on its way out the door? The stuff that those of us are doing in Linux(natively or in a VM), are these going to be no longer available then?

The current -bigadv WUs are running on the A2 core. These will be discontinued some time in the future. The upcoming -bigadv WUs will run on the A3 core for both Linux and Windows. You can still use Linux but the A3 WUs seem to work more efficiently in Windows, and produce higher PPD. It would be better to just run these new -bigadv WUs in Windows due to the faster processing and far less overhead (no more VM). Apparently, it's supposed to be even better than native Linux.

I have experienced poor performance on my A3 SMP clients no matter what architecture and how many cores I throw at it. My systems just plain suck at A3. In fact, I need to reduce core usage to one less than the total number of cores available to see 'optimal' performance. Whereas my systems with 8 cores are roughly comparable to i7 machines in ballpark -bigadv figures, they are nearly 50% less productive in A3. Understandably, I'm quite skeptical about the transition. I have crossed fingers the move won't translate to similar performance discrepancies once it transpires.

If we go to new -bigadv stuff, I wonder whether those of us with Core I7 rigs will be able to continue doing them?

Everyone who can run -bigadv now should continue to be capable of running them. I'd wager, the newer the system architecture, the more readily it will process the unreleased A3 -biagadv WUs. No need to worry about this for the foreseeable future.

 

[jebo_4jc]

It depends on how you think about it. Here the difference in electricity cost between the 1090T and the 860 is about $40 a year, extrapolated out 24/7.

For me, the 1090T costs more like $12 per year

 

[Zink]

In the 3.2-3.4ghz range these use the same amount of power as i7's. The difference right now is bigadv @ 25k ppd vs smp @ 13k? ppd.

 

[Nathan_P]

In the 3.2-3.4ghz range these use the same amount of power as i7's. The difference right now is bigadv @ 25k ppd vs smp @ 13k? ppd.

That maybe but if you already have an AMD rig that will take the chip its a no brainer. For a new build it does get a bit more complicated. If you can't run -bigadv for any reason do you go for i7 or X6???

 

[APOLLO]

That maybe but if you already have an AMD rig that will take the chip its a no brainer. For a new build it does get a bit more complicated. If you can't run -bigadv for any reason do you go for i7 or X6???

For a new build I still think i7 is superior if one intends to run -bigadv. So far, I haven't heard a definitive yes that Thuban can run -bigadv, let alone make the bonus with substantial amount of credit. If one already has a system ready to accommodate an X6, then there's no question a chip upgrade is the logical course to take.

 

[jebo_4jc]

particularly if the existing chip can be sold

 

[amdgamer]

The current -bigadv WUs are running on the A2 core. These will be discontinued some time in the future. The upcoming -bigadv WUs will run on the A3 core for both Linux and Windows. You can still use Linux but the A3 WUs seem to work more efficiently in Windows, and produce higher PPD. It would be better to just run these new -bigadv WUs in Windows due to the faster processing and far less overhead (no more VM). Apparently, it's supposed to be even better than native Linux.

I have experienced poor performance on my A3 SMP clients no matter what architecture and how many cores I throw at it. My systems just plain suck at A3. In fact, I need to reduce core usage to one less than the total number of cores available to see 'optimal' performance. Whereas my systems with 8 cores are roughly comparable to i7 machines in ballpark -bigadv figures, they are nearly 50% less productive in A3. Understandably, I'm quite skeptical about the transition. I have crossed fingers the move won't translate to similar performance discrepancies once it transpires.

Everyone who can run -bigadv now should continue to be capable of running them. I'd wager, the newer the system architecture, the more readily it will process the unreleased A3 -biagadv WUs. No need to worry about this for the foreseeable future.

Ahh okay, I guess I probably shouldn't have anything to worry about then. I just got into doing -bigadv's, so I was a little worried. For my purposes, bringing -bigadv over to the Windows SMP client would be awesome since I would no longer have to run a VM. That alone should cut down on some of the resource overhead.

Is 'Windows just not very good with dual processor systems like yours, or is it the actual A3 core itself?

 

[sirmonkey1985]

Most of the "lower-end" (read: cheaper) boards from most major manufacturers don't have the NF-200 or SLI support, but which have the same overclocking performance and same memory bandwidth, stability, power, etc. as much more expensive mobos. The only real difference are number of PCIe slots and SLI support.


It depends on how you think about it. Here the difference in electricity cost between the 1090T and the 860 is about $40 a year, extrapolated out 24/7.

yeah i understand that.. but if im going to build a new system im not going with a low budget motherboard on that platform.. i build my shit for overclocking and longevity.. i dont plan to upgrade from a 1090T any time soon unless the bulldozer is like OMGWTFBBQ faster then the x6.. but the biggest thing out of all of this is that the 890 chipset supports up to 6 sata 6Gbps drives.. while theres maybe 3 or 4 intel boards that support up to 2 sata 6Gbps..

 

[SmokeRngs]

Is 'Windows just not very good with dual processor systems like yours, or is it the actual A3 core itself?

For whatever reason, the A3 core doesn't seem to run as efficiently under Linux as it does under Windows. On my Linux server, the A3 core work units were doing no better than A2 core work units which was about 6200PPD. Certain A3 work units would get up to 6800PPD on occasion, though.

I was already running a WinXP VM on that machine for something and decided to do a little testing. I raised the amount of RAM for the VM to 1 or 1.5 gig and set it to 4 cores instead of 1 and installed the WinSMP client in the VM. That machine is currently doing 7400PPD on a P6023 work unit.

I don't know what the difference is, but the A3 core does not run anywhere near as well under Linux as it does under Windows. Before this, the A1 core ran a little better under Linux than Windows and obviously the A2 core only ran under Linux and did a hell of a lot better than the A1 core.

 

[APOLLO]

Is 'Windows just not very good with dual processor systems like yours, or is it the actual A3 core itself?

Probably the former reason, not A3 itself. I run many S-771 systems which are dual socket Core2 architecture. Few people on this team seem to be running these systems. Actually, very few people are, and I believe I have more of these than anyone else here who may be running them. This is one of the reasons why I often see different client 'effects' than a lot of people, and why sometimes I need to do things differently than others.

The implementation of Intel's Core2 architecture in multiple sockets was a semi-successful iteration from my experience. Some would argue it just didn't meet performance expectations in many areas. S-771 was largely hampered by an archaic memory subsystem that prevented the high inherent capability of Core2 processors from ever seeing their full potential. And, this could very well be the reason why my A3 performance is lackluster.

(As an arguable side note, up until the release of Nehalem, the memory subsystem was one area AMD still outshone Intel, albeit with increasingly less realized advantages in actual performance, as benchmarks at that time bore out).

 

[EvilAlchemist]

I'm not holding my breath for the A3 bigadvs, let alone GPU3.

What do we mean that SMP1 -bigadv is on its way out the door? The stuff that those of us are doing in Linux(natively or in a VM), are these going to be no longer available then? If we go to new -bigadv stuff, I wonder whether those of us with Core I7 rigs will be able to continue doing them?

F@H Beta Testing Rules don't let me comment on times of release or hardware required.

I have given warning earlier in this post. The official SMP1 -bigadv hardware suggestion is for 8 pyhsical cores - always has been.

 

[Mr. Pedantic]

yeah i understand that.. but if im going to build a new system im not going with a low budget motherboard on that platform.. i build my shit for overclocking and longevity.. i dont plan to upgrade from a 1090T any time soon unless the bulldozer is like OMGWTFBBQ faster then the x6.. but the biggest thing out of all of this is that the 890 chipset supports up to 6 sata 6Gbps drives.. while theres maybe 3 or 4 intel boards that support up to 2 sata 6Gbps..

But because everything is integrated onto the CPU the motherboard becomes much less of a bottleneck; the CPU is much more important in determining how far you'll get to OC.

 

[EvilAlchemist]

I am curious how these CPUs will preform on the Standard SMP2 A3 cores.

 

[Spun Ducky]

So I am wondering about joining this folding madness. I currently have the urge to get a bit of a toy and I have been looking at firearms but I've already got so many of those I figure maybe put it towards a different item. I am looking at possibly getting 3 1055's systems for folding or maybe 1 intel 980gx. So that is my question what would be more bang for the buck as I am a coder so occasionally I would use the comp(s) for tests but for the most part they would be dedicated to folding. If you guys have any other suggestions I am willing to listen as well, but I just have that hexacore bug I think others are getting right now.

 

[Kendrak]

So I am wondering about joining this folding madness. I currently have the urge to get a bit of a toy and I have been looking at firearms but I've already got so many of those I figure maybe put it towards a different item. I am looking at possibly getting 3 1055's systems for folding or maybe 1 intel 980gx. So that is my question what would be more bang for the buck as I am a coder so occasionally I would use the comp(s) for tests but for the most part they would be dedicated to folding. If you guys have any other suggestions I am willing to listen as well, but I just have that hexacore bug I think others are getting right now.

The trend has fewer bowen with more power centered in a single bomputer.

Stanford has changed it's point structure to reflect this. Also with new work units on the horrison I would want to have as much power as possible (in a single box) to throw at them to get the best bonus.

Go with the Intel hex core.

 

[jebo_4jc]

The trend has fewer bowen with more power centered in a single bomputer.

Stanford has changed it's point structure to reflect this. Also with new work units on the horrison I would want to have as much power as possible (in a single box) to throw at them to get the best bonus.

Go with the Intel hex core.

correct

Largely because of power requirements.

Three 1055 sytems would draw roughly 3x as much power as one 980x system and if you get a good OC going on the 980x system they would produce roughly the same amount of points per day. The 1055 systems might have a slight advantage in production, but it wouldn't be anywhere near 3x. Also, having more PPD coming from one system is easier to maintain.