AMD Shanghai/Deneb Review Thread

[iandh]

I would say Prime95 would be the best app to prove stability.

I know that I am way outside the common accepted thinking on this site, but I don't think unnaturally stressing a componenet to see if it is stable is particularly necessary for non-critical apps.

I suppose if you were folding/crunching and didn't want to lose WU's, or if you had a comp used for competition gaming where a crash could cost you a win, then it might matter.

This is somewhat like the furmark/4870 debacle... the app put an unnatural load on the cards, something that they NEVER would see in a real life application, and it caused them to fry.

I may be 4Ghz game stable with my stock bare mosfets on my mobo, but running Prime95 for an hour may cause smoke to come out of my machine whereas if I would have just left well enough alone I would be happily gaming along.

I don't mind the existence of heavy stress tests for records and to make standards, but for my personal use I prefer medium ones rather than heavy, such as a couple hour gaming session in Crysis, or maybe a few 3dmark runs. A lack of stability will usually present itself without the need to unduly stress your components.

[qurious63ss]

I know that I am way outside the common accepted thinking on this site, but I don't think unnaturally stressing a componenet to see if it is stable is particularly necessary for non-critical apps.

I suppose if you were folding/crunching and didn't want to lose WU's, or if you had a comp used for competition gaming where a crash could cost you a win, then it might matter.

This is somewhat like the furmark/4870 debacle... the app put an unnatural load on the cards, something that they NEVER would see in a real life application, and it caused them to fry.

I may be 4Ghz game stable with my stock bare mosfets on my mobo, but running Prime95 for an hour may cause smoke to come out of my machine whereas if I would have just left well enough alone I would be happily gaming along.

I don't mind the existence of heavy stress tests for records and to make standards, but for my personal use I prefer medium ones rather than heavy, such as a couple hour gaming session in Crysis, or maybe a few 3dmark runs. A lack of stability will usually present itself without the need to unduly stress your components.

Fair enough. I personally run Prime95 because I like to push my hardware to the limit. It has nothing to do with gaming or benchmarking but to push the limits on the hardware. In other words it's a sport.

[roofsniper]

yea just as iandh said i find my comp stable whenever i can run all my programs completely stable without a problem. its nice to run prime for a little while just because it confirms that it is stable even under the worst possible conditions but i don't believe it is necessary. i recently had been using corsair dominators and on stock settings with stock voltage it was unstable. so i set it to 2.2 and it was still slightly unstable. but it was stable for every program i had just not for prime. and i ran it for that for like a year without a problem. now i don't have that issue with my ocz platinums. i can actually undervolt those.

[imsochobo]

my phenom on SB600, does things nicely, its surely a nice 700 mhz running stock voltage, but unfortuantly, dfi have been late on good bios for 790gx (LP JR edition) and the 790FX m2r. =/

[iandh]

yea just as iandh said i find my comp stable whenever i can run all my programs completely stable without a problem. its nice to run prime for a little while just because it confirms that it is stable even under the worst possible conditions but i don't believe it is necessary. i recently had been using corsair dominators and on stock settings with stock voltage it was unstable. so i set it to 2.2 and it was still slightly unstable. but it was stable for every program i had just not for prime. and i ran it for that for like a year without a problem. now i don't have that issue with my ocz platinums. i can actually undervolt those.

OT

I've had meh results with Corsair for the last few years.

My G.Skill cheapo DDR2-800 are undervolted .1v and running DDR2-900 with their rated timings.

[RhoXeD]

OT

I've had meh results with Corsair for the last few years.

My G.Skill cheapo DDR2-800 are undervolted .1v and running DDR2-900 with their rated timings.

as do both of my pairs of g.skill ddr2-800

IM picking up a PH2 in the next week or so, i will be happy to finally see good scaling on the GPU side of things (my 9850BE @ 3.2 doesnt push the cards to their max)

[marten_larsson]

Got a strange idea for a review. I would like to see the power consumption for different setups over a full workday. Then compare it to most work done. It would require a workplace with at least 100 employees who will work one week with each system. To make it at bit beefier, why not do some comparisons between different uses also, like multimedia, office, CAD and such. Someone willing to do it for me?

[demonkevy666]

hey when is MONTREAL coming out ? (MCM shanghai) I've been looking around it said 09 last.

Building up Shanghai chip should be easy right now.
my only other thoughts about that is if's it shanghai are we goign to have two memory controllers ? and numa connections ?
I sure hope a phenom II 980 is that then lol

[Lightman]

hey when is MONTREAL coming out ? (MCM shanghai) I've been looking around it said 09 last.

Building up Shanghai chip should be easy right now.
my only other thoughts about that is if's it shanghai are we goign to have two memory controllers ? and numa connections ?
I sure hope a phenom II 980 is that then lol

Yes, MCM will have two DC memory controllers for Quad-Channel RDDR3. The catch is that this chip will work in new Socket G34 which is over 1900+ pads
This won't come to desktop in that form anytime soon ....

[informal]

The MCM Deneb is a 2010 thing and probably on a new socket.Instead ,we should be looking at 6 core Deneb part in Q3/Q4 of this year.That chip should fit in AM2+/AM3 socket without a problem and will stay in the same TDP envelope as current parts(maybe a 140W for a top clock 6 core part). A 6 core 3Ghz part @ 140W would be a great improvement in multithreaded scenarios since performance would scale great with the core count.My guess is that OCing will be about the same as with current parts,with a bit of added power draw from more cores(12-15W more for each added core in full load).

[imsochobo]

a bad chip hits 4300 water and 4500 water, my confirmation.

3x hdz940xcj4dgi

Max suicide dryice 4500 mhz.

Max crysis stable water 4ghz.
max water suicide: 4300 mhz.

any1 get those chips, i feel sorry for you :p i really do.

[informal]

I know that I am way outside the common accepted thinking on this site, but I don't think unnaturally stressing a componenet to see if it is stable is particularly necessary for non-critical apps.

I suppose if you were folding/crunching and didn't want to lose WU's, or if you had a comp used for competition gaming where a crash could cost you a win, then it might matter.

This is somewhat like the furmark/4870 debacle... the app put an unnatural load on the cards, something that they NEVER would see in a real life application, and it caused them to fry.

I may be 4Ghz game stable with my stock bare mosfets on my mobo, but running Prime95 for an hour may cause smoke to come out of my machine whereas if I would have just left well enough alone I would be happily gaming along.

I don't mind the existence of heavy stress tests for records and to make standards, but for my personal use I prefer medium ones rather than heavy, such as a couple hour gaming session in Crysis, or maybe a few 3dmark runs. A lack of stability will usually present itself without the need to unduly stress your components.

I must say that I completely agree with your post .
Quoted for truth(and I almost never use QFT ).

PS Yeah,some time passed since the original post,but it's never too late to agree on something .

[S_G]

Yes, MCM will have two DC memory controllers for Quad-Channel RDDR3. The catch is that this chip will work in new Socket G34 which is over 1900+ pads
This won't come to desktop in that form anytime soon ....

Montreal!

[Uncle Jimbo]

yea thats what i use them for. i think many people look at it and see that they put the TRU above the TRUE and just disregard them. thanks for the info on this.

I meant to comment on FrosyTech TRU vs TRUE ratings - they used a higher velocity fan with the TRU which is why they got lower ratings. There is no question that for identical airflow, the extreme is a better cooler. They discussed fan selection in both reviews. I suppose they could have tested with a variety of fans to show the tradeoff in cooling vs noise - the one they used on the original TRU review was pretty loud (53.7 dB), while they used a much quieter fan for the TRUE review (44.5 dB). If they had used the same high velocity fan as on the TRU review, they would have gotten lower thermal numbers, but the noise performance would not have been anywhere near as good.

One advantage of the TRUE is that the added heat pipe capacity means you can handle larger total power levels - basically 50% more raw watts. But since the overall number of fins did not change between the designs, the performance at lower wattage (and for that cooler, any normal load is 'lower wattage'), will not be a 50% difference between them.

I guess you can shoot holes in any reviewer's choice of tests - but FrostyTech does a very rigorous and repeatable set of tests, and if you read carefully, you can usually extract what you need to know.

[CH33TAH]

Here is another review that I did:

http://www.benchmarkextreme.com/Arti...%20940/P1.html

Interesting, how it compares to Yorkfield clock for clock.