Folding Chit-Chat

[Queelis]

[AMD RANT/]I have to say rumoured because this seems to be of ATI's own making if I'm understanding issues correctly - it physically doesn't process the data as quickly for whatever reason.

Well that's just not true. In folding AMD cards are slower. In BOINC and Bitcoin nVidia cards are slower. It's all about the application.

Regarding the CPU, yes, it's just strange how AMD's products end up being lacklustre over and over.

[Golden Dragoon]

I think AMD's current woes are caused by the dwindling amount of alpha engineers they have left.
Back in the day the K7 Athlon was neck and neck with the P3 due to their designs, and were beating it in many situations.
Then they went on to design the A64, which was an immensely good processor, and merrily went and trounced the P4 netburst designs from Intel, now some may argue that Intels mismanagement of that era was the main cause for AMD to be "winning" but when you look at it, it took until the Core2 to equal the X2.
It is my opinion that had Intel scrapped netburst sooner and went with the core line, that AMD wouldn't have just reused and altered the same design for this long and would have been under pressure to revise the design more throughly, AMD got complacent in a way, and over that time more and more of the alpha engineers left to do other things.

[Tattysnuc]

I think AMD's current woes are caused by the dwindling amount of alpha engineers they have left.
Back in the day the K7 Athlon was neck and neck with the P3 due to their designs, and were beating it in many situations.
Then they went on to design the A64, which was an immensely good processor, and merrily went and trounced the P4 netburst designs from Intel, now some may argue that Intels mismanagement of that era was the main cause for AMD to be "winning" but when you look at it, it took until the Core2 to equal the X2.
It is my opinion that had Intel scrapped netburst sooner and went with the core line, that AMD wouldn't have just reused and altered the same design for this long and would have been under pressure to revise the design more throughly, AMD got complacent in a way, and over that time more and more of the alpha engineers left to do other things.

Sounds very plausable.

Looks like Stanfords Stat service is down. I'm still down/up-loading work units, but nothing is recording points wise. Doh! Hope nothing "goes missing"

[Golden Dragoon]

Looks like I will hit 10 mil points sometime tonight, when the stats start updating again

[scaryjim]

... It is my opinion that had Intel scrapped netburst sooner and went with the core line, that AMD wouldn't have just reused and altered the same design for this long ...

To be honest, I don't think either posit there is at all plausible - either Intel scrapping netburst / releasing Core 2 sooner, or AMD suddenly developing a killer chip in response.

If Intel had a better performing chip ready earlier they would have released it earlier - Netburst was the architecture they were able to produce chips on during that period, and it cost them in performance against Athlon 64 because they didn't predict the heat and power issues of clocking Netburst to the levels it was designed for. Intel were designing Core 2 throughout the Netburst period - but it takes a long time to get a new chip architecture right. When they were finally happy with Core 2 they released it and it was fantastic.

I think that was the real start of AMDs woes. Core 2 was simply *much* better than they anticipated. I'm pretty sure, however, that they hadn't sat on their rears all that time. As far as I know Bulldozer was already in development before Core 2 was released. Reading around though, it looks like they had a couple of false starts with the design, and had to start from a blank sheet - not a happy thought given how long it takes to get a processor design from the drawing board to production silicon...

In the meantime, they had to keep selling chips or they'd have gone out of business, and they had to keep improving their chips or their best performing chips would've been worse than Intel's worst performing chips. So they had to split their R&D budget - already smaller than Intel's - between two designs: a new architecture to make a performance leap, and enhancements to the old architecture to keep Intel at least vaguely within reach - and again they had a set-back with Phenom, which was a great idea but suffered in implementation.

At the same time I'm pretty sure that the (relative) failure of Netburst gave Intel a kick up the arse and inspired them to start continuously re-evaluating their architecture, an approach typified by the tick-tock development cycle they've adopted. So Intel have a bigger R&D budget and a more focussed R&D goal. Makes life a lot easier.

So, even if you just look at the CPU issues there's a multitude of issues, some within AMD's control and some outside of it.

Of course, AMD also did their best to sabotage their chances "off-field", as it were, with the spin off of GLOFO and the acquisition of ATI. I don't think either of those were bad calls, but it gave them about 3 - 4 years of continuous turmoil, and a whole other set of processor design issues to deal with (let alone having to work out how to function fabless and how to get good value out of the silicon, which is something they'd not had to do previously from a CPU point of view).


tl;dr? Probably wise, it's mostly inconsequential drivel

[Golden Dragoon]

The point I was making was that the loss of almost the entire team that developed the Alpha/K7/K8 processors was the major problem for AMD, and that as they had a performance advantage and weren't anticipating the resurgence of Intel they allowed themselves to get sidetracked (hence the many false starts on BD)

[Tattysnuc]

Stats service is back up and online.

There'll be a backlog of points waiting to be credited to users accounts - don't give up - with the millions of people doing this, and there being several workunits each, there's probably a substantial backlog of work for them to do, but I've never known the points/credits to NOT filter through, albeit delayed.

[Golden Dragoon]

Just a quick thought here, but with the Intel Z68 boards, is it possible to use the integrated graphics to display the desktop while the dedicated gpu folds away with no overhead?
I know that there is a switching tech on the Z68's, but from the reviews I have read it seems to switch when the dedicated gpu is loaded, which when folding would be most of the time.
If it works, what would the estimated ppd benefits be?

[Tattysnuc]

Just a quick thought here, but with the Intel Z68 boards, is it possible to use the integrated graphics to display the desktop while the dedicated gpu folds away with no overhead?
I know that there is a switching tech on the Z68's, but from the reviews I have read it seems to switch when the dedicated gpu is loaded, which when folding would be most of the time.
If it works, what would the estimated ppd benefits be?

This is the setup I have on my home media server. CPU is folding SMP units, GTX 295 is folding on both cores, and the GPU that is on the CPU Packagae (pK) is used for display purposes.

There is no discernable benefit points wise for me as the machine sits there "serving" so there's very little screen refresh going on. The advantage for me was to has a more responsive rig, becuase at time the screen would get a little corrpupt where windows hadn;t re-drawn, and it just felt a bit sluggish. Now... You wouldn't even know it was folding.

[Tattysnuc]

The points for the work I submitted during the stats service outage have now been credited...

[Terbinator]

I've managed to get some good points the last few days (now Uni is over for Christmas)

Hopefully more to come.

[Golden Dragoon]

Tatty any chance you can email me the calcs you used for folder of the month, got a little sidetracked over christmas as I had a very bad chest infection and was almost hospitalised with it so didn't think to ask you since the last time.

[Tattysnuc]

I've been sidetracked too, but with Family things. Sorry about that.

I'll send you the spreadsheet in the morning.

[Tattysnuc]

Tatty any chance you can email me the calcs you used for folder of the month, got a little sidetracked over christmas as I had a very bad chest infection and was almost hospitalised with it so didn't think to ask you since the last time.

I sent the spreadsheet through a couple of nights ago - have you had chance to look at it yet?

I'm starting to get a bit more time free up after this week - I could bring it up to date for you and record my actions. Any ideas how I could do that without having to video camera the whole session?

[Terbinator]

I sent the spreadsheet through a couple of nights ago - have you had chance to look at it yet?

I'm starting to get a bit more time free up after this week - I could bring it up to date for you and record my actions. Any ideas how I could do that without having to video camera the whole session?

Link up on TeamViewer?

[Golden Dragoon]

Took me a little while to even get it loaded in open office, getting it off of skydrive wasn't the most intuitive thing in the world lol
Have had a look through it and can kinda see what you were doing with it, but it would help an awful lot to be walked through it once.