AMD 7 series latest spec info from AMD round up and discussion

[555BUK]

Nobody knows how good or bad the 7 series will be, and specs that look good on paper do not always translate into real world gains. Bulldozer for example promissed much but delivered almost nothing over the previous gen AMD CPU's. Adding more cores, increasing transistor counts, and architectural changes do not always provide the goods.

Wait for real benchmarks before wetting your pants.

 

[N19h7m4r3]

Nobody knows how good or bad the 7 series will be, and specs that look good on paper do not always translate into real world gains. Bulldozer for example promissed much but delivered almost nothing over the previous gen AMD CPU's. Adding more cores, increasing transistor counts, and architectural changes do not always provide the goods.

Wait for real benchmarks before wetting your pants.

^ This.

Don't get into the hype, otherwise you might just be disappointed.

Wait for the benchmarks and then decide on what you want.

 

[drunkenmaster]

Nobody knows how good or bad the 7 series will be, and specs that look good on paper do not always translate into real world gains. Bulldozer for example promissed much but delivered almost nothing over the previous gen AMD CPU's. Adding more cores, increasing transistor counts, and architectural changes do not always provide the goods.

Wait for real benchmarks before wetting your pants.

Sorry but, this is rubbish.

I've been saying, since Bulldozer specs were revealed, that it WOULD not beat a 2600k almost at all, that its 8 cores, but its 8 slim cores and that its simply not got the instruction width/issue capacity to rival the 2600k.

Its not a bad chip at all, and its only the first step in what will likely be a 6-7 year long architecture, and it was ALWAYS going to be the worst efficiency and worst speed of all of them because the first one always is.

The only people who thought Bulldozer was somehow going to blow away the richest, highest R&D spending CPU manufacturer the planet knows, are people who couldn't see past the two numbers 4 and 8 and jumped to conclusions.

Even though for instance everyone talking about the 580/6970 never once assumed the 6970 would be almost 3 times faster than the 580gtx based on core count alone, which highlights the ridiculous nature of peoples inability to read widely available information.

Anandtech's table on maybe the first page of the review explains EXACTLY where bulldozer is slower than Thuban and Sandybridge and exactly why it can't "thrash" both based on looking at one arbitrary number OUT OF MANY MANY numbers.

http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/2

Sorry, second page, peak instruction decode rate, Thuban hexcore, 18, i7 24, Bulldozer 16.

I've been saying this since basically a day after they released info on the architecture(so have many others), around 6 months ago. Most people who can actually read more than the first number in an architecture preview, knew exactly how Bulldozer would perform.


Anyway, that just goes to show, you're completely wrong, you can indeed have an incredibly good idea how a new architecture will perform and theres not been a surprise in the eventual speed of a card since, the 9700 pro to be honest. Part of that was, that's when I started building computers, partly because ATi were so far behind before.

The 6970, I said again fully 6 months before it was due, it couldn't go over 400mm2, it would likely be 5-10% smaller, it will only be 10-15% bigger than a 5870, and its likely to come in somewhere around 15-25% faster.

580gtx performance was obvious compared to a 285gtx and completely predictable, as was Bulldozer performance, as is Ivybridge. Trinity and Piledriver in general, less so, mostly because we don't know where AMD will tweak, we will hear about the improvements long before launch and we WILL know roughly how it will perform long before launch.

 

[kitch9]

I think I'm going to have to wait for Nvidias next gen though ... I've kind of gotten used to Physx in a few games too much

All 2 of them?

580 owner here by the way.....

 

[Farooq]

Bulldozer has about as much to do with 7 series as a intel's faulty chipset for mother boards which was a total fail had to do with sandybridge's awsomeness totally different products only thing they share is the same parent company different departments totally. It makes me chuckle when people bring one product from a company and think its all the same guys who design it. These companys have thousands of staff working on different projects at the same time please be more informed before you post nonsense it makes you sound like trolls. Have some understanding of R n D

 

[drunkenmaster]

To be fair I don't think he was saying because AMD didn't do great with Bulldozer that AMD won't do great with their GPU, just that because Bulldozer was hyped to be brilliant and wasn't is somehow proof you can't estimate performance somewhat accurately.

I was just pointing out that was very wrong, that anyone in the know and not getting sidetracked by certain members who enjoy hyping the masses up and watching them all cry about a "crap" product when it launches no where near the performance said member has convinced everyone it will have.

GPU side we have Raven for that, for MONTHS on here before anyone went OTT with performance estimates on Cayman I said its incredibly unlikely that a maximum 15% die size bump, on the SAME generation will provide much more than 20-25% more performance, then Raven weeks before launch started telling everyone his "friend" knew how fast it was and thought it would be 40% faster than a 580gtx.

This was complete rubbish, in every sense, and completely impossible, yet a month later when the 6970 launched with entirely predictable performance, half this forum somehow felt let down it wasn't 40% faster than a 580gtx.

People let themselves get carried away by users who state these things with ZERO logic to back it up.

Why did I predict 6970 performance at max 15% more die size and unlikely more than 5-10% boost from architecture after that for so long (things got a touch confusing with the shader count mis info nearer launch), because 400mm2 + was asking for trouble AMD wouldn't go near.

There were very good reasons I said what I said, and others were saying it aswell, a good estimate backed up by facts, and logic mean a lot more than random guy Y running onto a forum and saying "my friend knows all about it its X% faster than something else".

Bulldozer was even more well known as the architecture, core count, decoder, issues per core was all known 6 months before launch. It could have gone 10% further either way, it still has some to gain with better scheduling, but even that was known long before launch.

 

[bhavv]

Its completely daft to estimate or compare GPU performance to CPUs.

AMD graphics cards are still made by the same people that worked for ATI, they just work for AMD now.

 

[Farooq]

Its completely daft to estimate or compare GPU performance to CPUs.

AMD graphics cards are still made by the same people that worked for ATI, they just work for AMD now.

Couldn't have said it any better no one said what performance of these would be certinly not me. For that we wait for benchmarks. Its just annoying when a topic about one product gets sidetracked by another product that has very little to do with the first relevant topic. We dont know the perfomance and no1 is hyping the gfx side but we do know its a radical overhaul and the GPCPU and it is here to stay. ATI/AMD fusion chips are roadmapped for at least next 7 years. This is AMD first real step in that direction and that is what is exciting and what our focus should be on.

 

[Farooq]

The latest news is the 7 series will use the pci express 3 interface this comes as no suprise to many of us. Bandwidth details below for those who are unfamiliar with the changes.

Generation per lane maxmium slot bandwidth and sustained transfer rates

PCI-Express 1.1 2.5GT/sec 2GB/sec 250MB/sec 8GB/sec
PCI-Express 2.0 5GT/sec 4GB/sec 500MB/sec 16GB/sec
PCI-Express 3.0 8GT/sec 8GB/sec 1GB/sec 32GB/sec

Fear not however as pci express 2.0 boards that are 16 speed will suffer no detremental effect on performance due to insufficiant bandwidth. However AMD as made no statements regarding 8x pci express 2 performance so thats still to be seen if it wll cause significant bottleneck. Previous generation cards suffered a 2 percent hit when ran in 8x the next gen cards are to suffer a larger hit thats for sure how big is still to be seen will wait for benchmarks as always.

Detailed bitech post on changes from gen 2 to gen 3 pci express. I will copy n paste it in for you lads makes a informative read.


"A Doubling of Bandwidth...
Following the tradition that goes back to AGP (remember that?), the bandwidth is again doubled from 500MB/sec or 4Gb/sec on PCI-Express 2.0 to 1GB/sec or 8Gb/sec on PCI-Express 3.0. That's per lane, in each direction. This means the total bandwidth for an 16x PCI-Express graphics slots go up from 16GB/s to 32GB/s, so it should cope better with the future demands of high performance graphics cards.

To double the bandwidth the PCI-SIG hasn't just cranked up the transfer frequency by two though, instead it's lessened the encoding overheads to make faster transfers more efficient.

typically PCI-Express (and a whole lot of other signalling buses) use an '8b/10b' encoding, which means ten bits are transferred for eight bits (one byte) of actual data. That 20 per cent overhead is considerable, and as we've seen from CPUs, frequencies cannot perpetually increase. This means that this overhead becomes ever more an issue as the bandwith increases.

To work around this problem, PCI-Express 3.0 encodes the data in a much larger '128b/130b' chunk, and then 'a known polynomial is applied to a data stream in a feedback topology', with an 'inverse polynomial' sat at the other end to decode the data. In more human terms, this is basically a hard-coded mathematical function that is designed to evenly spread the 0s and 1s (which are electrical clock blips), so they don't interfere with each other during transport. This technique is called 'scrambling'.

This means the data transfer rate can be lower while still achieving the same real bandwidth. For example, instead of 5GT/sec of PCI-Express 2.0 increasing to 10GT/sec, PCI-E Express 3.0 only needs 8GT/sec. Less transfers mean greater efficiency - so PCI-E Express 3.0 also uses less power - and it doesn't have to use higher grade materials - so the products are cheaper to make. Win-win.

However, PCI-E Express 3.0 provides no more power than its predecesors - but that's actually a good thing. Having >300W drawn from each 16xPCI-E slot would drive up the cost of motherboards, as the copper traces on the motherboard would need to be thicker. Manufacturers might even need to add extra layers to try and route these high power traces and their associated electromagnetic interference away from sensitive data traces.

Finally, PCI-E Express 3.0 has the same physical characteristics as PCI-E Express 2.0, so it's backward compatible with previous versions of PCI-Express, regardless of the data encoding change. "