Originally posted by sammorris:

---

Originally posted by theonejrs:

---

Sam,

Take a look at the chip lineup on your graph. Look at all the crap they show for AMD cpus. A 3.0GHz Phenom IIx4 940 BE, an obsolete chip, not even produced anymore. Then there's another obsolete, no longer produced chip in the 2.6GHz Athlon IIx4 620, and yet another in the Phenom IIx2 550! There seems to be a little bias towards Intel, but I'm sure you won't find anything wrong with the way they've done things though. All Intel chips shown are modern except for the Q9550 and the E8400. Where's the Phenom IIx4s, x3s, and x2s? Where's the 1090T BE, which is a better all around chip than the 1100T BE?

Then there's the motherboards. They chose an obsolete GIGABYTE GA-MA790GP-DS4H AM2+/AM2 AMD 790GX with DDR2 to test the AMDs, but chose a more modern P55-GD55 motherboard with DDR3 to test the Intels. But it's a fair example, right? Not by a long shot! Get rid of all the junk and give us an honest showing next time!

Russ

---

The P55-GD55 is as old as LGA1156 gets, that was one of the early boards from late 2009. On top of that, you can't test Sandy bridge CPUs in an LGA1156 board as it's a different socket!

Meanwhile, in the land of people that can actually read:

Originally posted by bit-tech:

---

Testing
To test the AMD processors we used the Asus Crosshair V Formula motherboard which is based on the AMD 990FX chipset and Socket AM3+ CPU socket.

For the Intel LGA1155 processors, we used the Asus Maximus IV Gene-Z and for the LGA1336 processors, we used the Asus Sabertooth X58. All three motherboards have proved to be excellent overclockers, and yet cost a reasonable amount, so are great options for everyday PCs running overclocked CPUs

---

Originally posted by techreport:

---

http://techreport.com/articles.x/21813/5

---

You'll notice the use of an 890GPA-UD3H for the AMDs, A Crosshair 5 Formula for the Bulldozers, a DX58SO2 for the i7 6-core, a P5E3 Premium for the Core 2s and a P7P55D-E Pro for the sandy bridge CPUs. All seems fair to me.

Originally posted by HardOCP:

---

http://hardocp.com/article/2011/10/11/amd_bulldozer_fx8150_desktop_performance_review/3

---

Here, a Crosshair 5 Formula is used for all the AMDs, and an MSI Z68A-GD65 for the sandy bridge CPUs.

Funny, not an MSI P55 or 790 chipset AMD board to be seen!

---

Originally posted by shaffaaf:

---

for the CPU argument, I used to have an e5200 at 3.5 GHz, which is a dual core chip. I then added an extra 4870 into crossfire, and in battlefield bad company 2, I saw near no increase, maybe 5 fps. I then added a q6600, clocked it to 3.4 and it damn near doubled my frame rates.

CPU bottleneck indeed.

bulldozer is onlygood in insanely multithreaded programs. So what one program you use or two, compared to the tens to hundreds that end up being worse of compared to the 1100t.

not to mention very soon sandybridge e will be coming out to further the gap, though that's a different price point.

---

Originally posted by theonejrs:

---

Sam,

First, please watch your wise mouth and respect your elders. I can read perfectly well, thank you! Please look carefully at the top of the graph!

---

Originally posted by sammorris:

---

Originally posted by shaffaaf:

---

for the CPU argument, I used to have an e5200 at 3.5 GHz, which is a dual core chip. I then added an extra 4870 into crossfire, and in battlefield bad company 2, I saw near no increase, maybe 5 fps. I then added a q6600, clocked it to 3.4 and it damn near doubled my frame rates.

CPU bottleneck indeed.

bulldozer is onlygood in insanely multithreaded programs. So what one program you use or two, compared to the tens to hundreds that end up being worse of compared to the 1100t.

not to mention very soon sandybridge e will be coming out to further the gap, though that's a different price point.

---

Same story here Shaff, when I first went crossfire with the HD4870X2 I saw almost no benefit at all with my E4300, even clocked all the way to 75% beyond its normal speed. Once I stuck a Q6600 in there, and latterly a Q9550, the frame rates jumped enormously.

Originally posted by theonejrs:

---

Sam,

First, please watch your wise mouth and respect your elders. I can read perfectly well, thank you! Please look carefully at the top of the graph!

---

Russ, with all due respect, I have to be brash sometimes to get your attention. Every discussion we have on these matters is littered with mistakes from you, and by the time we've cleared these up, the original argument is lost in pages of back and forth about who said what and who meant what. The only reason I posted that chart from GameGPU was to illustrate that some games can be extremely demanding on the CPU and it's not all just about the graphics! Did you not notice there's no Bulldozer on that chart?

---

Originally posted by AnandTech:

---

AMD also shared with us that Windows 7 isn't really all that optimized for Bulldozer. Given AMD's unique multi-core module architecture, the OS scheduler needs to know when to place threads on a single module (with shared caches) vs. on separate modules with dedicated caches. Windows 7's scheduler isn't aware of Bulldozer's architecture and as a result sort of places threads wherever it sees fit, regardless of optimal placement. Windows 8 is expected to correct this, however given the short lead time on Bulldozer reviews we weren't able to do much experimenting with Windows 8 performance on the platform. There's also the fact that Windows 8 isn't expected out until the end of next year, at which point we'll likely see an upgraded successor to Bulldozer.

---

Originally posted by sammorris:

---

Originally posted by Mr-Movies:

---

Pings are directly related to bandwidth, when you ping a network/internet address you are measuring response time between the host and you. The host is a factor as well as bandwidth.

---

Nope - I had a ping of 6ms on a 4Mbps connection, yet I know plenty of people with a ping of 25ms on 24Mbps connections. How so? ADSL Interleave, for a start. Network capacity, distance from the exchange, distance between you and the remote sever, it all adds up.
The speed of light means even at full light speed (which fibre-optic cables can't make full use of) in 1ms light can only travel 300km. If we take the assumption that fibre optic cable carries information at a third of light speed, then to travel 5000km between here and the US east coast, that'd be 50ms. As you may notice, latencies between the UK and the US are indeed, a minimum of around 60-70ms. This has nothing to do with bandwidth, you could see this with a 40Mbps pipe, or a 100Gbps pipe. It's simple physics.

---

Originally posted by sammorris:

---

All this said, it's important to take CPU speed into perspective for things like gaming. It may be a considerable step ahead of the Phenom 2s, but I recently throttled back my i5 to stock clock speed (thermal paste problem, the cooler needs redoing), and really, I practically don't notice the difference. This though, is probably because even at stock it equates to a Phenom 2 at 3.6Ghz. This then, when considered the newer i5s are 35% faster to start with and will overclock to almost 50% beyond their stock speeds, you can understand why I tell people to stick i5s in their new gaming systems, $220 isn't a lot to pay for a CPU that can do that sort of work, especially when these people are spending at least that sort of money, often more, on graphics hardware.

---

Originally posted by sammorris:

---

Your explanation of ping isn't at fault, but you referred to the term 'bandwidth' when you meant 'latency' - two different things. That's where the confusion came from.

As for the benchmarks, just because they deviate that doesn't mean they're false. This suggests variance in similar systems, which does happen. Any decent site states their testing parameters so you can tell whether they're appropriate or not.

---

Originally posted by Mr-Movies:

---

Originally posted by sammorris:

---

Your explanation of ping isn't at fault, but you referred to the term 'bandwidth' when you meant 'latency' - two different things. That's where the confusion came from.

As for the benchmarks, just because they deviate that doesn't mean they're false. This suggests variance in similar systems, which does happen. Any decent site states their testing parameters so you can tell whether they're appropriate or not.

---

Latency was implied by delay so I did cover that and bandwidth, connections, and so on still enter in. Latency is the sum of that delay measured in ms as I stated so I covered it correctly but more generally.

Benchmarks can drastically be different and although it is always better to compare on same gear, same test methods, even with different gear it shouldn't be so drastic of a difference and my point still stands. If I really wanted to wast my time I could go around and find plenty of huge differences but again I'm not going to wast my time and I doubt it would change your bias.

Believe what you want but it pays to be objective...

---

Originally posted by Estuansis:

---

Russ I don't see any bias there. They're testing CPUs at stock speeds. Also, they're testing the fastest CPUs AMD has ever made, against decidedly not Intel's best, disregarding the 990X. If anything, the test is biased heavily in AMD's favor. The AMD CPUs are just slower. This is a fact. I don't know what you keep looking for Russ. Finding the right graph won't suddenly make AMD CPUs faster.

Also, I see nothing about the boards you mentioned at all. Disregarding the fact that none of the CPUs in the shown tests will even work in the boards you mentioned. So I really don't know what angle you're going for.

I mean wtf do you keep going on about it's not a fair test? Is it not a fair test unless AMD wins? If I was shown a graph with AMD winning, common sense tells me that's the unfair test. Every single benchmark we've seen so far has shown Bulldozer getting its ass handed to it. Are you trying to say that every single reputable review site is in Intel's pockets and that's why Bulldozer sucks? Sorry Russ, but Intel bribing a review site doesn't suddenly make a CPU better or worse. If Bulldozer sucks, there's nothing a slightly skewed graph is going to do to change it.

---

Originally posted by sammorris:

---

It doesn't have anything to do with bulldozer, nor was it supposed to. It was brought up because it was claimed that Phenom 2s were adequate for modern games, as games only really needed GPU power - this graph disproves that entirely - forget which CPUs they're comparing to which, the numbers are so low, that it proves you can generate situations in games where even powerful CPUs produce painfully low numbers.

They have no reason to include the 955 over the 940 - they're both the same tech, one is just faster than the other - they just need to cover all architectures (at the time, this test came before bulldozer remember) which they have done.

---

Originally posted by sammorris:

---

Yeah I'm aware of the performance gains of the new chipsets - that only applies to this GameGPU test anyway though - the other tests, biased or otherwise, use the 990 chipset in new boards, so their results are accurate in that sense.
I hope you're joking about playing Crysis on the integrated graphics chipset...

---