Intel's 'Haswell' CPU refresh expected early next month alongside 9-series chipset

[TheLastPanda]

If PCI-E 3.0 and 4.0 ever become valid "must have" features it will likely be because of storage hardware advances.

That is more or less what I thought. Thanks for backing it up with the info though, I will have a look that later today

 

[Jad Chaar]

I'm waiting on the Haswell-E series and the X99 Chipset to come out so I can grab one of those 8-Core chips at the 500 dollar price point.

Haswell-E is gonna be amazing.

 

[theBest11778]

This shows the titan black performing about 25% better. assuming the 680 is at 4100 and the titan is at 5200 which is my closest guess. A card at least double in the same benchmark would be over 8000 which is a mid 50's% jump over the flagship single gpu (not even considered to be a consumer level product). You said "GPUs" as well implying that not just the one on the very top of the 800 line will manage this. Even assuming that does happen plus all of the following: the bottleneck is currently right at pcie3 x4, bandwidth scales linearly with performance, cpu's will support more lanes on that socket eventually before an upgrade is neeed, and that more lanes will be needed for new devices..... then why is pcie3 x16 not good enough for consumer grade computing? Its a lot of assumptions to make to get a few fps from bleeding edge hardware that may need it years down the road....by the time Nvidias new socket may be in use already.

1.) In that same TechpowerUp article they did hit the wall/bottleneck with PCIe 1.0 at 8x, PCIe 2.0 at 4x. So, the bottleneck is there. It's just not a huge issue yet because most people don't use their PCIe lanes for anything other than their GPU at the moment. Perhaps a video capture card/RAID card/or a Sound Card, but those aren't commonly used as most motherboards deliver decent performance on those fronts how. You can see where using those cards and a high power GPU on a PCIe 1.0 bus wouldn't work well at all, and running out of room on PCIe 2.0.

The Titan Black is far more than 25% better than a GTX 680. I'm using a GTX 760 right now, overclocked performs about the same as my old GTX 680 (give or take 5%,) but its EXACTLY 1/2 as powerful as my GTX 780 OC (1/2 framerate in games, and 1/2 in synthetic benchmarks.) The Titan Black with a decent OC should be 2x as fast as the GTX 680 easily.

 

[dividebyzero]

The Titan Black is far more than 25% better than a GTX 680. I'm using a GTX 760 right now, overclocked performs about the same as my old GTX 680 (give or take 5%,) but its EXACTLY 1/2 as powerful as my GTX 780 OC (1/2 framerate in games, and 1/2 in synthetic benchmarks.) The Titan Black with a decent OC should be 2x as fast as the GTX 680 easily.

Very easily. For example, a GTX 680 will post around 30 f.p.s. in BF4 @ 2560x1440/1600 (max game menu settings), a Titan Black on the other hand

That is the measure of two cards that are limited by GPU performance or thermal throttling*. You could find instances where the GTX 680 is closer ( possible CPU limitation) or much more lacking, but the latter is usually a product of 680's lack of internal bandwidth and framebuffer. Crysis 3 for instance will happily consume 3+GB of framebuffer (and 4+GB at 5760x1080) when 8 x MSAA is applied. Notice the drop off for a 256-bit/2GB card

You could cripple virtually any 2 or 3GB card to a virtual standstill in relation to the Titan Black with enough antialiasing and resolution applied, but it becomes an artificial construct in many cases for a single card, since the game would be unplayable on either- amply demonstrated with Crysis 3 "just" 4xMSAA.

*Just as an aside, Gigabyte are bundling a newer (600W capacity?) Windforce3 cooler with their OC'ed Titan Black (13% OC base/boost), which should allow for better overclocking than the reference cooler thus far used.

 

[Guest]

My Dell 486DX33 running DOS5.2 still owns

 

[Lionvibez]

My Dell 486DX33 running DOS5.2 still owns

lol the only thing that 486 owns is a calculator

 

[TheLastPanda]

Now the real questions are A) why do the 3dmark vantage scores on nvidias own website make the titan black look like such a small bump up, and B) how does the titan in the benchmark you linked get such a lead on the 290x in bf4? The benches on this site show a healthy lead. Maybe with new drivers. Im also guessing they watercooled the titan for the OC tests.

 

[dividebyzero]

Now the real questions are A) why do the 3dmark vantage scores on nvidias own website make the titan black look like such a small bump up,

Possible differences in hardware used? drivers? Without the info and links I'd be guessing.

and B) how does the titan in the benchmark you linked get such a lead on the 290x in bf4?

Probably tested both the 290X and Titan using the DirectX 11 render path would be my guess. Most sites usually use the Mantle path for AMD cards when benching BF4 (and multiplayer which benefits the Nvidia card). As you can see from this recent Tech Report article, there is quite a difference in performance between DX11 and Mantle. Add in the fact that the AMD card would close the performance gap at 2560x1440 under DirectX from the 1920x1080 shown and the numbers seem about right albeit skewed without Mantle results.

Im also guessing they watercooled the titan for the OC tests.

Open air bench for the Digital Storm article (+200 GPU clock offset)- the bit-tech review hit 1119 base (+220 offset)/1210 boost/1262 peak on air. The Black overclocks just as well as the 780 Ti (as has been reported by users in general) and rev.2 GTX 780 (all are B1 silicon). Even multi SLI setups reach fairly good levels of stable OC on air, which isn't a given with any SKU.

 

[TheLastPanda]

Probably tested both the 290X and Titan using the DirectX 11 render path would be my guess. Most sites usually use the Mantle path for AMD cards when benching BF4 (and multiplayer which benefits the Nvidia card). As you can see from this recent Tech Report article, there is quite a difference in performance between DX11 and Mantle.

I will look in to the multiplayer issue. I didn't hear that it favored nvidia, and I'm surprised they benchmarked on multiplayer at all given how variable it is.

Something still seems off. I know that mantle boosts the amd side in certain cases, but the original bench was made before mantle the was implemented. https://www.techspot.com/review/734-battlefield-4-benchmarks/page4.html shows a decent lead at 1600p with dx11. To think that the gap could be closed by drivers and overclocking on air makes sense, but I am hesitant to believe they flipped the tables entirely.

 

[dividebyzero]

I will look in to the multiplayer issue. I didn't hear that it favored nvidia, and I'm surprised they benchmarked on multiplayer at all given how variable it is

Most sites bench MP since it has a higher relevance to the majority of readers

Something still seems off. I know that mantle boosts the amd side in certain cases, but the original bench was made before mantle the was implemented. https://www.techspot.com/review/734-battlefield-4-benchmarks/page4.html shows a decent lead at 1600p with dx11

Which pretty much evaporated with Nvidia's updated drivers to a degree. A lot also depends on the level of AO allied with multisampling is used. It also depends upon whether you're looking for playable framerates, or absolutes. Note how the relative performance changes in this Hardwareluxx review. I'd also note (as have others) that AMD's driver DX path seems to be regressing at the expense of Mantle

In DirectX 11 mode, the NVIDIA GeForce GTX 780 Ti turned in a 13.4% performance increase over the Radeon R9 290X. The Radeon R9 290X did not seem to perform as well as we have recorded in DirectX 11 mode during prior evaluations.

Anyhow, this is supposed to be a thread about Intel's Haswell refresh, so if you're worried about a few percentage points difference in graphics cards reviews I'd suggest a new thread.